Merge pull request #313 from CJackHwang/dev

toolcall优化补丁

AGENTS.md

@@ -21,3 +21,9 @@ These rules apply to all agent-made changes in this repository.
 
 - Keep changes additive and tightly scoped to the requested feature or bugfix.
 - Do not mix unrelated refactors into feature PRs unless they are required to make the change pass gates.
+
+## Documentation Sync
+
+- When business logic or user-visible behavior changes, update the corresponding documentation in the same change.
+- `docs/prompt-compatibility.md` is the source-of-truth document for the “API -> pure-text web-chat context” compatibility flow.
+- If a change affects message normalization, tool prompt injection, prompt-visible tool history, file/reference handling, history split, or completion payload assembly, update `docs/prompt-compatibility.md` in the same change.

API.en.md

@@ -31,13 +31,13 @@ Docs: [Overview](README.en.md) / [Architecture](docs/ARCHITECTURE.en.md) / [Depl
 | Base URL | `http://localhost:5001` or your deployment domain |
 | Default Content-Type | `application/json` |
 | Health probes | `GET /healthz`, `GET /readyz` |
-| CORS | Enabled (`Access-Control-Allow-Origin: *`, allows `Content-Type`, `Authorization`, `X-API-Key`, `X-Ds2-Target-Account`, `X-Ds2-Source`, `X-Vercel-Protection-Bypass`) |
+| CORS | Enabled (uniformly covers `/v1/*`, `/anthropic/*`, `/v1beta/models/*`, and `/admin/*`; echoes the browser `Origin` when present, otherwise `*`; default allow-list includes `Content-Type`, `Authorization`, `X-API-Key`, `X-Ds2-Target-Account`, `X-Ds2-Source`, `X-Vercel-Protection-Bypass`, `X-Goog-Api-Key`, `Anthropic-Version`, `Anthropic-Beta`, and also accepts third-party preflight-requested headers such as `x-stainless-*`; `/v1/chat/completions` on Vercel Node Runtime matches the same behavior; internal-only `X-Ds2-Internal-Token` remains blocked) |
 
 ### 3.0 Adapter-Layer Notes
 
 - OpenAI / Claude / Gemini protocols are now mounted on one shared `chi` router tree assembled in `internal/server/router.go`.
 - Adapter responsibilities are streamlined to: **request normalization → DeepSeek invocation → protocol-shaped rendering**, reducing legacy split-logic paths.
-- Tool-calling semantics are aligned between Go and Node runtime: parsing is now centered on XML/Markup-family tool syntax (`<tool_call>` / `<function_call>` / `<invoke>` / `tool_use` / antml variants), plus stream-time anti-leak filtering.
+- Tool-calling semantics are aligned between Go and Node runtime: the only executable model-output syntax is the canonical XML tool block `<tool_calls>` → `<invoke name="...">` → `<parameter name="...">`, plus stream-time anti-leak filtering.
 - `Admin API` separates static config from runtime policy: `/admin/config*` for configuration state, `/admin/settings*` for runtime behavior.
 
 ---
@@ -160,6 +160,7 @@ Gemini-compatible clients can also send `x-goog-api-key`, `?key=`, or `?api_key=
 | DELETE | `/admin/dev/captures` | Admin | Clear local packet-capture entries |
 | GET | `/admin/chat-history` | Admin | Read server-side conversation history |
 | DELETE | `/admin/chat-history` | Admin | Clear server-side conversation history |
+| GET | `/admin/chat-history/{id}` | Admin | Read one server-side conversation entry |
 | DELETE | `/admin/chat-history/{id}` | Admin | Delete one server-side conversation entry |
 | PUT | `/admin/chat-history/settings` | Admin | Update conversation history retention limit |
 | GET | `/admin/version` | Admin | Check current version and latest Release |
@@ -194,18 +195,12 @@ No auth required. Returns the currently supported DeepSeek native model list.
 {
   "object": "list",
   "data": [
-    {"id": "deepseek-chat", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-reasoner", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-chat-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-reasoner-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-expert-chat", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-expert-reasoner", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-expert-chat-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-expert-reasoner-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-vision-chat", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-vision-reasoner", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-vision-chat-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-vision-reasoner-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []}
+    {"id": "deepseek-v4-flash", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
+    {"id": "deepseek-v4-pro", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
+    {"id": "deepseek-v4-flash-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
+    {"id": "deepseek-v4-pro-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
+    {"id": "deepseek-v4-vision", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
+    {"id": "deepseek-v4-vision-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []}
   ]
 }
 ```
@@ -221,12 +216,15 @@ For `chat` / `responses` / `embeddings`, DS2API follows a wide-input/strict-outp
 3. If still unmatched, fall back by known family heuristics (`o*`, `gpt-*`, `claude-*`, etc.).
 4. If still unmatched, return `invalid_request_error`.
 
-Current built-in default aliases (excerpt):
+Built-in aliases come from `internal/config/models.go`; `config.model_aliases` can override or add mappings at runtime. Excerpt:
 
-- OpenAI: `gpt-4o`, `gpt-4.1`, `gpt-4.1-mini`, `gpt-4.1-nano`, `gpt-5`, `gpt-5-mini`, `gpt-5-codex`
-- OpenAI reasoning: `o1`, `o1-mini`, `o3`, `o3-mini`
-- Claude: `claude-sonnet-4-5`, `claude-haiku-4-5`, `claude-opus-4-6` (plus compatibility aliases `claude-3-5-sonnet` / `claude-3-5-haiku` / `claude-3-opus`)
-- Gemini: `gemini-2.5-pro`, `gemini-2.5-flash`
+- OpenAI / Codex: `gpt-4o`, `gpt-4.1`, `gpt-5`, `gpt-5.5`, `gpt-5-codex`, `gpt-5.3-codex`, `codex-mini-latest`
+- OpenAI reasoning: `o1`, `o3`, `o3-deep-research`, `o4-mini`
+- Claude: `claude-opus-4-6`, `claude-sonnet-4-6`, `claude-haiku-4-5`, `claude-3-5-sonnet-latest`
+- Gemini: `gemini-2.5-pro`, `gemini-2.5-flash`, `gemini-pro-vision`
+- Other compatibility families: `llama-*`, `qwen-*`, `mistral-*`, and `command-*` fall back through family heuristics
+
+Retired historical families such as `claude-1.*`, `claude-2.*`, `claude-instant-*`, and `gpt-3.5*` are explicitly rejected.
 
 ### `POST /v1/chat/completions`
 
@@ -241,7 +239,7 @@ Content-Type: application/json
 
 | Field | Type | Required | Notes |
 | --- | --- | --- | --- |
-| `model` | string | ✅ | DeepSeek native models + common aliases (`gpt-5`, `gpt-5-mini`, `gpt-5-codex`, `o3`, `claude-opus-4-6`, `gemini-2.5-pro`, `gemini-2.5-flash`, etc.) |
+| `model` | string | ✅ | DeepSeek native models + common aliases (`gpt-5.5`, `gpt-5.4-mini`, `gpt-5.3-codex`, `o3`, `claude-opus-4-6`, `gemini-2.5-pro`, `gemini-2.5-flash`, etc.) |
 | `messages` | array | ✅ | OpenAI-style messages |
 | `stream` | boolean | ❌ | Default `false` |
 | `tools` | array | ❌ | Function calling schema |
@@ -254,14 +252,14 @@ Content-Type: application/json
   "id": "<chat_session_id>",
   "object": "chat.completion",
   "created": 1738400000,
-  "model": "deepseek-reasoner",
+  "model": "deepseek-v4-pro",
   "choices": [
     {
       "index": 0,
       "message": {
         "role": "assistant",
         "content": "final response",
-        "reasoning_content": "reasoning trace (reasoner models)"
+        "reasoning_content": "reasoning trace (when thinking is enabled)"
       },
       "finish_reason": "stop"
     }
@@ -296,7 +294,7 @@ data: [DONE]
 **Field notes**:
 
 - First delta includes `role: assistant`
-- `deepseek-reasoner` / `deepseek-reasoner-search` models emit `delta.reasoning_content`
+- When thinking is enabled, the stream may emit `delta.reasoning_content`
 - Text emits `delta.content`
 - Last chunk includes `finish_reason` and `usage`
 - Token counting prefers pass-through from upstream DeepSeek SSE (`accumulated_token_usage` / `token_usage`), and only falls back to local estimation when upstream usage is absent
@@ -336,7 +334,7 @@ When `tools` is present, DS2API performs anti-leak handling:
 
 Additional notes:
 
-- The parser currently follows XML/Markup-family tool payloads (`<tool_call>`, `<function_call>`, `<invoke>`, `tool_use`, antml variants). Standalone JSON `tool_calls` payloads are not treated as executable tool calls by default.
+- The parser currently treats only canonical XML tool blocks (`<tool_calls>` / `<invoke name="...">` / `<parameter name="...">`) as executable tool calls. Legacy `<tools>`, `<tool_call>`, `<tool_name>`, `<param>`, `<function_call>`, `tool_use`, antml variants, and standalone JSON `tool_calls` payloads are treated as plain text.
 - `tool_calls` shown inside fenced markdown code blocks (for example, ```json ... ```) are treated as examples, not executable calls.
 
 ---
@@ -448,17 +446,17 @@ No auth required.
 {
   "object": "list",
   "data": [
-    {"id": "claude-sonnet-4-5", "object": "model", "created": 1715635200, "owned_by": "anthropic"},
+    {"id": "claude-sonnet-4-6", "object": "model", "created": 1715635200, "owned_by": "anthropic"},
     {"id": "claude-haiku-4-5", "object": "model", "created": 1715635200, "owned_by": "anthropic"},
     {"id": "claude-opus-4-6", "object": "model", "created": 1715635200, "owned_by": "anthropic"}
   ],
   "first_id": "claude-opus-4-6",
-  "last_id": "claude-instant-1.0",
+  "last_id": "claude-3-haiku-20240307",
   "has_more": false
 }
 ```
 
-> Note: the example is partial; besides the current primary aliases, the real response also includes Claude 4.x snapshots plus historical 3.x / 2.x / 1.x IDs and common aliases.
+> Note: the example is partial; besides the current primary aliases, the real response also includes Claude 4.x snapshots plus historical 3.x IDs and common aliases.
 
 ### `POST /anthropic/v1/messages`
 
@@ -476,7 +474,7 @@ anthropic-version: 2023-06-01
 
 | Field | Type | Required | Notes |
 | --- | --- | --- | --- |
-| `model` | string | ✅ | For example `claude-sonnet-4-5` / `claude-opus-4-6` / `claude-haiku-4-5` (compatible with `claude-3-5-haiku-latest`), plus historical Claude model IDs |
+| `model` | string | ✅ | For example `claude-sonnet-4-6` / `claude-opus-4-6` / `claude-haiku-4-5` (compatible with `claude-3-5-haiku-latest`), plus historical Claude model IDs |
 | `messages` | array | ✅ | Claude-style messages |
 | `max_tokens` | number | ❌ | Auto-filled to `8192` when omitted; not strictly enforced by upstream bridge |
 | `stream` | boolean | ❌ | Default `false` |
@@ -490,7 +488,7 @@ anthropic-version: 2023-06-01
   "id": "msg_1738400000000000000",
   "type": "message",
   "role": "assistant",
-  "model": "claude-sonnet-4-5",
+  "model": "claude-sonnet-4-6",
   "content": [
     {"type": "text", "text": "response"}
   ],
@@ -544,7 +542,7 @@ data: {"type":"message_stop"}
 
 ```json
 {
-  "model": "claude-sonnet-4-5",
+  "model": "claude-sonnet-4-6",
   "messages": [
     {"role": "user", "content": "Hello"}
   ]
@@ -672,16 +670,16 @@ Returns sanitized config, including both `keys` and `api_keys`.
       "token_preview": "abcde..."
     }
   ],
-  "claude_mapping": {
-    "fast": "deepseek-chat",
-    "slow": "deepseek-reasoner"
+  "model_aliases": {
+    "claude-sonnet-4-6": "deepseek-v4-flash",
+    "claude-opus-4-6": "deepseek-v4-pro"
   }
 }
 ```
 
 ### `POST /admin/config`
 
-Only updates `keys`, `api_keys`, `accounts`, and `claude_mapping`.
+Only updates `keys`, `api_keys`, `accounts`, and `model_aliases`.
 If both `api_keys` and `keys` are sent, the structured `api_keys` entries win so `name` / `remark` metadata is preserved; `keys` remains a legacy fallback.
 
 **Request**:
@@ -696,9 +694,9 @@ If both `api_keys` and `keys` are sent, the structured `api_keys` entries win so
   "accounts": [
     {"email": "user@example.com", "password": "pwd", "token": ""}
   ],
-  "claude_mapping": {
-    "fast": "deepseek-chat",
-    "slow": "deepseek-reasoner"
+  "model_aliases": {
+    "claude-sonnet-4-6": "deepseek-v4-flash",
+    "claude-opus-4-6": "deepseek-v4-pro"
   }
 }
 ```
@@ -713,7 +711,8 @@ Reads runtime settings and status, including:
 - `compat` (`wide_input_strict_output`, `strip_reference_markers`)
 - `responses` / `embeddings`
 - `auto_delete` (`mode`: `none` / `single` / `all`; legacy `sessions=true` is still treated as `all`)
-- `claude_mapping` / `model_aliases`
+- `history_split` (`enabled` always returns `true`, `trigger_after_turns`)
+- `model_aliases`
 - `env_backed`, `needs_vercel_sync`
 - `toolcall` policy is fixed to `feature_match + high` and is no longer returned or editable via settings
 
@@ -727,7 +726,7 @@ Hot-updates runtime settings. Supported fields:
 - `responses.store_ttl_seconds`
 - `embeddings.provider`
 - `auto_delete.mode`
-- `claude_mapping`
+- `history_split.trigger_after_turns` (`history_split.enabled` is forced on globally; legacy client writes are stored as `true`)
 - `model_aliases`
 - `toolcall` policy is fixed and is no longer writable through settings
 
@@ -752,9 +751,9 @@ Imports full config with:
 
 The request can send config directly, or wrapped as `{"config": {...}, "mode":"merge"}`.
 Query params `?mode=merge` / `?mode=replace` are also supported.
-Import accepts `keys`, `api_keys`, `accounts`, `claude_mapping` / `claude_model_mapping`, `model_aliases`, `admin`, `runtime`, `responses`, `embeddings`, and `auto_delete`; legacy `toolcall` fields are ignored.
+`replace` mode replaces the full config shape while preserving Vercel sync metadata. `merge` mode merges `keys`, `api_keys`, `accounts`, and `model_aliases`, and overwrites non-empty fields under `admin`, `runtime`, `responses`, and `embeddings`. Manage `compat`, `auto_delete`, and `history_split` via `/admin/settings` or the config file; legacy `toolcall` fields are ignored.
 
-> `compat` fields are managed via `/admin/settings` or the config file; this import endpoint does not update `compat`.
+> Note: `merge` mode does not update `compat`, `auto_delete`, or `history_split`.
 
 ### `GET /admin/config/export`
 
@@ -903,7 +902,7 @@ Updates proxy binding for a specific account.
 | Field | Required | Notes |
 | --- | --- | --- |
 | `identifier` | ✅ | email / mobile / token-only synthetic id |
-| `model` | ❌ | default `deepseek-chat` |
+| `model` | ❌ | default `deepseek-v4-flash` |
 | `message` | ❌ | if empty, only session creation is tested |
 
 **Response**:
@@ -914,7 +913,7 @@ Updates proxy binding for a specific account.
   "success": true,
   "response_time": 1240,
   "message": "API test successful (session creation only)",
-  "model": "deepseek-chat",
+  "model": "deepseek-v4-flash",
   "session_count": 0,
   "config_writable": true
 }
@@ -985,7 +984,7 @@ Test API availability through the service itself.
 
 | Field | Required | Default |
 | --- | --- | --- |
-| `model` | ❌ | `deepseek-chat` |
+| `model` | ❌ | `deepseek-v4-flash` |
 | `message` | ❌ | `你好` |
 | `api_key` | ❌ | First key in config |
 
@@ -1009,7 +1008,7 @@ Common request fields:
 | --- | --- | --- | --- |
 | `message` | No | `你好` | Convenience single-turn user message |
 | `messages` | No | Auto-derived from `message` | OpenAI-style message array |
-| `model` | No | `deepseek-chat` | Target model |
+| `model` | No | `deepseek-v4-flash` | Target model |
 | `stream` | No | `true` | Recommended to keep streaming enabled so raw SSE is recorded |
 | `api_key` | No | First configured key | Business API key to use |
 | `sample_id` | No | Auto-generated | Sample directory name |
@@ -1219,7 +1218,7 @@ curl http://localhost:5001/v1/chat/completions \
   -H "Authorization: Bearer your-api-key" \
   -H "Content-Type: application/json" \
   -d '{
-    "model": "deepseek-chat",
+    "model": "deepseek-v4-flash",
     "messages": [{"role": "user", "content": "Hello"}],
     "stream": false
   }'
@@ -1232,7 +1231,7 @@ curl http://localhost:5001/v1/chat/completions \
   -H "Authorization: Bearer your-api-key" \
   -H "Content-Type: application/json" \
   -d '{
-    "model": "deepseek-reasoner",
+    "model": "deepseek-v4-pro",
     "messages": [{"role": "user", "content": "Explain quantum entanglement"}],
     "stream": true
   }'
@@ -1270,7 +1269,7 @@ curl http://localhost:5001/v1/chat/completions \
   -H "Authorization: Bearer your-api-key" \
   -H "Content-Type: application/json" \
   -d '{
-    "model": "deepseek-chat-search",
+    "model": "deepseek-v4-flash-search",
     "messages": [{"role": "user", "content": "Latest news today"}],
     "stream": true
   }'
@@ -1283,7 +1282,7 @@ curl http://localhost:5001/v1/chat/completions \
   -H "Authorization: Bearer your-api-key" \
   -H "Content-Type: application/json" \
   -d '{
-    "model": "deepseek-chat",
+    "model": "deepseek-v4-flash",
     "messages": [{"role": "user", "content": "What is the weather in Beijing?"}],
     "tools": [
       {
@@ -1344,7 +1343,7 @@ curl http://localhost:5001/anthropic/v1/messages \
   -H "Content-Type: application/json" \
   -H "anthropic-version: 2023-06-01" \
   -d '{
-    "model": "claude-sonnet-4-5",
+    "model": "claude-sonnet-4-6",
     "max_tokens": 1024,
     "messages": [{"role": "user", "content": "Hello"}]
   }'
@@ -1381,7 +1380,7 @@ curl http://localhost:5001/v1/chat/completions \
   -H "X-Ds2-Target-Account: user@example.com" \
   -H "Content-Type: application/json" \
   -d '{
-    "model": "deepseek-chat",
+    "model": "deepseek-v4-flash",
     "messages": [{"role": "user", "content": "Hello"}]
   }'
 ```

API.md

@@ -31,13 +31,13 @@
 | Base URL | `http://localhost:5001` 或你的部署域名 |
 | 默认 Content-Type | `application/json` |
 | 健康检查 | `GET /healthz`、`GET /readyz` |
-| CORS | 已启用（`Access-Control-Allow-Origin: *`，允许 `Content-Type`, `Authorization`, `X-API-Key`, `X-Ds2-Target-Account`, `X-Ds2-Source`, `X-Vercel-Protection-Bypass`） |
+| CORS | 已启用（统一覆盖 `/v1/*`、`/anthropic/*`、`/v1beta/models/*`、`/admin/*`；浏览器有 `Origin` 时回显该 Origin，否则为 `*`；默认允许 `Content-Type`, `Authorization`, `X-API-Key`, `X-Ds2-Target-Account`, `X-Ds2-Source`, `X-Vercel-Protection-Bypass`, `X-Goog-Api-Key`, `Anthropic-Version`, `Anthropic-Beta`，并会放行预检里声明的第三方请求头，如 `x-stainless-*`；Vercel 上 `/v1/chat/completions` 的 Node Runtime 也对齐相同行为；内部专用头 `X-Ds2-Internal-Token` 仍被拦截） |
 
 ### 3.0 接口适配层说明
 
 - OpenAI / Claude / Gemini 三套协议已统一挂在同一 `chi` 路由树上，由 `internal/server/router.go` 负责装配。
 - 适配器层职责收敛为：**请求归一化 → DeepSeek 调用 → 协议形态渲染**，减少历史版本中“同能力多处实现”的分叉。
-- Tool Calling 的解析策略在 Go 与 Node Runtime 间保持一致：当前以 XML/Markup 家族解析为主（含 `<tool_call>` / `<function_call>` / `<invoke>` / `tool_use` / antml 变体），并在流式场景执行防泄漏筛分。
+- Tool Calling 的解析策略在 Go 与 Node Runtime 间保持一致：当前唯一可执行的模型输出语法是 canonical XML 工具块 `<tool_calls>` → `<invoke name="...">` → `<parameter name="...">`，并在流式场景执行防泄漏筛分。
 - `Admin API` 将配置与运行时策略分开：`/admin/config*` 管静态配置，`/admin/settings*` 管运行时行为。
 
 ---
@@ -160,6 +160,7 @@ Gemini 兼容客户端还可以使用 `x-goog-api-key`、`?key=` 或 `?api_key=`
 | DELETE | `/admin/dev/captures` | Admin | 清空本地抓包记录 |
 | GET | `/admin/chat-history` | Admin | 查看服务器端对话记录 |
 | DELETE | `/admin/chat-history` | Admin | 清空服务器端对话记录 |
+| GET | `/admin/chat-history/{id}` | Admin | 查看单条服务器端对话记录 |
 | DELETE | `/admin/chat-history/{id}` | Admin | 删除单条服务器端对话记录 |
 | PUT | `/admin/chat-history/settings` | Admin | 更新对话记录保留条数 |
 | GET | `/admin/version` | Admin | 查询当前版本与最新 Release |
@@ -194,18 +195,12 @@ Gemini 兼容客户端还可以使用 `x-goog-api-key`、`?key=` 或 `?api_key=`
 {
   "object": "list",
   "data": [
-    {"id": "deepseek-chat", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-reasoner", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-chat-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-reasoner-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-expert-chat", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-expert-reasoner", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-expert-chat-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-expert-reasoner-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-vision-chat", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-vision-reasoner", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-vision-chat-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
-    {"id": "deepseek-vision-reasoner-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []}
+    {"id": "deepseek-v4-flash", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
+    {"id": "deepseek-v4-pro", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
+    {"id": "deepseek-v4-flash-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
+    {"id": "deepseek-v4-pro-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
+    {"id": "deepseek-v4-vision", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
+    {"id": "deepseek-v4-vision-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []}
   ]
 }
 ```
@@ -221,12 +216,15 @@ Gemini 兼容客户端还可以使用 `x-goog-api-key`、`?key=` 或 `?api_key=`
 3. 未命中时按模型家族规则回退（如 `o*`、`gpt-*`、`claude-*`）。
 4. 仍未命中则返回 `invalid_request_error`。
 
-当前内置默认 alias（节选）：
+当前内置默认 alias 来自 `internal/config/models.go`，`config.model_aliases` 会在运行时覆盖或补充同名映射。节选：
 
-- OpenAI：`gpt-4o`、`gpt-4.1`、`gpt-4.1-mini`、`gpt-4.1-nano`、`gpt-5`、`gpt-5-mini`、`gpt-5-codex`
-- OpenAI Reasoning：`o1`、`o1-mini`、`o3`、`o3-mini`
-- Claude：`claude-sonnet-4-5`、`claude-haiku-4-5`、`claude-opus-4-6`（及 `claude-3-5-sonnet` / `claude-3-5-haiku` / `claude-3-opus` 兼容别名）
-- Gemini：`gemini-2.5-pro`、`gemini-2.5-flash`
+- OpenAI / Codex：`gpt-4o`、`gpt-4.1`、`gpt-5`、`gpt-5.5`、`gpt-5-codex`、`gpt-5.3-codex`、`codex-mini-latest`
+- OpenAI reasoning：`o1`、`o3`、`o3-deep-research`、`o4-mini`
+- Claude：`claude-opus-4-6`、`claude-sonnet-4-6`、`claude-haiku-4-5`、`claude-3-5-sonnet-latest`
+- Gemini：`gemini-2.5-pro`、`gemini-2.5-flash`、`gemini-pro-vision`
+- 其他兼容族：`llama-*`、`qwen-*`、`mistral-*`、`command-*` 会按家族启发式回退
+
+退役历史模型（如 `claude-1.*`、`claude-2.*`、`claude-instant-*`、`gpt-3.5*`）会被显式拒绝。
 
 ### `POST /v1/chat/completions`
 
@@ -241,7 +239,7 @@ Content-Type: application/json
 
 | 字段 | 类型 | 必填 | 说明 |
 | --- | --- | --- | --- |
-| `model` | string | ✅ | 支持 DeepSeek 原生模型 + 常见 alias（如 `gpt-5`、`gpt-5-mini`、`gpt-5-codex`、`o3`、`claude-opus-4-6`、`gemini-2.5-pro`、`gemini-2.5-flash` 等） |
+| `model` | string | ✅ | 支持 DeepSeek 原生模型 + 常见 alias（如 `gpt-5.5`、`gpt-5.4-mini`、`gpt-5.3-codex`、`o3`、`claude-opus-4-6`、`claude-sonnet-4-6`、`gemini-2.5-pro`、`gemini-2.5-flash` 等） |
 | `messages` | array | ✅ | OpenAI 风格消息数组 |
 | `stream` | boolean | ❌ | 默认 `false` |
 | `tools` | array | ❌ | Function Calling 定义 |
@@ -254,14 +252,14 @@ Content-Type: application/json
   "id": "<chat_session_id>",
   "object": "chat.completion",
   "created": 1738400000,
-  "model": "deepseek-reasoner",
+  "model": "deepseek-v4-pro",
   "choices": [
     {
       "index": 0,
       "message": {
         "role": "assistant",
         "content": "最终回复",
-        "reasoning_content": "思考内容（reasoner 模型）"
+        "reasoning_content": "思考内容（开启 thinking 时）"
       },
       "finish_reason": "stop"
     }
@@ -296,7 +294,7 @@ data: [DONE]
 **字段说明**：
 
 - 首个 delta 包含 `role: assistant`
-- `deepseek-reasoner` / `deepseek-reasoner-search` 模型输出 `delta.reasoning_content`
+- 开启 thinking 时会输出 `delta.reasoning_content`
 - 普通文本输出 `delta.content`
 - 最后一段包含 `finish_reason` 和 `usage`
 - token 计数优先透传上游 DeepSeek SSE（如 `accumulated_token_usage` / `token_usage`）；仅在上游缺失时回退本地估算
@@ -337,7 +335,7 @@ data: [DONE]
 补充说明：
 
 - **非代码块上下文**下，工具负载即使与普通文本混合，也会按特征识别并产出可执行 tool call（前后普通文本仍可透传）。
-- 解析器当前走 XML/Markup 家族（包含 `<tool_call>`、`<function_call>`、`<invoke>`、`tool_use`、antml 风格）；纯 JSON `tool_calls` 片段默认不会直接作为可执行调用解析。
+- 解析器当前只把 canonical XML 工具块（`<tool_calls>` / `<invoke name="...">` / `<parameter name="...">`）作为可执行调用解析；旧式 `<tools>`、`<tool_call>`、`<tool_name>`、`<param>`、`<function_call>`、`tool_use`、antml 风格与纯 JSON `tool_calls` 片段默认都会按普通文本处理。
 - Markdown fenced code block（例如 ```json ... ```）中的 `tool_calls` 仅视为示例文本，不会被执行。
 
 ---
@@ -449,17 +447,17 @@ data: [DONE]
 {
   "object": "list",
   "data": [
-    {"id": "claude-sonnet-4-5", "object": "model", "created": 1715635200, "owned_by": "anthropic"},
+    {"id": "claude-sonnet-4-6", "object": "model", "created": 1715635200, "owned_by": "anthropic"},
     {"id": "claude-haiku-4-5", "object": "model", "created": 1715635200, "owned_by": "anthropic"},
     {"id": "claude-opus-4-6", "object": "model", "created": 1715635200, "owned_by": "anthropic"}
   ],
   "first_id": "claude-opus-4-6",
-  "last_id": "claude-instant-1.0",
+  "last_id": "claude-3-haiku-20240307",
   "has_more": false
 }
 ```
 
-> 说明：示例仅展示部分模型；实际返回除当前主别名外，还包含 Claude 4.x snapshots，以及 3.x / 2.x / 1.x 历史模型 ID 与常见别名。
+> 说明：示例仅展示部分模型；实际返回除当前主别名外，还包含 Claude 4.x snapshots，以及 3.x 历史模型 ID 与常见别名。
 
 ### `POST /anthropic/v1/messages`
 
@@ -477,7 +475,7 @@ anthropic-version: 2023-06-01
 
 | 字段 | 类型 | 必填 | 说明 |
 | --- | --- | --- | --- |
-| `model` | string | ✅ | 例如 `claude-sonnet-4-5` / `claude-opus-4-6` / `claude-haiku-4-5`（兼容 `claude-3-5-haiku-latest`），并支持历史 Claude 模型 ID |
+| `model` | string | ✅ | 例如 `claude-sonnet-4-6` / `claude-opus-4-6` / `claude-haiku-4-5`（兼容 `claude-sonnet-4-5`、`claude-3-5-haiku-latest`），并支持历史 Claude 模型 ID |
 | `messages` | array | ✅ | Claude 风格消息数组 |
 | `max_tokens` | number | ❌ | 缺省自动补 `8192`；当前实现不会硬性截断上游输出 |
 | `stream` | boolean | ❌ | 默认 `false` |
@@ -491,7 +489,7 @@ anthropic-version: 2023-06-01
   "id": "msg_1738400000000000000",
   "type": "message",
   "role": "assistant",
-  "model": "claude-sonnet-4-5",
+  "model": "claude-sonnet-4-6",
   "content": [
     {"type": "text", "text": "回复内容"}
   ],
@@ -545,7 +543,7 @@ data: {"type":"message_stop"}
 
 ```json
 {
-  "model": "claude-sonnet-4-5",
+  "model": "claude-sonnet-4-6",
   "messages": [
     {"role": "user", "content": "你好"}
   ]
@@ -673,16 +671,16 @@ data: {"type":"message_stop"}
       "token_preview": "abcde..."
     }
   ],
-  "claude_mapping": {
-    "fast": "deepseek-chat",
-    "slow": "deepseek-reasoner"
+  "model_aliases": {
+    "claude-sonnet-4-6": "deepseek-v4-flash",
+    "claude-opus-4-6": "deepseek-v4-pro"
   }
 }
 ```
 
 ### `POST /admin/config`
 
-只更新 `keys`、`api_keys`、`accounts`、`claude_mapping`。
+只更新 `keys`、`api_keys`、`accounts`、`model_aliases`。
 如果同时发送 `api_keys` 与 `keys`，优先保留 `api_keys` 中的结构化 `name` / `remark`；`keys` 仅作为旧格式兼容回退。
 
 **请求**：
@@ -697,9 +695,9 @@ data: {"type":"message_stop"}
   "accounts": [
     {"email": "user@example.com", "password": "pwd", "token": ""}
   ],
-  "claude_mapping": {
-    "fast": "deepseek-chat",
-    "slow": "deepseek-reasoner"
+  "model_aliases": {
+    "claude-sonnet-4-6": "deepseek-v4-flash",
+    "claude-opus-4-6": "deepseek-v4-pro"
   }
 }
 ```
@@ -714,7 +712,8 @@ data: {"type":"message_stop"}
 - `compat`（`wide_input_strict_output`、`strip_reference_markers`）
 - `responses` / `embeddings`
 - `auto_delete`（`mode`：`none` / `single` / `all`；旧配置 `sessions=true` 仍按 `all` 处理）
-- `claude_mapping` / `model_aliases`
+- `history_split`（`enabled` 固定返回 `true`、`trigger_after_turns`）
+- `model_aliases`
 - `env_backed`、`needs_vercel_sync`
 - `toolcall` 策略已固定为 `feature_match + high`，不再通过 settings 返回或修改
 
@@ -728,7 +727,7 @@ data: {"type":"message_stop"}
 - `responses.store_ttl_seconds`
 - `embeddings.provider`
 - `auto_delete.mode`
-- `claude_mapping`
+- `history_split.trigger_after_turns`（`history_split.enabled` 已全局强制开启；旧客户端传入时会被保存为 `true`）
 - `model_aliases`
 - `toolcall` 策略已固定，不再作为可写入字段
 
@@ -753,9 +752,9 @@ data: {"type":"message_stop"}
 
 请求可直接传配置对象，或使用 `{"config": {...}, "mode":"merge"}` 包裹格式。
 也支持在查询参数里传 `?mode=merge` / `?mode=replace`。
-导入时会接受 `keys`、`api_keys`、`accounts`、`claude_mapping` / `claude_model_mapping`、`model_aliases`、`admin`、`runtime`、`responses`、`embeddings`、`auto_delete` 等字段；`toolcall` 相关字段会被忽略。
+`replace` 模式会按完整配置结构替换（保留 Vercel 同步元信息）；`merge` 模式会合并 `keys`、`api_keys`、`accounts`、`model_aliases`，并覆盖 `admin`、`runtime`、`responses`、`embeddings` 中的非空字段。`compat`、`auto_delete`、`history_split` 建议通过 `/admin/settings` 或配置文件管理；`toolcall` 相关字段会被忽略。
 
-> `compat` 相关字段请通过 `/admin/settings` 或配置文件管理；该导入接口不会更新 `compat`。
+> 注意：`merge` 模式不会更新 `compat`、`auto_delete`、`history_split`。
 
 ### `GET /admin/config/export`
 
@@ -907,7 +906,7 @@ data: {"type":"message_stop"}
 | 字段 | 必填 | 说明 |
 | --- | --- | --- |
 | `identifier` | ✅ | email / mobile / token-only 合成标识 |
-| `model` | ❌ | 默认 `deepseek-chat` |
+| `model` | ❌ | 默认 `deepseek-v4-flash` |
 | `message` | ❌ | 空字符串时仅测试会话创建 |
 
 **响应**：
@@ -918,7 +917,7 @@ data: {"type":"message_stop"}
   "success": true,
   "response_time": 1240,
   "message": "API 测试成功（仅会话创建）",
-  "model": "deepseek-chat",
+  "model": "deepseek-v4-flash",
   "session_count": 0,
   "config_writable": true
 }
@@ -988,7 +987,7 @@ data: {"type":"message_stop"}
 
 | 字段 | 必填 | 默认值 |
 | --- | --- | --- |
-| `model` | ❌ | `deepseek-chat` |
+| `model` | ❌ | `deepseek-v4-flash` |
 | `message` | ❌ | `你好` |
 | `api_key` | ❌ | 配置中第一个 key |
 
@@ -1012,7 +1011,7 @@ data: {"type":"message_stop"}
 | --- | --- | --- | --- |
 | `message` | 否 | `你好` | 便捷单轮用户消息 |
 | `messages` | 否 | 自动由 `message` 生成 | OpenAI 风格消息数组 |
-| `model` | 否 | `deepseek-chat` | 目标模型 |
+| `model` | 否 | `deepseek-v4-flash` | 目标模型 |
 | `stream` | 否 | `true` | 建议保留流式，以记录原始 SSE |
 | `api_key` | 否 | 配置中第一个 key | 调用业务接口使用的 key |
 | `sample_id` | 否 | 自动生成 | 样本目录名 |
@@ -1222,7 +1221,7 @@ curl http://localhost:5001/v1/chat/completions \
   -H "Authorization: Bearer your-api-key" \
   -H "Content-Type: application/json" \
   -d '{
-    "model": "deepseek-chat",
+    "model": "deepseek-v4-flash",
     "messages": [{"role": "user", "content": "你好"}],
     "stream": false
   }'
@@ -1235,7 +1234,7 @@ curl http://localhost:5001/v1/chat/completions \
   -H "Authorization: Bearer your-api-key" \
   -H "Content-Type: application/json" \
   -d '{
-    "model": "deepseek-reasoner",
+    "model": "deepseek-v4-pro",
     "messages": [{"role": "user", "content": "解释一下量子纠缠"}],
     "stream": true
   }'
@@ -1248,7 +1247,7 @@ curl http://localhost:5001/v1/responses \
   -H "Authorization: Bearer your-api-key" \
   -H "Content-Type: application/json" \
   -d '{
-    "model": "gpt-5-codex",
+    "model": "gpt-5.3-codex",
     "input": "写一个 golang 的 hello world",
     "stream": true
   }'
@@ -1273,7 +1272,7 @@ curl http://localhost:5001/v1/chat/completions \
   -H "Authorization: Bearer your-api-key" \
   -H "Content-Type: application/json" \
   -d '{
-    "model": "deepseek-chat-search",
+    "model": "deepseek-v4-flash-search",
     "messages": [{"role": "user", "content": "今天的新闻"}],
     "stream": true
   }'
@@ -1286,7 +1285,7 @@ curl http://localhost:5001/v1/chat/completions \
   -H "Authorization: Bearer your-api-key" \
   -H "Content-Type: application/json" \
   -d '{
-    "model": "deepseek-chat",
+    "model": "deepseek-v4-flash",
     "messages": [{"role": "user", "content": "北京今天天气怎么样？"}],
     "tools": [
       {
@@ -1347,7 +1346,7 @@ curl http://localhost:5001/anthropic/v1/messages \
   -H "Content-Type: application/json" \
   -H "anthropic-version: 2023-06-01" \
   -d '{
-    "model": "claude-sonnet-4-5",
+    "model": "claude-sonnet-4-6",
     "max_tokens": 1024,
     "messages": [{"role": "user", "content": "你好"}]
   }'
@@ -1384,7 +1383,7 @@ curl http://localhost:5001/v1/chat/completions \
   -H "X-Ds2-Target-Account: user@example.com" \
   -H "Content-Type: application/json" \
   -d '{
-    "model": "deepseek-chat",
+    "model": "deepseek-v4-flash",
     "messages": [{"role": "user", "content": "你好"}]
   }'
 ```

CODE_OF_CONDUCT.md

@@ -0,0 +1,128 @@
+# Contributor Covenant Code of Conduct
+
+## Our Pledge
+
+We as members, contributors, and leaders pledge to make participation in our
+community a harassment-free experience for everyone, regardless of age, body
+size, visible or invisible disability, ethnicity, sex characteristics, gender
+identity and expression, level of experience, education, socio-economic status,
+nationality, personal appearance, race, religion, or sexual identity
+and orientation.
+
+We pledge to act and interact in ways that contribute to an open, welcoming,
+diverse, inclusive, and healthy community.
+
+## Our Standards
+
+Examples of behavior that contributes to a positive environment for our
+community include:
+
+* Demonstrating empathy and kindness toward other people
+* Being respectful of differing opinions, viewpoints, and experiences
+* Giving and gracefully accepting constructive feedback
+* Accepting responsibility and apologizing to those affected by our mistakes,
+  and learning from the experience
+* Focusing on what is best not just for us as individuals, but for the
+  overall community
+
+Examples of unacceptable behavior include:
+
+* The use of sexualized language or imagery, and sexual attention or
+  advances of any kind
+* Trolling, insulting or derogatory comments, and personal or political attacks
+* Public or private harassment
+* Publishing others' private information, such as a physical or email
+  address, without their explicit permission
+* Other conduct which could reasonably be considered inappropriate in a
+  professional setting
+
+## Enforcement Responsibilities
+
+Community leaders are responsible for clarifying and enforcing our standards of
+acceptable behavior and will take appropriate and fair corrective action in
+response to any behavior that they deem inappropriate, threatening, offensive,
+or harmful.
+
+Community leaders have the right and responsibility to remove, edit, or reject
+comments, commits, code, wiki edits, issues, and other contributions that are
+not aligned to this Code of Conduct, and will communicate reasons for moderation
+decisions when appropriate.
+
+## Scope
+
+This Code of Conduct applies within all community spaces, and also applies when
+an individual is officially representing the community in public spaces.
+Examples of representing our community include using an official e-mail address,
+posting via an official social media account, or acting as an appointed
+representative at an online or offline event.
+
+## Enforcement
+
+Instances of abusive, harassing, or otherwise unacceptable behavior may be
+reported to the community leaders responsible for enforcement at
+cjackhwang@qq.com.
+All complaints will be reviewed and investigated promptly and fairly.
+
+All community leaders are obligated to respect the privacy and security of the
+reporter of any incident.
+
+## Enforcement Guidelines
+
+Community leaders will follow these Community Impact Guidelines in determining
+the consequences for any action they deem in violation of this Code of Conduct:
+
+### 1. Correction
+
+**Community Impact**: Use of inappropriate language or other behavior deemed
+unprofessional or unwelcome in the community.
+
+**Consequence**: A private, written warning from community leaders, providing
+clarity around the nature of the violation and an explanation of why the
+behavior was inappropriate. A public apology may be requested.
+
+### 2. Warning
+
+**Community Impact**: A violation through a single incident or series
+of actions.
+
+**Consequence**: A warning with consequences for continued behavior. No
+interaction with the people involved, including unsolicited interaction with
+those enforcing the Code of Conduct, for a specified period of time. This
+includes avoiding interactions in community spaces as well as external channels
+like social media. Violating these terms may lead to a temporary or
+permanent ban.
+
+### 3. Temporary Ban
+
+**Community Impact**: A serious violation of community standards, including
+sustained inappropriate behavior.
+
+**Consequence**: A temporary ban from any sort of interaction or public
+communication with the community for a specified period of time. No public or
+private interaction with the people involved, including unsolicited interaction
+with those enforcing the Code of Conduct, is allowed during this period.
+Violating these terms may lead to a permanent ban.
+
+### 4. Permanent Ban
+
+**Community Impact**: Demonstrating a pattern of violation of community
+standards, including sustained inappropriate behavior,  harassment of an
+individual, or aggression toward or disparagement of classes of individuals.
+
+**Consequence**: A permanent ban from any sort of public interaction within
+the community.
+
+## Attribution
+
+This Code of Conduct is adapted from the [Contributor Covenant][homepage],
+version 2.0, available at
+https://www.contributor-covenant.org/version/2/0/code_of_conduct.html.
+
+Community Impact Guidelines were inspired by [Mozilla's code of conduct
+enforcement ladder](https://github.com/mozilla/diversity).
+
+[homepage]: https://www.contributor-covenant.org
+
+For answers to common questions about this code of conduct, see the FAQ at
+https://www.contributor-covenant.org/faq. Translations are available at
+https://www.contributor-covenant.org/translations.

Dockerfile

@@ -3,6 +3,7 @@ FROM node:24 AS webui-builder
 WORKDIR /app/webui
 COPY webui/package.json webui/package-lock.json ./
 RUN npm ci
+COPY config.example.json /app/config.example.json
 COPY webui ./
 RUN npm run build
 

README.MD

@@ -35,25 +35,27 @@ flowchart LR
     Client["🖥️ 客户端 / SDK\n(OpenAI / Claude / Gemini)"]
     Upstream["☁️ DeepSeek API"]
 
-    subgraph DS2API["DS2API 3.x（统一 OpenAI 内核）"]
+    subgraph DS2API["DS2API 4.x（模块化 HTTP surface + PromptCompat 内核）"]
         Router["chi Router + 中间件\n(RequestID / RealIP / Logger / Recoverer / CORS)"]
 
-        subgraph Adapters["协议适配层"]
-            OA["OpenAI\n/v1/*"]
+        subgraph HTTP["HTTP API surface"]
+            OA["OpenAI\nchat / responses / files / embeddings"]
             CA["Claude\n/anthropic/* + /v1/messages"]
             GA["Gemini\n/v1beta/models/* + /v1/models/*"]
-            Admin["Admin API\n/admin/*"]
+            Admin["Admin API\n资源子包"]
             WebUI["WebUI\n/admin（静态托管）"]
+            Vercel["Vercel Node Stream\n/v1/chat/completions"]
         end
 
         subgraph Runtime["运行时核心能力"]
-            Bridge["CLIProxy 转换桥\n(多协议 <-> OpenAI)"]
-            OAEngine["OpenAI ChatCompletions\n(统一工具调用与流式语义)"]
+            Compat["PromptCompat\n(API -> 网页纯文本上下文)"]
+            Chat["Chat / Responses Runtime\n(统一工具调用与流式语义)"]
             Auth["Auth Resolver\n(API key / bearer / x-goog-api-key)"]
             Pool["Account Pool + Queue\n(并发槽位 + 等待队列)"]
-            DSClient["DeepSeek Client\n(Session / Auth / HTTP)"]
-            Pow["PoW 实现\n(纯 Go 毫秒级)"]
+            DSClient["DeepSeek Client\n(Session / Auth / Completion / Files)"]
+            Pow["PoW 实现\n(纯 Go)"]
             Tool["Tool Sieve\n(Go/Node 语义对齐)"]
+            History["History Split\n(长历史文件化)"]
         end
     end
 
@@ -61,19 +63,23 @@ flowchart LR
     Router --> OA & CA & GA
     Router --> Admin
     Router --> WebUI
+    Router --> Vercel
 
-    OA --> OAEngine
-    CA & GA --> Bridge
-    Bridge --> OAEngine
-    OAEngine --> Auth
-    OAEngine -.账号轮询.-> Pool
-    OAEngine -.工具调用解析.-> Tool
-    OAEngine -.PoW 计算.-> Pow
+    OA --> Compat
+    CA & GA --> Compat
+    Compat --> Chat
+    Compat -.长历史.-> History
+    Vercel -.Go prepare.-> Chat
+    Vercel -.Node SSE.-> Tool
+    Chat --> Auth
+    Chat -.账号轮询.-> Pool
+    Chat -.工具调用解析.-> Tool
+    Chat -.PoW 计算.-> Pow
     Auth --> DSClient
     DSClient --> Upstream
     Upstream --> DSClient
-    OAEngine --> Bridge
-    Bridge --> Client
+    Chat --> Client
+    Vercel --> Client
 ```
 
 详细架构拆分与目录职责见 [docs/ARCHITECTURE.md](docs/ARCHITECTURE.md)。
@@ -89,6 +95,7 @@ flowchart LR
 | OpenAI 兼容 | `GET /v1/models`、`GET /v1/models/{id}`、`POST /v1/chat/completions`、`POST /v1/responses`、`GET /v1/responses/{response_id}`、`POST /v1/embeddings`、`POST /v1/files` |
 | Claude 兼容 | `GET /anthropic/v1/models`、`POST /anthropic/v1/messages`、`POST /anthropic/v1/messages/count_tokens`（及快捷路径 `/v1/messages`、`/messages`） |
 | Gemini 兼容 | `POST /v1beta/models/{model}:generateContent`、`POST /v1beta/models/{model}:streamGenerateContent`（及 `/v1/models/{model}:*` 路径） |
+| 统一 CORS 兼容 | `/v1/*`、`/anthropic/*`、`/v1beta/models/*`、`/admin/*` 统一走同一套 CORS 策略；Vercel 上 `/v1/chat/completions` 的 Node Runtime 也对齐相同放行规则，尽量减少第三方预检请求头限制 |
 | 多账号轮询 | 自动 token 刷新、邮箱/手机号双登录方式 |
 | 并发队列控制 | 每账号 in-flight 上限 + 等待队列，动态计算建议并发值 |
 | DeepSeek PoW | 纯 Go 高性能实现（DeepSeekHashV1），毫秒级响应 |
@@ -114,38 +121,32 @@ flowchart LR
 
 | 模型类型 | 模型 ID | thinking | search |
 | --- | --- | --- | --- |
-| default | `deepseek-chat` | ❌ | ❌ |
-| default | `deepseek-reasoner` | ✅ | ❌ |
-| default | `deepseek-chat-search` | ❌ | ✅ |
-| default | `deepseek-reasoner-search` | ✅ | ✅ |
-| expert | `deepseek-expert-chat` | ❌ | ❌ |
-| expert | `deepseek-expert-reasoner` | ✅ | ❌ |
-| expert | `deepseek-expert-chat-search` | ❌ | ✅ |
-| expert | `deepseek-expert-reasoner-search` | ✅ | ✅ |
-| vision | `deepseek-vision-chat` | ❌ | ❌ |
-| vision | `deepseek-vision-reasoner` | ✅ | ❌ |
-| vision | `deepseek-vision-chat-search` | ❌ | ✅ |
-| vision | `deepseek-vision-reasoner-search` | ✅ | ✅ |
+| default | `deepseek-v4-flash` | 默认开启，可由请求参数控制 | ❌ |
+| expert | `deepseek-v4-pro` | 默认开启，可由请求参数控制 | ❌ |
+| default | `deepseek-v4-flash-search` | 默认开启，可由请求参数控制 | ✅ |
+| expert | `deepseek-v4-pro-search` | 默认开启，可由请求参数控制 | ✅ |
+| vision | `deepseek-v4-vision` | 默认开启，可由请求参数控制 | ❌ |
+| vision | `deepseek-v4-vision-search` | 默认开启，可由请求参数控制 | ✅ |
 
-除原生模型外，也支持常见 alias 输入（如 `gpt-5`、`gpt-5-mini`、`gpt-5-codex`、`gpt-4.1`、`o3`、`claude-opus-4-6`、`claude-sonnet-4-5`、`gemini-2.5-pro`、`gemini-2.5-flash` 等），但 `/v1/models` 返回的是规范化后的 DeepSeek 原生模型 ID。
+除原生模型外，也支持常见 alias 输入（如 `gpt-4.1`、`gpt-5`、`gpt-5-codex`、`o3`、`claude-*`、`gemini-*` 等），但 `/v1/models` 返回的是规范化后的 DeepSeek 原生模型 ID。完整 alias 行为以 [API.md](API.md#模型-alias-解析策略) 和 `config.example.json` 为准。
 
 ### Claude 接口（`GET /anthropic/v1/models`）
 
 | 当前常用模型 | 默认映射 |
 | --- | --- |
-| `claude-sonnet-4-5` | `deepseek-chat` |
-| `claude-haiku-4-5`（兼容 `claude-3-5-haiku-latest`） | `deepseek-chat` |
-| `claude-opus-4-6` | `deepseek-reasoner` |
+| `claude-sonnet-4-6` | `deepseek-v4-flash` |
+| `claude-haiku-4-5`（兼容 `claude-3-5-haiku-latest`） | `deepseek-v4-flash` |
+| `claude-opus-4-6` | `deepseek-v4-pro` |
 
-可通过配置中的 `claude_mapping` 或 `claude_model_mapping` 覆盖映射关系。
-`/anthropic/v1/models` 除上述当前主别名外，还会返回 Claude 4.x snapshots，以及 3.x / 2.x / 1.x 历史模型 ID 与常见 alias，便于旧客户端直接兼容。
+可通过配置中的 `model_aliases` 覆盖映射关系。
+`/anthropic/v1/models` 除上述主别名外，还会返回 Claude 4.x snapshots、3.x 历史模型 ID 与常见 alias，便于旧客户端直接兼容。
 
 #### Claude Code 接入避坑（实测）
 
 - `ANTHROPIC_BASE_URL` 推荐直接指向 DS2API 根地址（例如 `http://127.0.0.1:5001`），Claude Code 会请求 `/v1/messages?beta=true`。
 - `ANTHROPIC_API_KEY` 需要与 `config.json` 中 `keys` 一致；建议同时保留常规 key 与 `sk-ant-*` 形态 key，兼容不同客户端校验习惯。
 - 若系统设置了代理，建议对 DS2API 地址配置 `NO_PROXY=127.0.0.1,localhost,<你的主机IP>`，避免本地回环请求被代理拦截。
-- 如遇“工具调用输出成文本、未执行”问题，请优先检查模型输出是否为受支持的 XML/Markup 工具块（例如 `<tool_call>` / `<function_call>` / `<invoke>` / `tool_use`），而不是纯 JSON `tool_calls` 片段。
+- 如遇“工具调用输出成文本、未执行”问题，请优先检查模型输出是否为当前唯一受支持的 XML 工具块：`<tool_calls><invoke name="..."><parameter name="...">...`，而不是旧式 `<tools>` / `<tool_call>` / `<tool_name>` / `<param>`、`<function_call>`、`tool_use` 或纯 JSON `tool_calls` 片段。
 
 ### Gemini 接口
 
@@ -239,7 +240,7 @@ cp config.example.json config.json
 base64 < config.json | tr -d '\n'
 ```
 
-> **流式说明**：`/v1/chat/completions` 在 Vercel 上默认走 `api/chat-stream.js`（Node Runtime）以保证实时 SSE。鉴权、账号选择、会话/PoW 准备仍由 Go 内部 prepare 接口完成；流式响应（含 `tools`）在 Node 侧执行与 Go 对齐的输出组装与防泄漏处理。
+> **流式说明**：`/v1/chat/completions` 在 Vercel 上默认走 `api/chat-stream.js`（Node Runtime）以保证实时 SSE。鉴权、账号选择、会话/PoW 准备仍由 Go 内部 prepare 接口完成；流式响应（含 `tools`）在 Node 侧执行与 Go 对齐的输出组装与防泄漏处理。虽然这里只有 OpenAI chat 流式走 Node，但 CORS 放行策略仍与 Go 主路由保持一致，统一覆盖第三方客户端预检场景。
 
 详细部署说明请参阅 [部署指南](docs/DEPLOY.md)。
 
@@ -268,123 +269,18 @@ go run ./cmd/ds2api
 
 ## 配置说明
 
-### `config.json` 示例
+`README` 只保留快速入口，完整字段请以 [config.example.json](config.example.json) 为模板，并参考 [部署指南](docs/DEPLOY.md#0-前置要求) 与 [API 配置最佳实践](API.md#配置最佳实践)。
 
-```json
-{
-  "keys": ["your-api-key-1", "your-api-key-2"],
-  "api_keys": [
-    {
-      "key": "your-api-key-1",
-      "name": "主 Key",
-      "remark": "生产流量"
-    }
-  ],
-  "accounts": [
-    {
-      "name": "账号 A",
-      "remark": "主账号",
-      "email": "user@example.com",
-      "password": "your-password"
-    },
-    {
-      "mobile": "12345678901",
-      "password": "your-password"
-    }
-  ],
-  "model_aliases": {
-    "gpt-4o": "deepseek-chat",
-    "gpt-5": "deepseek-chat",
-    "gpt-5-mini": "deepseek-chat",
-    "gpt-5-codex": "deepseek-reasoner",
-    "o3": "deepseek-reasoner",
-    "claude-opus-4-6": "deepseek-reasoner",
-    "gemini-2.5-flash": "deepseek-chat"
-  },
-  "compat": {
-    "wide_input_strict_output": true,
-    "strip_reference_markers": true
-  },
-  "responses": {
-    "store_ttl_seconds": 900
-  },
-  "embeddings": {
-    "provider": "deterministic"
-  },
-  "claude_mapping": {
-    "fast": "deepseek-chat",
-    "slow": "deepseek-reasoner"
-  },
-  "admin": {
-    "jwt_expire_hours": 24
-  },
-  "runtime": {
-    "account_max_inflight": 2,
-    "account_max_queue": 0,
-    "global_max_inflight": 0,
-    "token_refresh_interval_hours": 6
-  },
-  "auto_delete": {
-    "mode": "none"
-  }
-}
-```
+常用字段：
 
-- `keys`：API 访问密钥列表，客户端通过 `Authorization: Bearer <key>` 鉴权
-- `api_keys`：推荐使用的新结构化密钥列表，支持 `key` + `name` + `remark`（`keys` 仍兼容）
-- `accounts`：DeepSeek 账号列表，支持 `email` 或 `mobile` 登录；可额外填写 `name` / `remark` 便于管理
-- `token`：配置文件中即使填写也会在加载时被清空（不会从 `config.json` 读取 token）；实际 token 仅在运行时内存中维护并自动刷新
-- `model_aliases`：常见模型名（如 GPT/Codex/Claude）到 DeepSeek 模型的映射
-- `compat.wide_input_strict_output`：建议保持 `true`（当前实现默认宽进严出）
-- `compat.strip_reference_markers`：建议保持 `true`，用于清理可见输出中的引用/标记
-- `toolcall`：旧字段，当前实现已固定为特征匹配 + 高置信早发；即使保留在配置里也会被忽略
-- `responses.store_ttl_seconds`：`/v1/responses/{id}` 的内存缓存 TTL
-- `embeddings.provider`：embedding 提供方（当前内置 `deterministic/mock/builtin`）
-- `claude_mapping`：字典中 `fast`/`slow` 后缀映射到对应 DeepSeek 模型（兼容读取 `claude_model_mapping`）
-- `admin`：管理后台设置（JWT 过期时间、密码哈希等），可通过 Admin Settings API 热更新
-- `runtime`：运行时参数（并发限制、队列大小、托管账号 token 刷新间隔），可通过 Admin Settings API 热更新；`account_max_queue=0`/`global_max_inflight=0` 表示按推荐值自动计算，`token_refresh_interval_hours=6` 为默认强制重登间隔
-- `auto_delete.mode`：请求结束后如何清理 DeepSeek 远端聊天记录，支持 `none`（默认，不删除）、`single`（仅删除当前会话）、`all`（清空全部会话）；旧配置里的 `auto_delete.sessions=true` 仍会被视为 `all`
+- `keys` / `api_keys`：客户端访问密钥，`api_keys` 支持 `name` 与 `remark` 元信息，`keys` 继续兼容。
+- `accounts`：DeepSeek 托管账号，支持 `email` 或 `mobile` 登录，可配置代理、名称和备注。
+- `model_aliases`：OpenAI / Claude / Gemini 共用的模型 alias 映射。
+- `runtime`：账号并发、队列与 token 刷新策略，可通过 Admin Settings 热更新。
+- `auto_delete.mode`：请求结束后的远端会话清理策略，支持 `none` / `single` / `all`。
+- `history_split`：多轮历史拆分策略，已全局强制开启；可调整触发阈值，避免长历史全部内联进 prompt。
 
-### 环境变量
-
-> 建议：长期维护请优先以 `config.json`（或其 Base64）为单一配置源。环境变量仅保留部署必需项；`DS2API_CONFIG_JSON` 主要用于 Vercel/无持久盘场景，后续可能进一步收敛。
-
-| 变量 | 用途 | 默认值 |
-| --- | --- | --- |
-| `PORT` | 服务端口 | `5001` |
-| `LOG_LEVEL` | 日志级别 | `INFO`（可选：`DEBUG`/`WARN`/`ERROR`） |
-| `DS2API_ADMIN_KEY` | Admin 登录密钥 | `admin` |
-| `DS2API_JWT_SECRET` | Admin JWT 签名密钥 | 等同 `DS2API_ADMIN_KEY` |
-| `DS2API_JWT_EXPIRE_HOURS` | Admin JWT 过期小时数 | `24` |
-| `DS2API_CONFIG_PATH` | 配置文件路径 | `config.json` |
-| `DS2API_CONFIG_JSON` | 直接注入配置（JSON 或 Base64） | — |
-| `DS2API_CHAT_HISTORY_PATH` | 服务器端对话记录文件路径 | `data/chat_history.json` |
-| `DS2API_ENV_WRITEBACK` | 环境变量模式下自动写回配置文件并切换文件模式（`1/true/yes/on`） | 关闭 |
-| `DS2API_STATIC_ADMIN_DIR` | 管理台静态文件目录 | `static/admin` |
-| `DS2API_AUTO_BUILD_WEBUI` | 启动时自动构建 WebUI | 本地开启，Vercel 关闭 |
-| `DS2API_DEV_PACKET_CAPTURE` | 本地开发抓包开关（记录最近会话请求/响应体） | 本地非 Vercel 默认开启 |
-| `DS2API_DEV_PACKET_CAPTURE_LIMIT` | 本地抓包保留条数（超出自动淘汰） | `20` |
-| `DS2API_DEV_PACKET_CAPTURE_MAX_BODY_BYTES` | 单条响应体最大记录字节数 | `5242880` |
-| `DS2API_ACCOUNT_MAX_INFLIGHT` | 每账号最大并发 in-flight 请求数 | `2` |
-| `DS2API_ACCOUNT_MAX_QUEUE` | 等待队列上限 | `recommended_concurrency` |
-| `DS2API_GLOBAL_MAX_INFLIGHT` | 全局最大 in-flight 请求数 | `recommended_concurrency` |
-| `DS2API_VERCEL_INTERNAL_SECRET` | Vercel 混合流式内部鉴权密钥 | 回退用 `DS2API_ADMIN_KEY` |
-| `DS2API_VERCEL_STREAM_LEASE_TTL_SECONDS` | 流式 lease 过期秒数 | `900` |
-| `VERCEL_TOKEN` | Vercel 同步 token | — |
-| `VERCEL_PROJECT_ID` | Vercel 项目 ID | — |
-| `VERCEL_TEAM_ID` | Vercel 团队 ID | — |
-| `DS2API_VERCEL_PROTECTION_BYPASS` | Vercel 部署保护绕过密钥（内部 Node→Go 调用） | — |
-
-> 提示：当检测到 `DS2API_CONFIG_JSON` 时，管理台会显示当前模式风险与自动持久化状态（含 `DS2API_CONFIG_PATH` 路径与模式切换说明）。
-
-#### 必填 / 可选（按部署方式）
-
-- **所有部署都必填**：`DS2API_ADMIN_KEY`
-- **配置来源二选一（推荐前者）**：
-  - `config.json` 文件（推荐，持久化更直观）
-  - `DS2API_CONFIG_JSON`（可选，适合 Vercel；支持 JSON 或 Base64）
-- **仅在环境变量配置模式建议开启**：`DS2API_ENV_WRITEBACK=1`（避免管理台改动重启后丢失）
-- 其余环境变量均为可选调优项。
+环境变量完整列表见 [部署指南](docs/DEPLOY.md)，接口鉴权规则见 [API.md](API.md#鉴权规则)。
 
 ## 鉴权模式
 
@@ -416,7 +312,7 @@ Gemini 路由还可以使用 `x-goog-api-key`，或在没有认证头时使用 `
 当请求中带 `tools` 时，DS2API 会做防泄漏处理与结构化转译：
 
 1. 只在**非代码块上下文**启用执行型 toolcall 识别（代码块示例默认不触发）
-2. 解析层当前以 XML/Markup 家族为准（`<tool_call>` / `<function_call>` / `<invoke>` / `tool_use` / antml 变体）；纯 JSON `tool_calls` 片段默认不作为可执行调用解析
+2. 解析层当前只把 canonical XML 工具块视为可执行调用：`<tool_calls>` → `<invoke name="...">` → `<parameter name="...">`；旧式 `<tools>` / `<tool_call>` / `<tool_name>` / `<param>`、`<function_call>`、`tool_use` / antml 变体与纯 JSON `tool_calls` 片段都会按普通文本处理
 3. `responses` 流式严格使用官方 item 生命周期事件（`response.output_item.*`、`response.content_part.*`、`response.function_call_arguments.*`）
 4. `responses` 支持并执行 `tool_choice`（`auto`/`none`/`required`/强制函数）；`required` 违规时非流式返回 `422`，流式返回 `response.failed`
 5. 客户端请求哪种协议，就按该协议返回工具调用（OpenAI/Claude/Gemini 各自原生结构）；模型侧优先约束输出规范 XML，再由兼容层转译
@@ -467,44 +363,18 @@ go run ./cmd/ds2api
 
 ## 测试
 
-```bash
-# 单元测试（Go + Node）
-./tests/scripts/run-unit-all.sh
-
-# 一键端到端全链路测试（真实账号，生成完整请求/响应日志）
-./tests/scripts/run-live.sh
-
-# 或自定义参数
-go run ./cmd/ds2api-tests \
-  --config config.json \
-  --admin-key admin \
-  --out artifacts/testsuite \
-  --timeout 120 \
-  --retries 2
-```
-
-```bash
-# 发布前阻断门禁
-./tests/scripts/check-stage6-manual-smoke.sh
-./tests/scripts/check-refactor-line-gate.sh
-./tests/scripts/run-unit-all.sh
-npm ci --prefix webui && npm run build --prefix webui
-```
-
-## 测试
-
 详细测试指南请参阅 [docs/TESTING.md](docs/TESTING.md)。
 
 ### 快速测试命令
 
 ```bash
-# 运行所有单元测试
-go test ./...
+# 本地 PR 门禁
+./scripts/lint.sh
+./tests/scripts/check-refactor-line-gate.sh
+./tests/scripts/run-unit-all.sh
+npm run build --prefix webui
 
-# 运行 tool calls 相关测试（调试工具调用问题）
-go test -v -run 'TestParseToolCalls|TestRepair' ./internal/toolcall/
-
-# 运行端到端测试
+# 端到端全链路测试（真实账号，生成完整请求/响应日志）
 ./tests/scripts/run-live.sh
 ```
 

README.en.md

@@ -33,25 +33,27 @@ flowchart LR
     Client["🖥️ Clients / SDKs\n(OpenAI / Claude / Gemini)"]
     Upstream["☁️ DeepSeek API"]
 
-    subgraph DS2API["DS2API 3.x (Unified OpenAI Core)"]
+    subgraph DS2API["DS2API 4.x (Modular HTTP Surface + PromptCompat Core)"]
         Router["chi Router + Middleware\n(RequestID / RealIP / Logger / Recoverer / CORS)"]
 
-        subgraph Adapters["Protocol Adapters"]
-            OA["OpenAI\n/v1/*"]
+        subgraph HTTP["HTTP API Surface"]
+            OA["OpenAI\nchat / responses / files / embeddings"]
             CA["Claude\n/anthropic/* + /v1/messages"]
             GA["Gemini\n/v1beta/models/* + /v1/models/*"]
-            Admin["Admin API\n/admin/*"]
+            Admin["Admin API\nresource packages"]
             WebUI["WebUI\n/admin (static hosting)"]
+            Vercel["Vercel Node Stream\n/v1/chat/completions"]
         end
 
         subgraph Runtime["Runtime + Core Capabilities"]
-            Bridge["CLIProxy Bridge\n(multi-protocol <-> OpenAI)"]
-            OAEngine["OpenAI ChatCompletions\n(unified tools + stream semantics)"]
+            Compat["PromptCompat\n(API -> web-chat plain text context)"]
+            Chat["Chat / Responses Runtime\n(unified tools + stream semantics)"]
             Auth["Auth Resolver\n(API key / bearer / x-goog-api-key)"]
             Pool["Account Pool + Queue\n(in-flight slots + wait queue)"]
-            DSClient["DeepSeek Client\n(session / auth / HTTP)"]
-            Pow["PoW Solver\n(Pure Go ms-level)"]
+            DSClient["DeepSeek Client\n(session / auth / completion / files)"]
+            Pow["PoW Solver\n(Pure Go)"]
             Tool["Tool Sieve\n(Go/Node semantic parity)"]
+            History["History Split\n(long history as files)"]
         end
     end
 
@@ -59,19 +61,23 @@ flowchart LR
     Router --> OA & CA & GA
     Router --> Admin
     Router --> WebUI
+    Router --> Vercel
 
-    OA --> OAEngine
-    CA & GA --> Bridge
-    Bridge --> OAEngine
-    OAEngine --> Auth
-    OAEngine -.account rotation.-> Pool
-    OAEngine -.tool-call parsing.-> Tool
-    OAEngine -.PoW solving.-> Pow
+    OA --> Compat
+    CA & GA --> Compat
+    Compat --> Chat
+    Compat -.long history.-> History
+    Vercel -.Go prepare.-> Chat
+    Vercel -.Node SSE.-> Tool
+    Chat --> Auth
+    Chat -.account rotation.-> Pool
+    Chat -.tool-call parsing.-> Tool
+    Chat -.PoW solving.-> Pow
     Auth --> DSClient
     DSClient --> Upstream
     Upstream --> DSClient
-    OAEngine --> Bridge
-    Bridge --> Client
+    Chat --> Client
+    Vercel --> Client
 ```
 
 For the full module-by-module architecture and directory responsibilities, see [docs/ARCHITECTURE.en.md](docs/ARCHITECTURE.en.md).
@@ -87,6 +93,7 @@ For the full module-by-module architecture and directory responsibilities, see [
 | OpenAI compatible | `GET /v1/models`, `GET /v1/models/{id}`, `POST /v1/chat/completions`, `POST /v1/responses`, `GET /v1/responses/{response_id}`, `POST /v1/embeddings`, `POST /v1/files` |
 | Claude compatible | `GET /anthropic/v1/models`, `POST /anthropic/v1/messages`, `POST /anthropic/v1/messages/count_tokens` (plus shortcut paths `/v1/messages`, `/messages`) |
 | Gemini compatible | `POST /v1beta/models/{model}:generateContent`, `POST /v1beta/models/{model}:streamGenerateContent` (plus `/v1/models/{model}:*` paths) |
+| Unified CORS compatibility | `/v1/*`, `/anthropic/*`, `/v1beta/models/*`, and `/admin/*` share one CORS policy; on Vercel, the Node Runtime for `/v1/chat/completions` mirrors the same relaxed preflight behavior for third-party clients |
 | Multi-account rotation | Auto token refresh, email/mobile dual login |
 | Concurrency control | Per-account in-flight limit + waiting queue, dynamic recommended concurrency |
 | DeepSeek PoW | Pure Go high-performance solver (DeepSeekHashV1), ms-level response |
@@ -112,38 +119,32 @@ For the full module-by-module architecture and directory responsibilities, see [
 
 | Family | Model ID | thinking | search |
 | --- | --- | --- | --- |
-| default | `deepseek-chat` | ❌ | ❌ |
-| default | `deepseek-reasoner` | ✅ | ❌ |
-| default | `deepseek-chat-search` | ❌ | ✅ |
-| default | `deepseek-reasoner-search` | ✅ | ✅ |
-| expert | `deepseek-expert-chat` | ❌ | ❌ |
-| expert | `deepseek-expert-reasoner` | ✅ | ❌ |
-| expert | `deepseek-expert-chat-search` | ❌ | ✅ |
-| expert | `deepseek-expert-reasoner-search` | ✅ | ✅ |
-| vision | `deepseek-vision-chat` | ❌ | ❌ |
-| vision | `deepseek-vision-reasoner` | ✅ | ❌ |
-| vision | `deepseek-vision-chat-search` | ❌ | ✅ |
-| vision | `deepseek-vision-reasoner-search` | ✅ | ✅ |
+| default | `deepseek-v4-flash` | enabled by default, request-controlled | ❌ |
+| expert | `deepseek-v4-pro` | enabled by default, request-controlled | ❌ |
+| default | `deepseek-v4-flash-search` | enabled by default, request-controlled | ✅ |
+| expert | `deepseek-v4-pro-search` | enabled by default, request-controlled | ✅ |
+| vision | `deepseek-v4-vision` | enabled by default, request-controlled | ❌ |
+| vision | `deepseek-v4-vision-search` | enabled by default, request-controlled | ✅ |
 
-Besides native IDs, DS2API also accepts common aliases as input (for example `gpt-5`, `gpt-5-mini`, `gpt-5-codex`, `gpt-4.1`, `o3`, `claude-opus-4-6`, `claude-sonnet-4-5`, `gemini-2.5-pro`, `gemini-2.5-flash`), but `/v1/models` returns normalized DeepSeek native model IDs.
+Besides native IDs, DS2API also accepts common aliases as input (for example `gpt-4.1`, `gpt-5`, `gpt-5-codex`, `o3`, `claude-*`, `gemini-*`), but `/v1/models` returns normalized DeepSeek native model IDs. The complete alias behavior is documented in [API.en.md](API.en.md#model-alias-resolution) and `config.example.json`.
 
 ### Claude Endpoint (`GET /anthropic/v1/models`)
 
 | Current common model | Default Mapping |
 | --- | --- |
-| `claude-sonnet-4-5` | `deepseek-chat` |
-| `claude-haiku-4-5` (compatible with `claude-3-5-haiku-latest`) | `deepseek-chat` |
-| `claude-opus-4-6` | `deepseek-reasoner` |
+| `claude-sonnet-4-6` | `deepseek-v4-flash` |
+| `claude-haiku-4-5` (compatible with `claude-3-5-haiku-latest`) | `deepseek-v4-flash` |
+| `claude-opus-4-6` | `deepseek-v4-pro` |
 
-Override mapping via `claude_mapping` or `claude_model_mapping` in config.
-Besides the current primary aliases above, `/anthropic/v1/models` also returns Claude 4.x snapshots plus historical 3.x / 2.x / 1.x IDs and common aliases for legacy client compatibility.
+Override mapping via the global `model_aliases` config.
+Besides the primary aliases above, `/anthropic/v1/models` also returns Claude 4.x snapshots plus historical 3.x IDs and common aliases for legacy client compatibility.
 
 #### Claude Code integration pitfalls (validated)
 
 - Set `ANTHROPIC_BASE_URL` to the DS2API root URL (for example `http://127.0.0.1:5001`). Claude Code sends requests to `/v1/messages?beta=true`.
 - `ANTHROPIC_API_KEY` must match an entry in `keys` from `config.json`. Keeping both a regular key and an `sk-ant-*` style key improves client compatibility.
 - If your environment has proxy variables, set `NO_PROXY=127.0.0.1,localhost,<your_host_ip>` for DS2API to avoid proxy interception of local traffic.
-- If tool calls are rendered as plain text and not executed, first verify the model output uses supported XML/Markup tool blocks (`<tool_call>` / `<function_call>` / `<invoke>` / `tool_use`) rather than standalone JSON `tool_calls`.
+- If tool calls are rendered as plain text and not executed, first verify the model output uses the only supported XML block: `<tool_calls><invoke name="..."><parameter name="...">...`, not legacy `<tools>` / `<tool_call>` / `<tool_name>` / `<param>`, `<function_call>`, `tool_use`, or standalone JSON `tool_calls`.
 
 ### Gemini Endpoint
 
@@ -237,7 +238,7 @@ Recommended: convert `config.json` to Base64 locally, then paste into `DS2API_CO
 base64 < config.json | tr -d '\n'
 ```
 
-> **Streaming note**: `/v1/chat/completions` on Vercel is routed to `api/chat-stream.js` (Node Runtime) for real-time SSE. Auth, account selection, and session/PoW preparation are still handled by the Go internal prepare endpoint; streaming output (including `tools`) is assembled on Node with Go-aligned anti-leak handling.
+> **Streaming note**: `/v1/chat/completions` on Vercel is routed to `api/chat-stream.js` (Node Runtime) for real-time SSE. Auth, account selection, and session/PoW preparation are still handled by the Go internal prepare endpoint; streaming output (including `tools`) is assembled on Node with Go-aligned anti-leak handling. This is the only interface family currently routed through Node, and its CORS allow behavior is kept aligned with the Go router so third-party preflight handling stays unified.
 
 For detailed deployment instructions, see the [Deployment Guide](docs/DEPLOY.en.md).
 
@@ -266,102 +267,18 @@ The server actually binds to `0.0.0.0:5001`, so devices on the same LAN can usua
 
 ## Configuration
 
-### `config.json` Example
+`README` keeps only the onboarding path. Use [config.example.json](config.example.json) as the field template, and see the [deployment guide](docs/DEPLOY.en.md#0-prerequisites) plus [API configuration notes](API.en.md#configuration-best-practice) for full details.
 
-```json
-{
-  "keys": ["your-api-key-1", "your-api-key-2"],
-  "accounts": [
-    {
-      "email": "user@example.com",
-      "password": "your-password"
-    },
-    {
-      "mobile": "12345678901",
-      "password": "your-password"
-    }
-  ],
-  "model_aliases": {
-    "gpt-4o": "deepseek-chat",
-    "gpt-5": "deepseek-chat",
-    "gpt-5-mini": "deepseek-chat",
-    "gpt-5-codex": "deepseek-reasoner",
-    "o3": "deepseek-reasoner",
-    "claude-opus-4-6": "deepseek-reasoner",
-    "gemini-2.5-flash": "deepseek-chat"
-  },
-  "compat": {
-    "wide_input_strict_output": true,
-    "strip_reference_markers": true
-  },
-  "responses": {
-    "store_ttl_seconds": 900
-  },
-  "embeddings": {
-    "provider": "deterministic"
-  },
-  "claude_mapping": {
-    "fast": "deepseek-chat",
-    "slow": "deepseek-reasoner"
-  },
-  "admin": {
-    "jwt_expire_hours": 24
-  },
-  "runtime": {
-    "account_max_inflight": 2,
-    "account_max_queue": 0,
-    "global_max_inflight": 0,
-    "token_refresh_interval_hours": 6
-  },
-  "auto_delete": {
-    "mode": "none"
-  }
-}
-```
+Common fields:
 
-- `keys`: API access keys; clients authenticate via `Authorization: Bearer <key>`
-- `accounts`: DeepSeek account list, supports `email` or `mobile` login
-- `token`: Even if set in `config.json`, it is cleared during load (DS2API does not read persisted tokens from config); runtime tokens are maintained/refreshed in memory only
-- `model_aliases`: Map common model names (GPT/Codex/Claude) to DeepSeek models
-- `compat.wide_input_strict_output`: Keep `true` (current default policy)
-- `compat.strip_reference_markers`: Keep `true`; it strips reference markers from visible output
-- `toolcall`: Legacy field; the current behavior is fixed to feature matching + high-confidence early emit, and any config value is ignored
-- `responses.store_ttl_seconds`: In-memory TTL for `/v1/responses/{id}`
-- `embeddings.provider`: Embeddings provider (`deterministic/mock/builtin` built-in)
-- `claude_mapping`: Maps `fast`/`slow` suffixes to corresponding DeepSeek models (still compatible with `claude_model_mapping`)
-- `admin`: Admin panel settings (JWT expiry, password hash, etc.), hot-reloadable via Admin Settings API
-- `runtime`: Runtime parameters (concurrency limits, queue sizes, managed token refresh interval), hot-reloadable via Admin Settings API; `account_max_queue=0`/`global_max_inflight=0` means auto-calculate from recommended values, `token_refresh_interval_hours=6` is the default forced re-login interval
-- `auto_delete.mode`: How to clean up DeepSeek remote chat records after each request completes. Supported values: `none` (default, no deletion), `single` (delete only the current session), `all` (delete all sessions); legacy `auto_delete.sessions=true` is still treated as `all`
+- `keys` / `api_keys`: client API keys; `api_keys` adds `name` and `remark` metadata while `keys` remains compatible.
+- `accounts`: managed DeepSeek accounts, supporting `email` or `mobile` login plus proxy/name/remark metadata.
+- `model_aliases`: one shared alias map for OpenAI / Claude / Gemini model names.
+- `runtime`: account concurrency, queueing, and token refresh behavior, hot-reloadable via Admin Settings.
+- `auto_delete.mode`: remote session cleanup after each request, supporting `none` / `single` / `all`.
+- `history_split`: multi-turn history split policy, now forced on globally; tune its trigger threshold to avoid inlining all long history into the prompt.
 
-### Environment Variables
-
-| Variable | Purpose | Default |
-| --- | --- | --- |
-| `PORT` | Service port | `5001` |
-| `LOG_LEVEL` | Log level | `INFO` (`DEBUG`/`WARN`/`ERROR`) |
-| `DS2API_ADMIN_KEY` | Admin login key | `admin` |
-| `DS2API_JWT_SECRET` | Admin JWT signing secret | Same as `DS2API_ADMIN_KEY` |
-| `DS2API_JWT_EXPIRE_HOURS` | Admin JWT TTL in hours | `24` |
-| `DS2API_CONFIG_PATH` | Config file path | `config.json` |
-| `DS2API_CONFIG_JSON` | Inline config (JSON or Base64) | — |
-| `DS2API_CHAT_HISTORY_PATH` | Server-side conversation history file path | `data/chat_history.json` |
-| `DS2API_ENV_WRITEBACK` | Auto-write env-backed config to file and transition to file mode (`1/true/yes/on`) | Disabled |
-| `DS2API_STATIC_ADMIN_DIR` | Admin static assets dir | `static/admin` |
-| `DS2API_AUTO_BUILD_WEBUI` | Auto-build WebUI on startup | Enabled locally, disabled on Vercel |
-| `DS2API_ACCOUNT_MAX_INFLIGHT` | Max in-flight requests per account | `2` |
-| `DS2API_ACCOUNT_MAX_QUEUE` | Waiting queue limit | `recommended_concurrency` |
-| `DS2API_GLOBAL_MAX_INFLIGHT` | Global max in-flight requests | `recommended_concurrency` |
-| `DS2API_VERCEL_INTERNAL_SECRET` | Vercel hybrid streaming internal auth | Falls back to `DS2API_ADMIN_KEY` |
-| `DS2API_VERCEL_STREAM_LEASE_TTL_SECONDS` | Stream lease TTL seconds | `900` |
-| `DS2API_DEV_PACKET_CAPTURE` | Local dev packet capture switch (record recent request/response bodies) | Enabled by default on non-Vercel local runtime |
-| `DS2API_DEV_PACKET_CAPTURE_LIMIT` | Number of captured sessions to retain (auto-evict overflow) | `20` |
-| `DS2API_DEV_PACKET_CAPTURE_MAX_BODY_BYTES` | Max recorded bytes per captured response body | `5242880` |
-| `VERCEL_TOKEN` | Vercel sync token | — |
-| `VERCEL_PROJECT_ID` | Vercel project ID | — |
-| `VERCEL_TEAM_ID` | Vercel team ID | — |
-| `DS2API_VERCEL_PROTECTION_BYPASS` | Vercel deployment protection bypass for internal Node→Go calls | — |
-
-> Note: when `DS2API_CONFIG_JSON` is detected, the Admin UI shows mode risk and auto-persistence status (including `DS2API_CONFIG_PATH` and mode-transition hints).
+For the full environment variable list, see [docs/DEPLOY.en.md](docs/DEPLOY.en.md). For auth behavior, see [API.en.md](API.en.md#authentication).
 
 ## Authentication Modes
 
@@ -393,7 +310,7 @@ Queue limit = DS2API_ACCOUNT_MAX_QUEUE (default = recommended concurrency)
 When `tools` is present in the request, DS2API performs anti-leak handling:
 
 1. Toolcall feature matching is enabled only in **non-code-block context** (fenced examples are ignored)
-2. The parser currently targets XML/Markup-family tool syntax (`<tool_call>` / `<function_call>` / `<invoke>` / `tool_use` / antml variants); standalone JSON `tool_calls` payloads are not treated as executable calls by default
+2. The parser now treats only the canonical XML wrapper as executable tool-calling syntax: `<tool_calls>` → `<invoke name="...">` → `<parameter name="...">`; legacy `<tools>` / `<tool_call>` / `<tool_name>` / `<param>`, `<function_call>`, `tool_use`, antml variants, and standalone JSON `tool_calls` payloads are treated as plain text
 3. `responses` streaming strictly uses official item lifecycle events (`response.output_item.*`, `response.content_part.*`, `response.function_call_arguments.*`)
 4. `responses` supports and enforces `tool_choice` (`auto`/`none`/`required`/forced function); `required` violations return `422` for non-stream and `response.failed` for stream
 5. The output protocol follows the client request (OpenAI / Claude / Gemini native shapes); model-side prompting can prefer XML, and the compatibility layer handles the protocol-specific translation
@@ -442,28 +359,19 @@ The save endpoint can target a chain by `query`, `chain_key`, or `capture_id`. E
 
 ## Testing
 
-```bash
-# Unit tests (Go + Node)
-./tests/scripts/run-unit-all.sh
+For the full testing guide, see [docs/TESTING.md](docs/TESTING.md).
 
-# One-command live end-to-end tests (real accounts, full request/response logs)
-./tests/scripts/run-live.sh
-
-# Or with custom flags
-go run ./cmd/ds2api-tests \
-  --config config.json \
-  --admin-key admin \
-  --out artifacts/testsuite \
-  --timeout 120 \
-  --retries 2
-```
+Quick commands:
 
 ```bash
-# Release-blocking gates
-./tests/scripts/check-stage6-manual-smoke.sh
+# Local PR gates
+./scripts/lint.sh
 ./tests/scripts/check-refactor-line-gate.sh
 ./tests/scripts/run-unit-all.sh
-npm ci --prefix webui && npm run build --prefix webui
+npm run build --prefix webui
+
+# Live end-to-end tests (real accounts, full request/response logs)
+./tests/scripts/run-live.sh
 ```
 
 ## Release Artifact Automation (GitHub Actions)

SECURITY.md

@@ -0,0 +1,65 @@
+# Security Policy
+
+## Supported Versions
+
+**Only the latest version** receives security updates.  
+If you are using an older version, please upgrade to the latest release.
+
+| Version        | Supported          |
+| -------------- | ------------------ |
+| latest         | :white_check_mark: |
+| < latest       | :x:                |
+
+> **Why?** This project is maintained by a single developer. Keeping only one active version ensures fast response times and avoids legacy maintenance overhead.
+
+## What is a Security Vulnerability?
+
+A **security vulnerability** is a bug that can be exploited to compromise:
+- Data confidentiality (e.g., leaking secrets, user data)
+- Data integrity (e.g., unauthorized modification)
+- System availability (e.g., remote crash, denial of service)
+- Privilege escalation (e.g., normal user gains admin rights)
+
+**Examples**: SQL injection, command injection, path traversal, authentication bypass, insecure deserialization, sensitive data exposure.
+
+**What is NOT a security vulnerability?**  
+Regular bugs like crashes (without exploit potential), incorrect return values, performance issues, missing features, or documentation typos. Please report those via **GitHub Issues** publicly.
+
+## Reporting a Vulnerability
+
+If you believe you have found a security vulnerability, **please do NOT open a public issue**.
+
+Instead, send an email to: **cjackhwang@qq.com**
+
+Please include as much as possible:
+- A clear description of the issue
+- Steps to reproduce (code / input / environment)
+- Potential impact (what could an attacker do?)
+- Suggested fix (if any)
+
+You can expect:
+- **Initial response** within 3 business days (acknowledgment)
+- **Confirmation or clarification** within 7 days
+- **Fix or decision** within 14 days (depending on complexity)
+
+## What to Expect After Reporting
+
+| Outcome            | What happens |
+| ------------------ | ------------- |
+| **Accepted**       | I will develop a fix, release a patch version, and may credit you in the release notes (unless you prefer anonymity). |
+| **Declined**       | I will explain why (e.g., not a security issue, already fixed, out of scope, or requires a larger redesign). |
+| **Need more info** | I will ask follow-up questions. If no response within 14 days, the report may be considered stale. |
+
+## Disclosure Policy
+
+- Vulnerabilities will be **fixed privately** and then released as a new version.
+- After the fix is released, I will typically publish a short security advisory (via GitHub Security Advisories) without revealing exploit details.
+- Public disclosure can be coordinated if you request it.
+
+## Recognition
+
+I appreciate security researchers who follow responsible disclosure. Contributors who report valid, previously unknown vulnerabilities may be acknowledged in the project's README or release notes (unless they prefer to stay anonymous).
+
+---
+
+*Thank you for helping keep this project safe!*

VERSION

@@ -1 +1 @@
-3.6.0
+4.0.0

config.example.json

@@ -38,9 +38,10 @@
     }
   ],
   "model_aliases": {
-    "gpt-4o": "deepseek-chat",
-    "gpt-5-codex": "deepseek-reasoner",
-    "o3": "deepseek-reasoner"
+    "gpt-4o": "deepseek-v4-flash",
+    "gpt-5.5": "deepseek-v4-flash",
+    "gpt-5.3-codex": "deepseek-v4-pro",
+    "o3": "deepseek-v4-pro"
   },
   "compat": {
     "wide_input_strict_output": true,
@@ -56,10 +57,6 @@
   "embeddings": {
     "provider": "deterministic"
   },
-  "claude_mapping": {
-    "fast": "deepseek-chat",
-    "slow": "deepseek-reasoner"
-  },
   "admin": {
     "jwt_expire_hours": 24
   },

docs/ARCHITECTURE.en.md

@@ -4,9 +4,9 @@ Language: [中文](ARCHITECTURE.md) | [English](ARCHITECTURE.en.md)
 
 > This file is the single architecture source for directory layout, module boundaries, and execution flow.
 
-## 1. Top-level Layout (expanded)
+## 1. Top-level Layout (core directories)
 
-> Notes: this is the **fully expanded** project directory list (excluding metadata/dependency dirs such as `.git/` and `webui/node_modules/`), with each folder annotated by purpose.
+> Notes: this lists the main business directories (excluding metadata/dependency dirs such as `.git/` and `webui/node_modules/`), with each folder annotated by purpose. Newly added directories should be verified from the code tree rather than treated as a per-file inventory here.
 
 ```text
 ds2api/
@@ -21,34 +21,46 @@ ds2api/
 ├── docs/                                 # Project documentation
 ├── internal/                             # Core implementation (non-public packages)
 │   ├── account/                          # Account pool, inflight slots, waiting queue
-│   ├── adapter/                          # Multi-protocol adapters
-│   │   ├── claude/                       # Claude protocol adapter
-│   │   ├── gemini/                       # Gemini protocol adapter
-│   │   └── openai/                       # OpenAI adapter and shared execution core
-│   ├── admin/                            # Admin API (config/accounts/ops)
 │   ├── auth/                             # Auth/JWT/credential resolution
+│   ├── chathistory/                      # Server-side conversation history storage/query
 │   ├── claudeconv/                       # Claude message conversion helpers
 │   ├── compat/                           # Compatibility and regression helpers
 │   ├── config/                           # Config loading/validation/hot reload
-│   ├── deepseek/                         # DeepSeek upstream client capabilities
+│   ├── deepseek/                         # DeepSeek upstream client/protocol/transport
+│   │   ├── client/                       # Login/session/completion/upload/delete calls
+│   │   ├── protocol/                     # DeepSeek URLs, constants, skip path/pattern
 │   │   └── transport/                    # DeepSeek transport details
 │   ├── devcapture/                       # Dev capture and troubleshooting
 │   ├── format/                           # Response formatting layer
 │   │   ├── claude/                       # Claude output formatting
 │   │   └── openai/                       # OpenAI output formatting
+│   ├── httpapi/                          # HTTP surfaces: OpenAI/Claude/Gemini/Admin
+│   │   ├── admin/                        # Admin API root assembly and resource packages
+│   │   ├── claude/                       # Claude HTTP protocol adapter
+│   │   ├── gemini/                       # Gemini HTTP protocol adapter
+│   │   └── openai/                       # OpenAI HTTP surface
+│   │       ├── chat/                     # Chat Completions execution entrypoint
+│   │       ├── responses/                # Responses API and response store
+│   │       ├── files/                    # Files API and inline-file preprocessing
+│   │       ├── embeddings/               # Embeddings API
+│   │       ├── history/                  # OpenAI history split
+│   │       └── shared/                   # OpenAI HTTP errors/models/tool formatting
 │   ├── js/                               # Node runtime related logic
 │   │   ├── chat-stream/                  # Node streaming bridge
 │   │   ├── helpers/                      # JS helper modules
 │   │   │   └── stream-tool-sieve/        # JS implementation of tool sieve
 │   │   └── shared/                       # Shared semantics between Go/Node
 │   ├── prompt/                           # Prompt composition
+│   ├── promptcompat/                     # API request -> DeepSeek web-chat plain-text compatibility
 │   ├── rawsample/                        # Raw sample read/write and management
 │   ├── server/                           # Router and middleware assembly
+│   │   └── data/                         # Router/runtime helper data
 │   ├── sse/                              # SSE parsing utilities
 │   ├── stream/                           # Unified stream consumption engine
 │   ├── testsuite/                        # Testsuite execution framework
 │   ├── textclean/                        # Text cleanup
 │   ├── toolcall/                         # Tool-call parsing and repair
+│   ├── toolstream/                       # Go streaming tool-call anti-leak and delta detection
 │   ├── translatorcliproxy/               # Cross-protocol translation bridge
 │   ├── util/                             # Shared utility helpers
 │   ├── version/                          # Version query/compare
@@ -90,34 +102,82 @@ ds2api/
 
 ```mermaid
 flowchart LR
-    C[Client/SDK] --> R[internal/server/router.go]
-    R --> OA[OpenAI Adapter]
-    R --> CA[Claude Adapter]
-    R --> GA[Gemini Adapter]
-    R --> AD[Admin API]
+    C[Client / SDK] --> R[internal/server/router.go]
 
-    CA --> BR[translatorcliproxy]
-    GA --> BR
-    BR --> CORE[internal/adapter/openai ChatCompletions]
-    OA --> CORE
+    subgraph HTTP[HTTP API surface]
+        OA[internal/httpapi/openai]
+        CHAT[openai/chat]
+        RESP[openai/responses]
+        FILES[openai/files + embeddings]
+        CA[internal/httpapi/claude]
+        GA[internal/httpapi/gemini]
+        AD[internal/httpapi/admin/*]
+        WEB[internal/webui static admin]
+    end
 
-    CORE --> AUTH[internal/auth + config key/account resolver]
-    CORE --> POOL[internal/account queue + concurrency]
-    CORE --> TOOL[internal/toolcall parser + sieve]
-    CORE --> DS[internal/deepseek client]
+    subgraph COMPAT[Prompt compatibility core]
+        PC[internal/promptcompat]
+        PROMPT[internal/prompt]
+        HIST[internal/httpapi/openai/history]
+    end
+
+    subgraph RUNTIME[Shared runtime]
+        AUTH[internal/auth]
+        POOL[internal/account queue + concurrency]
+        STREAM[internal/stream + internal/sse]
+        TOOL[internal/toolcall + internal/toolstream]
+        DS[internal/deepseek/client]
+        POW[pow + internal/deepseek/protocol]
+    end
+
+    subgraph NODE[Vercel Node Runtime]
+        NCS[api/chat-stream.js]
+        JS[internal/js/chat-stream + stream-tool-sieve]
+    end
+
+    R --> OA --> CHAT
+    OA --> RESP
+    OA --> FILES
+    R --> CA
+    R --> GA
+    R --> AD
+    R --> WEB
+    R -.Vercel stream.-> NCS
+
+    CA --> PC
+    GA --> PC
+    CHAT --> PC
+    RESP --> PC
+    PC --> PROMPT
+    PC -.long history.-> HIST
+    PC --> AUTH
+
+    NCS -.Go prepare/release.-> CHAT
+    NCS --> JS
+    JS --> TOOL
+
+    AUTH --> POOL
+    CHAT --> STREAM
+    RESP --> STREAM
+    STREAM --> TOOL
+    POOL --> DS
+    DS --> POW
     DS --> U[DeepSeek upstream]
 ```
 
 ## 3. Responsibilities in `internal/`
 
 - `internal/server`: router tree + middlewares (health, protocol routes, Admin/WebUI).
-- `internal/adapter/openai`: shared execution core (chat/responses/embeddings + tool semantics).
-- `internal/adapter/{claude,gemini}`: protocol wrappers only (no duplicated upstream execution).
+- `internal/httpapi/openai/*`: OpenAI HTTP surface split into chat, responses, files, embeddings, history, and shared packages; chat/responses share the promptcompat, stream, and toolcall semantics.
+- `internal/httpapi/{claude,gemini}`: protocol wrappers that normalize into the same prompt compatibility semantics without duplicating upstream execution.
+- `internal/promptcompat`: compatibility core for turning OpenAI/Claude/Gemini requests into DeepSeek web-chat plain-text context.
 - `internal/translatorcliproxy`: structure translation between Claude/Gemini and OpenAI.
-- `internal/deepseek`: upstream request/session/PoW/SSE handling.
-- `internal/stream` + `internal/sse`: stream parsing and incremental assembly.
-- `internal/toolcall`: XML/Markup-family tool-call parsing + anti-leak sieve (`<tool_call>` / `<function_call>` / `<invoke>` / `tool_use` / antml variants).
-- `internal/admin`: config/accounts/vercel sync/version/dev-capture endpoints.
+- `internal/deepseek/{client,protocol,transport}`: upstream requests, sessions, PoW adaptation, protocol constants, and transport details.
+- `internal/js/chat-stream` + `api/chat-stream.js`: Vercel Node streaming bridge; Go prepare/release owns auth, account lease, and completion payload assembly, while Node relays real-time SSE with Go-aligned finalization and tool sieve semantics.
+- `internal/stream` + `internal/sse`: Go stream parsing and incremental assembly.
+- `internal/toolcall` + `internal/toolstream`: canonical XML tool-call parsing + anti-leak sieve (the only executable format is `<tool_calls>` / `<invoke name="...">` / `<parameter name="...">`).
+- `internal/httpapi/admin/*`: Admin API root assembly plus auth/accounts/config/settings/proxies/rawsamples/vercel/history/devcapture/version resource packages.
+- `internal/chathistory`: server-side conversation history persistence, pagination, detail lookup, and retention policy.
 - `internal/config`: config loading/validation + runtime settings hot-reload.
 - `internal/account`: managed account pool, inflight slots, waiting queue.
 

docs/ARCHITECTURE.md

@@ -4,9 +4,9 @@
 
 > 本文档用于集中维护“代码目录结构 + 模块边界 + 主链路调用关系”。
 
-## 1. 顶层目录结构（展开）
+## 1. 顶层目录结构（核心目录）
 
-> 说明：以下为仓库内业务相关目录的**完整展开**（排除 `.git/` 与 `webui/node_modules/` 这类依赖/元数据目录），并标注每个文件夹作用。
+> 说明：以下为仓库内主要业务目录（排除 `.git/` 与 `webui/node_modules/` 这类依赖/元数据目录），并标注每个文件夹作用。新增目录以代码为准，不要求在本文做逐文件展开。
 
 ```text
 ds2api/
@@ -21,34 +21,46 @@ ds2api/
 ├── docs/                                 # 项目文档目录
 ├── internal/                             # 核心业务实现（不对外暴露）
 │   ├── account/                          # 账号池、并发槽位、等待队列
-│   ├── adapter/                          # 多协议适配层
-│   │   ├── claude/                       # Claude 协议适配
-│   │   ├── gemini/                       # Gemini 协议适配
-│   │   └── openai/                       # OpenAI 协议与统一执行核心
-│   ├── admin/                            # Admin API（配置/账号/运维）
 │   ├── auth/                             # 鉴权/JWT/凭证解析
+│   ├── chathistory/                      # 服务器端对话记录存储与查询
 │   ├── claudeconv/                       # Claude 消息格式转换工具
 │   ├── compat/                           # 兼容性辅助与回归支持
 │   ├── config/                           # 配置加载、校验、热更新
-│   ├── deepseek/                         # DeepSeek 上游客户端能力
+│   ├── deepseek/                         # DeepSeek 上游 client/protocol/transport
+│   │   ├── client/                       # 登录、会话、completion、上传/删除等上游调用
+│   │   ├── protocol/                     # DeepSeek URL、常量、skip path/pattern
 │   │   └── transport/                    # DeepSeek 传输层细节
 │   ├── devcapture/                       # 开发抓包与调试采集
 │   ├── format/                           # 响应格式化层
 │   │   ├── claude/                       # Claude 输出格式化
 │   │   └── openai/                       # OpenAI 输出格式化
+│   ├── httpapi/                          # HTTP surface：OpenAI/Claude/Gemini/Admin
+│   │   ├── admin/                        # Admin API 根装配与资源子包
+│   │   ├── claude/                       # Claude HTTP 协议适配
+│   │   ├── gemini/                       # Gemini HTTP 协议适配
+│   │   └── openai/                       # OpenAI HTTP surface
+│   │       ├── chat/                     # Chat Completions 执行入口
+│   │       ├── responses/                # Responses API 与 response store
+│   │       ├── files/                    # Files API 与 inline file 预处理
+│   │       ├── embeddings/               # Embeddings API
+│   │       ├── history/                  # OpenAI history split
+│   │       └── shared/                   # OpenAI HTTP 公共错误/模型/工具格式
 │   ├── js/                               # Node Runtime 相关逻辑
 │   │   ├── chat-stream/                  # Node 流式输出桥接
 │   │   ├── helpers/                      # JS 辅助函数
 │   │   │   └── stream-tool-sieve/        # Tool sieve JS 实现
 │   │   └── shared/                       # Go/Node 共用语义片段
 │   ├── prompt/                           # Prompt 组装
+│   ├── promptcompat/                     # API 请求到 DeepSeek 网页纯文本上下文兼容层
 │   ├── rawsample/                        # raw sample 读写与管理
 │   ├── server/                           # 路由与中间件装配
+│   │   └── data/                         # 路由/运行时辅助数据
 │   ├── sse/                              # SSE 解析工具
 │   ├── stream/                           # 统一流式消费引擎
 │   ├── testsuite/                        # 测试集执行框架
 │   ├── textclean/                        # 文本清洗
 │   ├── toolcall/                         # 工具调用解析与修复
+│   ├── toolstream/                       # Go 流式 tool call 防泄漏与增量检测
 │   ├── translatorcliproxy/               # 多协议互转桥
 │   ├── util/                             # 通用工具函数
 │   ├── version/                          # 版本查询/比较
@@ -90,34 +102,82 @@ ds2api/
 
 ```mermaid
 flowchart LR
-    C[Client/SDK] --> R[internal/server/router.go]
-    R --> OA[OpenAI Adapter]
-    R --> CA[Claude Adapter]
-    R --> GA[Gemini Adapter]
-    R --> AD[Admin API]
+    C[Client / SDK] --> R[internal/server/router.go]
 
-    CA --> BR[translatorcliproxy]
-    GA --> BR
-    BR --> CORE[internal/adapter/openai ChatCompletions]
-    OA --> CORE
+    subgraph HTTP[HTTP API surface]
+        OA[internal/httpapi/openai]
+        CHAT[openai/chat]
+        RESP[openai/responses]
+        FILES[openai/files + embeddings]
+        CA[internal/httpapi/claude]
+        GA[internal/httpapi/gemini]
+        AD[internal/httpapi/admin/*]
+        WEB[internal/webui static admin]
+    end
 
-    CORE --> AUTH[internal/auth + config key/account resolver]
-    CORE --> POOL[internal/account queue + concurrency]
-    CORE --> TOOL[internal/toolcall parser + sieve]
-    CORE --> DS[internal/deepseek client]
+    subgraph COMPAT[Prompt compatibility core]
+        PC[internal/promptcompat]
+        PROMPT[internal/prompt]
+        HIST[internal/httpapi/openai/history]
+    end
+
+    subgraph RUNTIME[Shared runtime]
+        AUTH[internal/auth]
+        POOL[internal/account queue + concurrency]
+        STREAM[internal/stream + internal/sse]
+        TOOL[internal/toolcall + internal/toolstream]
+        DS[internal/deepseek/client]
+        POW[pow + internal/deepseek/protocol]
+    end
+
+    subgraph NODE[Vercel Node Runtime]
+        NCS[api/chat-stream.js]
+        JS[internal/js/chat-stream + stream-tool-sieve]
+    end
+
+    R --> OA --> CHAT
+    OA --> RESP
+    OA --> FILES
+    R --> CA
+    R --> GA
+    R --> AD
+    R --> WEB
+    R -.Vercel stream.-> NCS
+
+    CA --> PC
+    GA --> PC
+    CHAT --> PC
+    RESP --> PC
+    PC --> PROMPT
+    PC -.长历史.-> HIST
+    PC --> AUTH
+
+    NCS -.Go prepare/release.-> CHAT
+    NCS --> JS
+    JS --> TOOL
+
+    AUTH --> POOL
+    CHAT --> STREAM
+    RESP --> STREAM
+    STREAM --> TOOL
+    POOL --> DS
+    DS --> POW
     DS --> U[DeepSeek upstream]
 ```
 
 ## 3. internal/ 子模块职责
 
 - `internal/server`：路由树和中间件挂载（健康检查、协议入口、Admin/WebUI）。
-- `internal/adapter/openai`：统一执行内核（chat/responses/embeddings 与 tool calling 语义）。
-- `internal/adapter/{claude,gemini}`：协议输入输出适配，不重复实现上游调用逻辑。
+- `internal/httpapi/openai/*`：OpenAI HTTP surface，按 chat、responses、files、embeddings、history、shared 拆分；chat/responses 共享 promptcompat、stream、toolcall 等核心语义。
+- `internal/httpapi/{claude,gemini}`：协议输入输出适配，归一到同一套 prompt compatibility 语义，不重复实现上游调用逻辑。
+- `internal/promptcompat`：OpenAI/Claude/Gemini 请求到 DeepSeek 网页纯文本上下文的兼容内核。
 - `internal/translatorcliproxy`：Claude/Gemini 与 OpenAI 结构互转。
-- `internal/deepseek`：上游请求、会话、PoW、SSE 消费。
-- `internal/stream` + `internal/sse`：流式解析与增量处理。
-- `internal/toolcall`：以 XML/Markup 家族为核心的工具调用解析与防泄漏筛分（`<tool_call>` / `<function_call>` / `<invoke>` / `tool_use` / antml 变体）。
-- `internal/admin`：配置管理、账号管理、Vercel 同步、版本检查、开发抓包。
+- `internal/deepseek/{client,protocol,transport}`：上游请求、会话、PoW 适配、协议常量与传输层。
+- `internal/js/chat-stream` + `api/chat-stream.js`：Vercel Node 流式桥；Go prepare/release 管理鉴权、账号租约和 completion payload，Node 侧负责实时 SSE 转发并保持 Go 对齐的终结态和 tool sieve 语义。
+- `internal/stream` + `internal/sse`：Go 流式解析与增量处理。
+- `internal/toolcall` + `internal/toolstream`：canonical XML 工具调用解析与防泄漏筛分（唯一可执行格式：`<tool_calls>` / `<invoke name="...">` / `<parameter name="...">`）。
+- `internal/httpapi/admin/*`：Admin API 根装配与 auth/accounts/config/settings/proxies/rawsamples/vercel/history/devcapture/version 等资源子包。
+- `internal/chathistory`：服务器端对话记录持久化、分页、单条详情和保留策略。
 - `internal/config`：配置加载、校验、运行时 settings 热更新。
 - `internal/account`：托管账号池、并发槽位、等待队列。
 

docs/CONTRIBUTING.en.md

@@ -59,10 +59,12 @@ docker-compose -f docker-compose.dev.yml up
 
 | Language | Standards |
 | --- | --- |
-| **Go** | Run `./scripts/lint.sh` (gofmt + golangci-lint) and ensure `go test ./...` passes before committing |
+| **Go** | Run `gofmt -w` after editing Go files; before committing, run `./scripts/lint.sh` (format check + golangci-lint) |
 | **JavaScript/React** | Follow existing project style (functional components) |
 | **Commit messages** | Use semantic prefixes: `feat:`, `fix:`, `docs:`, `refactor:`, `style:`, `perf:`, `chore:` |
 
+Do not silently ignore cleanup errors from I/O-style calls such as `Close`, `Flush`, or `Sync`; return them when possible, otherwise log them explicitly.
+
 ## Submitting a PR
 
 1. Fork the repo
@@ -85,10 +87,13 @@ Manually build WebUI to `static/admin/`:
 ## Running Tests
 
 ```bash
-# Go + Node unit tests (recommended)
+# Local PR gates (kept aligned with the quality-gates workflow)
+./scripts/lint.sh
+./tests/scripts/check-refactor-line-gate.sh
 ./tests/scripts/run-unit-all.sh
+npm run build --prefix webui
 
-# End-to-end live tests (real accounts)
+# End-to-end live tests (real accounts; recommended for releases or high-risk changes)
 ./tests/scripts/run-live.sh
 ```
 

docs/CONTRIBUTING.md

@@ -59,10 +59,12 @@ docker-compose -f docker-compose.dev.yml up
 
 | 语言 | 规范 |
 | --- | --- |
-| **Go** | 提交前运行 `./scripts/lint.sh`（包含 gofmt+golangci-lint）并确保 `go test ./...` 通过 |
+| **Go** | 修改 Go 文件后运行 `gofmt -w`；提交前运行 `./scripts/lint.sh`（包含格式化检查和 golangci-lint） |
 | **JavaScript/React** | 保持现有代码风格（函数组件） |
 | **提交信息** | 使用语义化前缀：`feat:`、`fix:`、`docs:`、`refactor:`、`style:`、`perf:`、`chore:` |
 
+I/O 类清理调用（如 `Close`、`Flush`、`Sync`）的错误不要直接忽略；无法向上返回时请显式记录日志。
+
 ## 提交 PR
 
 1. Fork 仓库
@@ -85,10 +87,13 @@ docker-compose -f docker-compose.dev.yml up
 ## 运行测试
 
 ```bash
-# Go + Node 单元测试（推荐）
+# PR 本地门禁（与 quality-gates 工作流保持一致）
+./scripts/lint.sh
+./tests/scripts/check-refactor-line-gate.sh
 ./tests/scripts/run-unit-all.sh
+npm run build --prefix webui
 
-# 端到端全链路测试（真实账号）
+# 端到端全链路测试（真实账号，发布或高风险改动时建议执行）
 ./tests/scripts/run-live.sh
 ```
 

docs/DEPLOY.en.md

@@ -259,12 +259,13 @@ VERCEL_TEAM_ID=team_xxxxxxxxxxxx   # optional for personal accounts
 | `DS2API_ENV_WRITEBACK` | When `DS2API_CONFIG_JSON` is present, auto-write to `DS2API_CONFIG_PATH` and switch to file-backed mode after success (`1/true/yes/on`) | Disabled |
 | `DS2API_VERCEL_INTERNAL_SECRET` | Hybrid streaming internal auth | Falls back to `DS2API_ADMIN_KEY` |
 | `DS2API_VERCEL_STREAM_LEASE_TTL_SECONDS` | Stream lease TTL | `900` |
+| `DS2API_RAW_STREAM_SAMPLE_ROOT` | Raw stream sample root for saving/reading samples | `tests/raw_stream_samples` |
 | `VERCEL_TOKEN` | Vercel sync token | — |
 | `VERCEL_PROJECT_ID` | Vercel project ID | — |
 | `VERCEL_TEAM_ID` | Vercel team ID | — |
 | `DS2API_VERCEL_PROTECTION_BYPASS` | Deployment protection bypass for internal Node→Go calls | — |
 
-### 3.3 Vercel Architecture
+### 3.4 Vercel Architecture
 
 ```text
 Request ──────┐
@@ -300,13 +301,14 @@ Vercel Go Runtime applies platform-level response buffering, so this project use
 
 - `api/chat-stream.js` falls back to Go entry (`?__go=1`) for non-stream requests only
 - Streaming requests (including requests with `tools`) stay on the Node path and use Go-aligned tool-call anti-leak handling
+- The Node stream path also mirrors Go finalization semantics: empty visible output returns the same shaped error SSE, and empty `content_filter` returns a `content_filter` error
 - WebUI non-stream test calls `?__go=1` directly to avoid Node hop timeout on long requests
 
 #### Function Duration
 
 `vercel.json` sets `maxDuration: 300` for both `api/chat-stream.js` and `api/index.go` (subject to your Vercel plan limits).
 
-### 3.4 Vercel Troubleshooting
+### 3.5 Vercel Troubleshooting
 
 #### Go Build Failure
 
@@ -350,7 +352,7 @@ If API responses return Vercel HTML `Authentication Required`:
 - **Option B**: Add `x-vercel-protection-bypass` header to requests
 - **Option C**: Set `VERCEL_AUTOMATION_BYPASS_SECRET` (or `DS2API_VERCEL_PROTECTION_BYPASS`) for internal Node→Go calls
 
-### 3.5 Build Artifacts Not Committed
+### 3.6 Build Artifacts Not Committed
 
 - `static/admin` directory is not in Git
 - Vercel / Docker automatically generate WebUI assets during build
@@ -546,7 +548,7 @@ curl -s -o /dev/null -w "%{http_code}" http://127.0.0.1:5001/admin
 curl http://127.0.0.1:5001/v1/chat/completions \
   -H "Authorization: Bearer your-api-key" \
   -H "Content-Type: application/json" \
-  -d '{"model":"deepseek-chat","messages":[{"role":"user","content":"hello"}]}'
+  -d '{"model":"deepseek-v4-flash","messages":[{"role":"user","content":"hello"}]}'
 ```
 
 ---
@@ -577,4 +579,4 @@ The testsuite automatically performs:
 - ✅ Live scenario verification (OpenAI/Claude/Admin/concurrency/toolcall/streaming)
 - ✅ Full request/response artifact logging for debugging
 
-For detailed testsuite documentation, see [TESTING.md](TESTING.md).
+For detailed testsuite documentation, see [TESTING.md](TESTING.md). The fixed local PR gates are listed in [TESTING.md](TESTING.md#pr-门禁--pr-gates).

docs/DEPLOY.md

@@ -258,7 +258,8 @@ VERCEL_TEAM_ID=team_xxxxxxxxxxxx   # 个人账号可留空
 | `DS2API_GLOBAL_MAX_INFLIGHT` | 全局并发上限 | `recommended_concurrency` |
 | `DS2API_ENV_WRITEBACK` | 检测到 `DS2API_CONFIG_JSON` 时自动写入 `DS2API_CONFIG_PATH`，并在成功后转为文件模式（`1/true/yes/on`） | 关闭 |
 | `DS2API_VERCEL_INTERNAL_SECRET` | 混合流式内部鉴权 | 回退用 `DS2API_ADMIN_KEY` |
-| `DS2API_VERCEL_STREAM_LEASE_TTL_SECONDS` | 流式 lease TTL | 默认与 `responses.store_ttl_seconds` 同步，若未设置则为 `900` |
+| `DS2API_VERCEL_STREAM_LEASE_TTL_SECONDS` | 流式 lease TTL | `900` |
+| `DS2API_RAW_STREAM_SAMPLE_ROOT` | raw stream 样本保存/读取根目录 | `tests/raw_stream_samples` |
 | `VERCEL_TOKEN` | Vercel 同步 token | — |
 | `VERCEL_PROJECT_ID` | Vercel 项目 ID | — |
 | `VERCEL_TEAM_ID` | Vercel 团队 ID | — |
@@ -274,7 +275,7 @@ VERCEL_TEAM_ID=team_xxxxxxxxxxxx   # 个人账号可留空
 
 详细说明参见 [API.md](../API.md#admin-接口) 中 `/admin/settings` 部分。
 
-### 3.3 Vercel 架构说明
+### 3.4 Vercel 架构说明
 
 ```text
 请求 ─────┐
@@ -310,13 +311,14 @@ api/index.go  api/chat-stream.js
 
 - `api/chat-stream.js` 仅对非流式请求回退到 Go 入口（`?__go=1`）
 - 流式请求（包括带 `tools`）走 Node 路径，并执行与 Go 对齐的 tool-call 防泄漏处理
+- Node 流式路径同时对齐 Go 的终结态语义：空可见输出会返回同形状错误 SSE，空 `content_filter` 会返回 `content_filter` 错误
 - WebUI 的"非流式测试"直接请求 `?__go=1`，避免 Node 中转造成长请求超时
 
 #### 函数时长
 
 `vercel.json` 已将 `api/chat-stream.js` 与 `api/index.go` 的 `maxDuration` 设为 `300`（受 Vercel 套餐上限约束）。
 
-### 3.4 Vercel 常见报错排查
+### 3.5 Vercel 常见报错排查
 
 #### Go 构建失败
 
@@ -360,7 +362,7 @@ No Output Directory named "public" found after the Build completed.
 - **方案 B**：请求中添加 `x-vercel-protection-bypass` 头
 - **方案 C**：设置 `VERCEL_AUTOMATION_BYPASS_SECRET`（或 `DS2API_VERCEL_PROTECTION_BYPASS`），仅影响内部 Node→Go 调用
 
-### 3.5 仓库不提交构建产物
+### 3.6 仓库不提交构建产物
 
 - `static/admin` 目录不在 Git 中
 - Vercel / Docker 构建阶段自动生成 WebUI 静态文件
@@ -556,7 +558,7 @@ curl -s -o /dev/null -w "%{http_code}" http://127.0.0.1:5001/admin
 curl http://127.0.0.1:5001/v1/chat/completions \
   -H "Authorization: Bearer your-api-key" \
   -H "Content-Type: application/json" \
-  -d '{"model":"deepseek-chat","messages":[{"role":"user","content":"hello"}]}'
+  -d '{"model":"deepseek-v4-flash","messages":[{"role":"user","content":"hello"}]}'
 ```
 
 ---
@@ -587,4 +589,4 @@ go run ./cmd/ds2api-tests \
 - ✅ 真实调用场景验证（OpenAI/Claude/Admin/并发/toolcall/流式）
 - ✅ 全量请求与响应日志落盘（用于故障复盘）
 
-详细测试集说明参阅 [TESTING.md](TESTING.md)。
+详细测试集说明参阅 [TESTING.md](TESTING.md)。PR 前的固定本地门禁以 [TESTING.md](TESTING.md#pr-门禁--pr-gates) 为准。

docs/README.md

@@ -15,14 +15,17 @@
 
 ### 专题文档
 
+- [API -> 网页对话纯文本兼容主链路说明](./prompt-compatibility.md)
 - [Tool Calling 统一语义](./toolcall-semantics.md)
 - [DeepSeek SSE 行为结构说明（逆向观察）](./DeepSeekSSE行为结构说明-2026-04-05.md)
 
 ### 文档维护约定
 
+- 文档更新必须以实际代码实现为依据：总路由装配看 `internal/server/router.go`，协议/resource 路由看 `internal/httpapi/*/**/routes.go` 与 `internal/httpapi/admin/handler.go`，配置默认值看 `internal/config/*`，模型/alias 看 `internal/config/models.go`，prompt 兼容链路看 `docs/prompt-compatibility.md` 列出的代码入口。
 - `README.MD` / `README.en.md`：面向首次接触用户，保留“是什么 + 怎么快速跑起来”。
 - `docs/ARCHITECTURE*.md`：面向开发者，集中维护项目结构、模块职责与调用链。
 - `API*.md`：面向客户端接入者，聚焦接口行为、鉴权和示例。
+- `docs/prompt-compatibility.md`：面向维护者，集中维护“API -> 网页对话纯文本上下文”的统一兼容语义；相关行为修改时必须同步更新。
 - 其他 `docs/*.md`：主题化说明，避免在多个文档重复粘贴同一段内容。
 
 ---
@@ -42,12 +45,15 @@ Recommended reading order:
 
 ### Topical docs
 
+- [API -> pure-text web-chat compatibility pipeline](./prompt-compatibility.md)
 - [Tool-calling unified semantics](./toolcall-semantics.md)
 - [DeepSeek SSE behavior notes (reverse-engineered)](./DeepSeekSSE行为结构说明-2026-04-05.md)
 
 ### Maintenance conventions
 
+- Documentation updates must be grounded in the actual implementation: root routing lives in `internal/server/router.go`, protocol/resource routes live in `internal/httpapi/*/**/routes.go` and `internal/httpapi/admin/handler.go`, config defaults in `internal/config/*`, models/aliases in `internal/config/models.go`, and the prompt compatibility pipeline in the code entrypoints listed by `docs/prompt-compatibility.md`.
 - `README.MD` / `README.en.md`: onboarding-oriented (“what + quick start”).
 - `docs/ARCHITECTURE*.md`: developer-oriented source of truth for module boundaries and execution flow.
 - `API*.md`: integration-oriented behavior/contracts.
+- `docs/prompt-compatibility.md`: maintainer-oriented source of truth for the “API -> pure-text web-chat context” compatibility flow; update it whenever related behavior changes.
 - Other `docs/*.md`: focused topics, avoid copy-pasting the same section into multiple files.

docs/TESTING.md

@@ -20,6 +20,25 @@ Node 单元测试脚本会先做 `node --check` 语法门禁，再以 `--test-co
 
 ---
 
+## PR 门禁 | PR Gates
+
+打开或更新 PR 前，按 `.github/workflows/quality-gates.yml` 的同等本地门禁执行：
+
+```bash
+./scripts/lint.sh
+./tests/scripts/check-refactor-line-gate.sh
+./tests/scripts/run-unit-all.sh
+npm run build --prefix webui
+```
+
+说明：
+
+- `./scripts/lint.sh` 会运行 Go 格式化检查和 `golangci-lint`；修改 Go 文件后仍建议先执行 `gofmt -w <files>`。
+- `run-unit-all.sh` 串行调用 Go 与 Node 单元测试入口。
+- `run-live.sh` 是真实账号端到端测试，适合作为发布或高风险改动后的补充验证，不属于每次 PR 的固定本地门禁。
+
+---
+
 ## 快速开始 | Quick Start
 
 ### 单元测试 | Unit Tests
@@ -39,7 +58,7 @@ Node 单元测试脚本会先做 `node --check` 语法门禁，再以 `--test-co
 ./tests/scripts/check-refactor-line-gate.sh
 ./tests/scripts/check-node-split-syntax.sh
 
-# 发布阻断：阶段 6 手工烟测签字检查（默认读取 plans/stage6-manual-smoke.md）
+# 历史阶段门禁：阶段 6 手工烟测签字检查（默认读取 plans/stage6-manual-smoke.md）
 ./tests/scripts/check-stage6-manual-smoke.sh
 ```
 
@@ -190,8 +209,8 @@ go test -v -run TestParseToolCallsWithDeepSeekHallucination ./internal/toolcall/
 # 运行 format 相关测试
 go test -v ./internal/format/...
 
-# 运行 adapter 相关测试
-go test -v ./internal/adapter/openai/...
+# 运行 HTTP API 相关测试
+go test -v ./internal/httpapi/openai/...
 ```
 
 ### 调试 Tool Call 问题 | Debugging Tool Call Issues

docs/prompt-compatibility.md

@@ -0,0 +1,402 @@
+# API -> 网页对话纯文本兼容主链路说明
+
+文档导航：[总览](../README.MD) / [架构说明](./ARCHITECTURE.md) / [接口文档](../API.md) / [测试指南](./TESTING.md)
+
+> 本文档是 DS2API“把 OpenAI / Claude / Gemini 风格 API 请求兼容成 DeepSeek 网页对话纯文本上下文”的专项说明。
+> 这是项目最重要的兼容产物之一。凡是修改消息标准化、tool prompt 注入、tool history 保留、文件引用、history split、下游 completion payload 组装等行为，都必须同步更新本文档。
+
+## 1. 核心结论
+
+DS2API 当前的核心思路，不是把客户端传来的 `messages`、`tools`、`attachments` 原样转发给下游。
+
+而是把这些高层 API 语义，统一压缩成 DeepSeek 网页对话更容易理解的三类输入：
+
+1. `prompt`
+   一个单字符串，里面带有角色标记、system 指令、历史消息、assistant reasoning 标签、历史 tool call XML 等。
+2. `ref_file_ids`
+   一个文件引用数组，承载附件、inline 上传文件，以及必要时被拆出去的历史文件。
+3. 控制位
+   例如 `thinking_enabled`、`search_enabled`、部分 passthrough 参数。
+
+也就是说，项目最重要的兼容动作，是把“结构化 API 会话”翻译成“网页对话纯文本上下文 + 文件引用”。
+
+## 2. 为什么这是核心产物
+
+因为对下游来说，真正稳定的输入面不是 OpenAI/Claude/Gemini 的原生 schema，而是：
+
+- 一段连续的对话 prompt
+- 一组可引用文件
+- 少量开关位
+
+这也是为什么很多表面上看像“协议兼容”的代码，最终都会收敛到同一类逻辑：
+
+- 先把不同协议的消息统一成内部消息序列
+- 再把工具声明改写成 system prompt 文本
+- 再把历史 tool call / tool result 改写成 prompt 可见内容
+- 最后输出成 DeepSeek completion payload
+
+## 3. 统一心智模型
+
+当前主链路可以这样理解：
+
+```text
+客户端请求
+  -> HTTP API surface（OpenAI / Claude / Gemini）
+  -> promptcompat 统一消息标准化
+  -> tool prompt 注入
+  -> DeepSeek 风格 prompt 拼装
+  -> 文件收集 / inline 上传 / history split（OpenAI 链路）
+  -> completion payload
+  -> 下游网页对话接口
+```
+
+对应的关键代码入口：
+
+- OpenAI Chat / Responses：
+  [internal/promptcompat/request_normalize.go](../internal/promptcompat/request_normalize.go)
+- OpenAI prompt 组装：
+  [internal/promptcompat/prompt_build.go](../internal/promptcompat/prompt_build.go)
+- OpenAI 消息标准化：
+  [internal/promptcompat/message_normalize.go](../internal/promptcompat/message_normalize.go)
+- Claude 标准化：
+  [internal/httpapi/claude/standard_request.go](../internal/httpapi/claude/standard_request.go)
+- Claude 消息与 tool_use/tool_result 归一：
+  [internal/httpapi/claude/handler_utils.go](../internal/httpapi/claude/handler_utils.go)
+- Gemini 复用 OpenAI prompt builder：
+  [internal/httpapi/gemini/convert_request.go](../internal/httpapi/gemini/convert_request.go)
+- DeepSeek prompt 角色标记拼装：
+  [internal/prompt/messages.go](../internal/prompt/messages.go)
+- prompt 可见 tool history XML：
+  [internal/prompt/tool_calls.go](../internal/prompt/tool_calls.go)
+- completion payload：
+  [internal/promptcompat/standard_request.go](../internal/promptcompat/standard_request.go)
+
+## 4. 下游真正收到的东西
+
+在“完成标准化后”，下游 completion payload 的核心形态是：
+
+```json
+{
+  "chat_session_id": "session-id",
+  "model_type": "default",
+  "parent_message_id": null,
+  "prompt": "<｜begin▁of▁sentence｜>...",
+  "ref_file_ids": [
+    "file-history",
+    "file-systemprompt",
+    "file-other-attachment"
+  ],
+  "thinking_enabled": true,
+  "search_enabled": false
+}
+```
+
+重点是：
+
+- `prompt` 才是对话上下文主载体。
+- `ref_file_ids` 只承载文件引用，不承载普通文本消息。
+- `tools` 不会作为“原生工具 schema”直接下发给下游，而是被改写进 `prompt`。
+- OpenAI Chat / Responses 原生走统一 OpenAI 标准化与 DeepSeek payload 组装；Claude / Gemini 会尽量复用 OpenAI prompt/tool 语义，其中 Gemini 直接复用 `promptcompat.BuildOpenAIPromptForAdapter`，Claude 消息接口在可代理场景会转换为 OpenAI chat 形态再执行。
+- 客户端传入的 thinking / reasoning 开关会被归一到下游 `thinking_enabled`。Claude surface 没有 `thinking` 字段时按 Anthropic 语义视为关闭；Gemini `generationConfig.thinkingConfig.thinkingBudget` 会翻译成同一套 thinking 开关；关闭时即使上游返回 `response/thinking_content`，兼容层也不会把它当作可见正文输出。
+
+## 5. prompt 是怎么拼出来的
+
+### 5.1 角色标记
+
+最终 prompt 使用 DeepSeek 风格角色标记：
+
+- `<｜begin▁of▁sentence｜>`
+- `<｜System｜>`
+- `<｜User｜>`
+- `<｜Assistant｜>`
+- `<｜Tool｜>`
+- `<｜end▁of▁instructions｜>`
+- `<｜end▁of▁sentence｜>`
+- `<｜end▁of▁toolresults｜>`
+
+实现位置：
+[internal/prompt/messages.go](../internal/prompt/messages.go)
+
+### 5.2 thinking continuity 说明
+
+如果启用了 thinking，会在最前面额外插入一个 system block，提醒模型：
+
+- 继续既有会话，不要重开
+- earlier messages 是 binding context
+- 不要把最终回答只留在 reasoning 里
+
+这部分不是客户端原始消息，而是兼容层主动补进去的连续性契约。
+
+### 5.3 相邻同角色消息会合并
+
+在最终 `MessagesPrepareWithThinking` 中，相邻同 role 的消息会被合并成一个块，中间插入空行。
+
+这意味着：
+
+- prompt 中看到的是“合并后的 role block”
+- 不是客户端传来的逐条 message 原样排列
+
+## 6. tools 为什么是“文本注入”，不是原生下发
+
+当前项目把工具能力视为“prompt 约束的一部分”。
+
+具体做法：
+
+1. 把每个 tool 的名称、描述、参数 schema 序列化成文本。
+2. 拼成 `You have access to these tools:` 大段说明。
+3. 再附上统一的 XML tool call 格式约束。
+4. 把这整段内容并入 system prompt。
+
+工具调用正例仍只示范 canonical XML：`<tool_calls>` → `<invoke name="...">` → `<parameter name="...">`。
+提示词会额外强调：如果要调用工具，工具块的首个非空白字符必须就是 `<tool_calls>`，不能只输出 `</tool_calls>` 而漏掉 opening tag。
+正例中的工具名只会来自当前请求实际声明的工具；如果当前请求没有足够的已知工具形态，就省略对应的单工具、多工具或嵌套示例，避免把不可用工具名写进 prompt。
+对执行类工具，脚本内容必须进入执行参数本身：`Bash` / `execute_command` 使用 `command`，`exec_command` 使用 `cmd`；不要把脚本示范成 `path` / `content` 文件写入参数。
+
+OpenAI 路径实现：
+[internal/promptcompat/tool_prompt.go](../internal/promptcompat/tool_prompt.go)
+
+Claude 路径实现：
+[internal/httpapi/claude/handler_utils.go](../internal/httpapi/claude/handler_utils.go)
+
+统一工具调用格式模板：
+[internal/toolcall/tool_prompt.go](../internal/toolcall/tool_prompt.go)
+
+这也是项目“网页对话纯文本兼容”的关键设计：
+
+- tools 对下游来说，本质上是 prompt 内规则
+- 不是 native tool schema transport
+
+## 7. assistant 的 tool_calls / reasoning 如何保留
+
+### 7.1 reasoning 保留方式
+
+assistant 的 reasoning 会变成一个显式标签块：
+
+```text
+[reasoning_content]
+...
+[/reasoning_content]
+```
+
+然后再接可见回答正文。
+
+### 7.2 历史 tool_calls 保留方式
+
+assistant 历史 `tool_calls` 不会保留成 OpenAI 原生 JSON，而会转成 prompt 可见的 XML：
+
+```xml
+<tool_calls>
+  <invoke name="read_file">
+    <parameter name="path"><![CDATA[src/main.go]]></parameter>
+  </invoke>
+</tool_calls>
+```
+
+这也是当前项目里唯一受支持的 canonical tool-calling 形态；其他形态都会作为普通文本保留，不会作为可执行调用语法。
+例外是 parser 会对一个非常窄的模型失误做修复：如果 assistant 输出了 `<invoke ...>` ... `</tool_calls>`，但漏掉最前面的 opening `<tool_calls>`，解析阶段会补回 wrapper 后再尝试识别。
+
+这件事很重要，因为它决定了：
+
+- 历史工具调用在 prompt 中是“可见文本历史”
+- 不是“隐藏结构化元数据”
+
+实现位置：
+[internal/prompt/tool_calls.go](../internal/prompt/tool_calls.go)
+
+### 7.3 tool result 保留方式
+
+tool / function role 的结果会作为 `<｜Tool｜>...<｜end▁of▁toolresults｜>` 进入 prompt。
+
+如果 tool content 为空，当前会补成字符串 `"null"`，避免整个 tool turn 丢失。
+
+## 8. files、附件、systemprompt 文件的实际语义
+
+这里要明确区分两类东西：
+
+1. 文本型 system prompt
+   例如 OpenAI `developer` / `system` / Responses `instructions` / Claude top-level `system`
+   这类会进入 `prompt`。
+2. 文件型 systemprompt
+   例如通过附件、`input_file`、base64、data URL 上传的文件
+   这类不会直接内联进 `prompt`，而是进入 `ref_file_ids`。
+
+OpenAI 文件相关实现：
+
+- inline/base64/data URL 上传：
+  [internal/httpapi/openai/files/file_inline_upload.go](../internal/httpapi/openai/files/file_inline_upload.go)
+- 文件 ID 收集：
+  [internal/promptcompat/file_refs.go](../internal/promptcompat/file_refs.go)
+
+结论：
+
+- “systemprompt 文字”在 prompt 里
+- “systemprompt 文件”通常只在 `ref_file_ids` 里
+
+除非调用方自己把文件内容展开后再塞进 system/developer 文本，否则文件内容不会自动出现在 prompt 正文。
+
+## 9. 多轮历史为什么不会一直完整内联在 prompt
+
+history split 现在全局强制开启；旧配置中的 `history_split.enabled=false` 会被忽略。默认从第 2 个 user turn 起就可能触发，仍可通过 `history_split.trigger_after_turns` 调整触发阈值。
+
+相关实现：
+
+- 配置访问器：
+  [internal/config/store_accessors.go](../internal/config/store_accessors.go)
+- 历史拆分：
+  [internal/httpapi/openai/history/history_split.go](../internal/httpapi/openai/history/history_split.go)
+
+触发后行为：
+
+1. 旧历史消息被切出去。
+2. 旧历史会被重新序列化成一个文本文件。
+3. 真正上传的文件名固定是 `HISTORY.txt`。
+4. 文件内容内部会使用 `IGNORE` 这层包装名来闭合 DeepSeek 官网原生文件标记。
+5. 该文件上传后，其 `file_id` 会排在 `ref_file_ids` 最前面。
+6. live prompt 只保留：
+   - system / developer
+   - 最新 user turn 起的上下文
+
+历史文件内容不是普通自由文本，而是用同一套角色标记再次序列化出的 transcript：
+
+```text
+[uploaded filename]: HISTORY.txt
+[file content end]
+
+<｜begin▁of▁sentence｜><｜User｜>...<｜Assistant｜>...<｜Tool｜>...
+
+[file name]: IGNORE
+[file content begin]
+```
+
+所以“完整上下文”在当前实现里，其实通常分散在两处：
+
+- `prompt` 里的 live context
+- `ref_file_ids` 指向的 history transcript file
+
+## 10. 各协议入口的差异
+
+### 10.1 OpenAI Chat / Responses
+
+特点：
+
+- `developer` 会映射到 `system`
+- Responses `instructions` 会 prepend 为 system message
+- `tools` 会注入 system prompt
+- `attachments` / `input_file` / inline 文件会进入 `ref_file_ids`
+- history split 主要在这条链路里生效
+
+### 10.2 Claude Messages
+
+特点：
+
+- top-level `system` 优先作为系统提示
+- `tool_use` / `tool_result` 会被转换成统一的 assistant/tool 历史语义
+- `tools` 同样会被并进 system prompt
+- 常规执行通过 `internal/httpapi/claude/handler_messages.go` 转到 OpenAI chat 路径，模型 alias 会先解析成 DeepSeek 原生模型
+- 当前代码里没有像 OpenAI 那样完整的 `ref_file_ids` 附件链路
+
+### 10.3 Gemini
+
+特点：
+
+- `systemInstruction`、`contents.parts`、`functionCall`、`functionResponse` 会先归一
+- tools 会转成 OpenAI 风格 function schema
+- prompt 构建复用 OpenAI 的 `promptcompat.BuildOpenAIPromptForAdapter`
+- 未识别的非文本 part 会被安全序列化进 prompt，并对二进制/疑似 base64 内容做省略或截断处理
+
+也就是说，Gemini 在“最终 prompt 语义”上，尽量和 OpenAI 保持一致。
+
+## 11. 一份贴近真实的最终上下文示意
+
+假设用户发来一个多轮请求：
+
+- 有 system/developer 文本
+- 有 tools
+- 有一个文件型 systemprompt 附件
+- 有历史 assistant tool call / tool result
+- history split 已触发
+
+那么最终上下文更接近：
+
+```json
+{
+  "prompt": "<｜begin▁of▁sentence｜><｜System｜>continuity instructions...\\n\\n原 system / developer\\n\\nYou have access to these tools: ...<｜end▁of▁instructions｜><｜User｜>最新问题<｜Assistant｜>",
+  "ref_file_ids": [
+    "file-history-ignore",
+    "file-systemprompt",
+    "file-other-attachment"
+  ],
+  "thinking_enabled": true,
+  "search_enabled": false
+}
+```
+
+这正是“API 转网页对话纯文本”的核心成果：
+
+- 大部分结构化语义被压进 `prompt`
+- 文件保持文件
+- 历史必要时拆文件
+
+## 12. 修改时必须同步本文档的场景
+
+只要触碰以下任一类行为，就必须在同一提交或同一 PR 中更新本文档：
+
+- 角色映射变更
+- system / developer / instructions 合并规则变更
+- assistant reasoning 保留格式变更
+- assistant 历史 `tool_calls` 的 XML 呈现方式变更
+- tool result 注入方式变更
+- tool prompt 模板或 tool_choice 约束变更
+- inline 文件上传 / 文件引用收集规则变更
+- history split 触发条件、上传格式、`IGNORE` 包装格式变更
+- completion payload 字段语义变更
+- Claude / Gemini 对这套统一语义的复用关系变更
+
+优先检查这些文件：
+
+- `internal/promptcompat/request_normalize.go`
+- `internal/promptcompat/prompt_build.go`
+- `internal/promptcompat/message_normalize.go`
+- `internal/promptcompat/tool_prompt.go`
+- `internal/httpapi/openai/files/file_inline_upload.go`
+- `internal/promptcompat/file_refs.go`
+- `internal/httpapi/openai/history/history_split.go`
+- `internal/promptcompat/responses_input_normalize.go`
+- `internal/httpapi/claude/standard_request.go`
+- `internal/httpapi/claude/handler_utils.go`
+- `internal/httpapi/gemini/convert_request.go`
+- `internal/httpapi/gemini/convert_messages.go`
+- `internal/httpapi/gemini/convert_tools.go`
+- `internal/prompt/messages.go`
+- `internal/prompt/tool_calls.go`
+- `internal/promptcompat/standard_request.go`
+
+## 13. 建议的最小验证
+
+改动这条链路后，至少补齐或检查这些测试：
+
+- `go test ./internal/prompt/...`
+- `go test ./internal/httpapi/openai/...`
+- `go test ./internal/httpapi/claude/...`
+- `go test ./internal/httpapi/gemini/...`
+- `go test ./internal/util/...`
+
+如果改的是 tool call 相关兼容语义，还应同时检查：
+
+- `go test ./internal/toolcall/...`
+- `node --test tests/node/stream-tool-sieve.test.js`
+
+## 14. 文档同步约定
+
+本文档是这条兼容链路的专项说明。
+
+如果外部接口行为也变了，还应同步检查：
+
+- [API.md](../API.md)
+- [API.en.md](../API.en.md)
+- [docs/toolcall-semantics.md](./toolcall-semantics.md)
+
+原则是：
+
+- 内部主链路变化，至少更新本文档
+- 外部可见契约变化，再同步更新 API 文档

docs/toolcall-semantics.md

@@ -1,74 +1,75 @@
 # Tool call parsing semantics（Go/Node 统一语义）
 
-本文档描述当前代码中工具调用解析链路的**实际行为**（以 `internal/toolcall` 与 `internal/js/helpers/stream-tool-sieve` 为准）。
+本文档描述当前代码中的**实际行为**，以 `internal/toolcall`、`internal/toolstream` 与 `internal/js/helpers/stream-tool-sieve` 为准。
 
 文档导航：[总览](../README.MD) / [架构说明](./ARCHITECTURE.md) / [测试指南](./TESTING.md)
 
-## 1) 当前输出结构
+## 1) 当前唯一可执行格式
+
+当前版本只把下面这类 canonical XML 视为可执行工具调用：
+
+```xml
+<tool_calls>
+  <invoke name="read_file">
+    <parameter name="path"><![CDATA[README.MD]]></parameter>
+  </invoke>
+</tool_calls>
+```
+
+约束：
+
+- 必须有 `<tool_calls>...</tool_calls>` wrapper
+- 每个调用必须在 `<invoke name="...">...</invoke>` 内
+- 工具名必须放在 `invoke` 的 `name` 属性
+- 参数必须使用 `<parameter name="...">...</parameter>`
+
+兼容修复：
+
+- 如果模型漏掉 opening `<tool_calls>`，但后面仍输出了一个或多个 `<invoke ...>` 并以 `</tool_calls>` 收尾，Go 解析链路会在解析前补回缺失的 opening wrapper。
+- 这是一个针对常见模型失误的窄修复，不改变推荐输出格式；prompt 仍要求模型直接输出完整 canonical XML。
+
+## 2) 非 canonical 内容
+
+任何不满足上述 canonical XML 形态的内容，都会保留为普通文本，不会执行。一个例外是上一节提到的“缺失 opening `<tool_calls>`、但 closing `</tool_calls>` 仍存在”的窄修复场景。
+
+当前 parser 不把 allow-list 当作硬安全边界：即使传入了已声明工具名列表，XML 里出现未声明工具名时也会尽量解析并交给上层协议输出；真正的执行侧仍必须自行校验工具名和参数。
+
+## 3) 流式与防泄漏行为
+
+在流式链路中（Go / Node 一致）：
+
+- canonical `<tool_calls>` wrapper 会进入结构化捕获
+- 如果流里直接从 `<invoke ...>` 开始，但后面补上了 `</tool_calls>`，Go 流式筛分也会按缺失 opening wrapper 的修复路径尝试恢复
+- 已识别成功的工具调用不会再次回流到普通文本
+- 不符合新格式的块不会执行，并继续按原样文本透传
+- fenced code block 中的 XML 示例始终按普通文本处理
+
+## 4) 输出结构
 
 `ParseToolCallsDetailed` / `parseToolCallsDetailed` 返回：
 
-- `calls`：解析出的工具调用列表（`name` + `input`）。
-- `sawToolCallSyntax`：检测到工具调用语法特征时为 `true`。
-- `rejectedByPolicy`：当前实现固定为 `false`（预留字段）。
-- `rejectedToolNames`：当前实现固定为空数组（预留字段）。
+- `calls`：解析出的工具调用列表（`name` + `input`）
+- `sawToolCallSyntax`：检测到 canonical wrapper，或命中“缺失 opening wrapper 但可修复”的形态时会为 `true`
+- `rejectedByPolicy`：当前固定为 `false`
+- `rejectedToolNames`：当前固定为空数组
 
-> 当前 `filterToolCallsDetailed` 仅做结构清洗，不做 allow-list 工具名硬拒绝。
+## 5) 落地建议
 
-## 2) 解析范围（重点）
+1. Prompt 里只示范 canonical XML 语法。
+2. 上游客户端仍应直接输出 canonical XML；DS2API 只对“closing tag 在、opening tag 漏掉”的常见失误做窄修复，不会泛化接受其他旧格式。
+3. 不要依赖 parser 做安全控制；执行器侧仍应做工具名和参数校验。
 
-当前版本的可执行解析以 **XML/Markup 家族**为主：
-
-- `<tool_call>...</tool_call>`
-- `<function_call>...</function_call>`
-- `<invoke ...>...</invoke>`（含自闭合）
-- `<tool_use>...</tool_use>`
-- antml 变体（如 `antml:function_call` / `antml:argument`）
-
-并支持在这些标记块内部解析：
-
-- JSON 参数字符串
-- 标签参数（`<parameter name="...">...`）
-- key/value 风格子标签
-
-## 3) 不应再假设的行为
-
-以下说法在当前实现中已不成立：
-
-1. “纯 JSON `tool_calls` 片段会被直接当作可执行工具调用解析”。
-2. “存在 `toolcall.mode` / `toolcall.early_emit_confidence` 等可配置开关可以改变解析策略”。
-
-当前策略在代码中固定为：
-
-- 特征匹配开启（feature-match on）
-- 高置信度早发开启（early emit on）
-- policy 拒绝字段保留但未启用
-
-## 4) 流式与防泄漏语义
-
-在流式链路中（OpenAI / Claude / Gemini 统一内核）：
-
-- 工具调用片段会被优先提取为结构化增量输出；
-- 已识别的工具调用原始片段不会作为普通文本再次回流；
-- fenced code block 中的示例内容按文本处理，不作为可执行工具调用。
-
-## 5) 落地建议（按当前实现）
-
-1. Prompt 里优先约束模型输出 XML/Markup 工具块。
-2. 执行器侧继续做工具名白名单与参数 schema 校验（不要依赖 parser 代替安全策略）。
-3. 需要兼容历史“纯 JSON tool_calls”模型输出时，请在上游模板层把输出规范化为 XML/Markup 风格再进入 DS2API。
-
-## 6) 回归验证建议
+## 6) 回归验证
 
 可直接运行：
 
 ```bash
-go test -v -run 'TestParseToolCalls|TestRepair' ./internal/toolcall/
+go test -v -run 'TestParseToolCalls|TestProcessToolSieve' ./internal/toolcall ./internal/toolstream ./internal/httpapi/openai/...
 node --test tests/node/stream-tool-sieve.test.js
 ```
 
 重点覆盖：
 
-- `<tool_call>` / `<function_call>` / `<invoke>` / `tool_use` / antml 变体
-- 参数 JSON 修复与解析
-- 流式增量下的工具调用提取与文本防泄漏
+- canonical `<tool_calls>` wrapper 正常解析
+- 非 canonical 内容按普通文本透传
+- 代码块示例不执行

internal/adapter/claude/deps_injection_test.go

@@ -1,34 +0,0 @@
-package claude
-
-import "testing"
-
-type mockClaudeConfig struct {
-	m map[string]string
-}
-
-func (m mockClaudeConfig) ClaudeMapping() map[string]string { return m.m }
-func (mockClaudeConfig) CompatStripReferenceMarkers() bool  { return true }
-
-func TestNormalizeClaudeRequestUsesConfigInterfaceMapping(t *testing.T) {
-	req := map[string]any{
-		"model": "claude-opus-4-6",
-		"messages": []any{
-			map[string]any{"role": "user", "content": "hello"},
-		},
-	}
-	out, err := normalizeClaudeRequest(mockClaudeConfig{
-		m: map[string]string{
-			"fast": "deepseek-chat",
-			"slow": "deepseek-reasoner-search",
-		},
-	}, req)
-	if err != nil {
-		t.Fatalf("normalizeClaudeRequest error: %v", err)
-	}
-	if out.Standard.ResolvedModel != "deepseek-reasoner-search" {
-		t.Fatalf("resolved model mismatch: got=%q", out.Standard.ResolvedModel)
-	}
-	if !out.Standard.Thinking || !out.Standard.Search {
-		t.Fatalf("unexpected flags: thinking=%v search=%v", out.Standard.Thinking, out.Standard.Search)
-	}
-}

internal/adapter/openai/handler_routes.go

@@ -1,74 +0,0 @@
-package openai
-
-import (
-	"net/http"
-	"strings"
-	"sync"
-	"time"
-
-	"github.com/go-chi/chi/v5"
-
-	"ds2api/internal/auth"
-	"ds2api/internal/chathistory"
-	"ds2api/internal/config"
-	"ds2api/internal/util"
-)
-
-const (
-	// openAIUploadMaxSize limits total multipart request body size (100 MiB).
-	openAIUploadMaxSize = 100 << 20
-	// openAIGeneralMaxSize limits total JSON request body size (100 MiB).
-	openAIGeneralMaxSize = 100 << 20
-)
-
-// writeJSON is a package-internal alias kept to avoid mass-renaming across
-// every call-site in this package.
-var writeJSON = util.WriteJSON
-
-type Handler struct {
-	Store       ConfigReader
-	Auth        AuthResolver
-	DS          DeepSeekCaller
-	ChatHistory *chathistory.Store
-
-	leaseMu      sync.Mutex
-	streamLeases map[string]streamLease
-	responsesMu  sync.Mutex
-	responses    *responseStore
-}
-
-func (h *Handler) compatStripReferenceMarkers() bool {
-	if h == nil || h.Store == nil {
-		return true
-	}
-	return h.Store.CompatStripReferenceMarkers()
-}
-
-type streamLease struct {
-	Auth      *auth.RequestAuth
-	ExpiresAt time.Time
-}
-
-func RegisterRoutes(r chi.Router, h *Handler) {
-	r.Get("/v1/models", h.ListModels)
-	r.Get("/v1/models/{model_id}", h.GetModel)
-	r.Post("/v1/chat/completions", h.ChatCompletions)
-	r.Post("/v1/responses", h.Responses)
-	r.Get("/v1/responses/{response_id}", h.GetResponseByID)
-	r.Post("/v1/files", h.UploadFile)
-	r.Post("/v1/embeddings", h.Embeddings)
-}
-
-func (h *Handler) ListModels(w http.ResponseWriter, _ *http.Request) {
-	writeJSON(w, http.StatusOK, config.OpenAIModelsResponse())
-}
-
-func (h *Handler) GetModel(w http.ResponseWriter, r *http.Request) {
-	modelID := strings.TrimSpace(chi.URLParam(r, "model_id"))
-	model, ok := config.OpenAIModelByID(h.Store, modelID)
-	if !ok {
-		writeOpenAIError(w, http.StatusNotFound, "Model not found.")
-		return
-	}
-	writeJSON(w, http.StatusOK, model)
-}

internal/adapter/openai/handler_toolcall_format.go

@@ -1,170 +0,0 @@
-package openai
-
-import (
-	"ds2api/internal/toolcall"
-	"encoding/json"
-	"fmt"
-	"strings"
-
-	"github.com/google/uuid"
-
-	"ds2api/internal/util"
-)
-
-func injectToolPrompt(messages []map[string]any, tools []any, policy util.ToolChoicePolicy) ([]map[string]any, []string) {
-	if policy.IsNone() {
-		return messages, nil
-	}
-	toolSchemas := make([]string, 0, len(tools))
-	names := make([]string, 0, len(tools))
-	isAllowed := func(name string) bool {
-		if strings.TrimSpace(name) == "" {
-			return false
-		}
-		if len(policy.Allowed) == 0 {
-			return true
-		}
-		_, ok := policy.Allowed[name]
-		return ok
-	}
-
-	for _, t := range tools {
-		tool, ok := t.(map[string]any)
-		if !ok {
-			continue
-		}
-		fn, _ := tool["function"].(map[string]any)
-		if len(fn) == 0 {
-			fn = tool
-		}
-		name, _ := fn["name"].(string)
-		desc, _ := fn["description"].(string)
-		schema, _ := fn["parameters"].(map[string]any)
-		name = strings.TrimSpace(name)
-		if !isAllowed(name) {
-			continue
-		}
-		names = append(names, name)
-		if desc == "" {
-			desc = "No description available"
-		}
-		b, _ := json.Marshal(schema)
-		toolSchemas = append(toolSchemas, fmt.Sprintf("Tool: %s\nDescription: %s\nParameters: %s", name, desc, string(b)))
-	}
-	if len(toolSchemas) == 0 {
-		return messages, names
-	}
-	toolPrompt := "You have access to these tools:\n\n" + strings.Join(toolSchemas, "\n\n") + "\n\n" + buildToolCallInstructions(names)
-	if policy.Mode == util.ToolChoiceRequired {
-		toolPrompt += "\n7) For this response, you MUST call at least one tool from the allowed list."
-	}
-	if policy.Mode == util.ToolChoiceForced && strings.TrimSpace(policy.ForcedName) != "" {
-		toolPrompt += "\n7) For this response, you MUST call exactly this tool name: " + strings.TrimSpace(policy.ForcedName)
-		toolPrompt += "\n8) Do not call any other tool."
-	}
-
-	for i := range messages {
-		if messages[i]["role"] == "system" {
-			old, _ := messages[i]["content"].(string)
-			messages[i]["content"] = strings.TrimSpace(old + "\n\n" + toolPrompt)
-			return messages, names
-		}
-	}
-	messages = append([]map[string]any{{"role": "system", "content": toolPrompt}}, messages...)
-	return messages, names
-}
-
-// buildToolCallInstructions delegates to the shared util implementation.
-func buildToolCallInstructions(toolNames []string) string {
-	return toolcall.BuildToolCallInstructions(toolNames)
-}
-
-func formatIncrementalStreamToolCallDeltas(deltas []toolCallDelta, ids map[int]string) []map[string]any {
-	if len(deltas) == 0 {
-		return nil
-	}
-	out := make([]map[string]any, 0, len(deltas))
-	for _, d := range deltas {
-		if d.Name == "" && d.Arguments == "" {
-			continue
-		}
-		callID, ok := ids[d.Index]
-		if !ok || callID == "" {
-			callID = "call_" + strings.ReplaceAll(uuid.NewString(), "-", "")
-			ids[d.Index] = callID
-		}
-		item := map[string]any{
-			"index": d.Index,
-			"id":    callID,
-			"type":  "function",
-		}
-		fn := map[string]any{}
-		if d.Name != "" {
-			fn["name"] = d.Name
-		}
-		if d.Arguments != "" {
-			fn["arguments"] = d.Arguments
-		}
-		if len(fn) > 0 {
-			item["function"] = fn
-		}
-		out = append(out, item)
-	}
-	return out
-}
-
-func filterIncrementalToolCallDeltasByAllowed(deltas []toolCallDelta, seenNames map[int]string) []toolCallDelta {
-	if len(deltas) == 0 {
-		return nil
-	}
-	out := make([]toolCallDelta, 0, len(deltas))
-	for _, d := range deltas {
-		if d.Name != "" {
-			if seenNames != nil {
-				seenNames[d.Index] = d.Name
-			}
-			out = append(out, d)
-			continue
-		}
-		if seenNames == nil {
-			out = append(out, d)
-			continue
-		}
-		name := strings.TrimSpace(seenNames[d.Index])
-		if name == "" {
-			continue
-		}
-		out = append(out, d)
-	}
-	return out
-}
-
-func formatFinalStreamToolCallsWithStableIDs(calls []toolcall.ParsedToolCall, ids map[int]string) []map[string]any {
-	if len(calls) == 0 {
-		return nil
-	}
-	out := make([]map[string]any, 0, len(calls))
-	for i, c := range calls {
-		callID := ""
-		if ids != nil {
-			callID = strings.TrimSpace(ids[i])
-		}
-		if callID == "" {
-			callID = "call_" + strings.ReplaceAll(uuid.NewString(), "-", "")
-			if ids != nil {
-				ids[i] = callID
-			}
-		}
-		args, _ := json.Marshal(c.Input)
-		out = append(out, map[string]any{
-			"index": i,
-			"id":    callID,
-			"type":  "function",
-			"function": map[string]any{
-				"name":      c.Name,
-				"arguments": string(args),
-			},
-		})
-	}
-	return out
-}

internal/adapter/openai/handler_toolcall_policy.go

@@ -1,9 +0,0 @@
-package openai
-
-func (h *Handler) toolcallFeatureMatchEnabled() bool {
-	return true
-}
-
-func (h *Handler) toolcallEarlyEmitHighConfidence() bool {
-	return true
-}

internal/adapter/openai/history_split.go

@@ -1,289 +0,0 @@
-package openai
-
-import (
-	"context"
-	"errors"
-	"fmt"
-	"strings"
-
-	"ds2api/internal/auth"
-	"ds2api/internal/deepseek"
-	"ds2api/internal/util"
-)
-
-const (
-	historySplitFilename    = "HISTORY.txt"
-	historySplitContentType = "text/plain; charset=utf-8"
-	historySplitPurpose     = "assistants"
-)
-
-func (h *Handler) applyHistorySplit(ctx context.Context, a *auth.RequestAuth, stdReq util.StandardRequest) (util.StandardRequest, error) {
-	if h == nil || h.DS == nil || h.Store == nil || a == nil {
-		return stdReq, nil
-	}
-	if !h.Store.HistorySplitEnabled() {
-		return stdReq, nil
-	}
-
-	promptMessages, historyMessages := splitOpenAIHistoryMessages(stdReq.Messages, h.Store.HistorySplitTriggerAfterTurns())
-	if len(historyMessages) == 0 {
-		return stdReq, nil
-	}
-
-	reasoningContent := extractHistorySplitReasoningContent(historyMessages)
-	historyText := buildOpenAIHistoryTranscript(historyMessages)
-	if strings.TrimSpace(historyText) == "" {
-		return stdReq, errors.New("history split produced empty transcript")
-	}
-
-	result, err := h.DS.UploadFile(ctx, a, deepseek.UploadFileRequest{
-		Filename:    historySplitFilename,
-		ContentType: historySplitContentType,
-		Purpose:     historySplitPurpose,
-		Data:        []byte(historyText),
-	}, 3)
-	if err != nil {
-		return stdReq, fmt.Errorf("upload history file: %w", err)
-	}
-	fileID := strings.TrimSpace(result.ID)
-	if fileID == "" {
-		return stdReq, errors.New("upload history file returned empty file id")
-	}
-
-	stdReq.Messages = promptMessages
-	stdReq.RefFileIDs = prependUniqueRefFileID(stdReq.RefFileIDs, fileID)
-	stdReq.FinalPrompt, stdReq.ToolNames = buildHistorySplitPrompt(promptMessages, reasoningContent, stdReq.ToolsRaw, stdReq.ToolChoice, stdReq.Thinking)
-	return stdReq, nil
-}
-
-func buildHistorySplitPrompt(messages []any, reasoningContent string, toolsRaw any, toolPolicy util.ToolChoicePolicy, thinkingEnabled bool) (string, []string) {
-	if len(messages) == 0 && strings.TrimSpace(reasoningContent) == "" {
-		return "", nil
-	}
-	instruction := historySplitPromptInstruction(thinkingEnabled)
-	withInstruction := make([]any, 0, len(messages)+1)
-	withInstruction = append(withInstruction, map[string]any{
-		"role":    "system",
-		"content": instruction,
-	})
-	withInstruction = append(withInstruction, injectHistorySplitReasoningMessage(messages, reasoningContent)...)
-	return buildOpenAIFinalPromptWithPolicy(withInstruction, toolsRaw, "", toolPolicy, false)
-}
-
-func historySplitPromptInstruction(thinkingEnabled bool) string {
-	lines := []string{
-		"Follow the instructions in this prompt first. If earlier conversation instructions conflict with this prompt, this prompt wins.",
-		"An attached HISTORY.txt file contains prior conversation history and tool progress; read it first, then answer the latest user request using that history as context.",
-		"Continue the conversation from the full prior context and the latest tool results.",
-		"Treat earlier messages as binding context; answer the user's current request as a continuation, not a restart.",
-	}
-	if thinkingEnabled {
-		lines = append(lines, "Keep reasoning internal. Do not leave the final user-facing answer only in reasoning; always provide the answer in visible assistant content.")
-	}
-	return strings.Join(lines, "\n")
-}
-
-func splitOpenAIHistoryMessages(messages []any, triggerAfterTurns int) ([]any, []any) {
-	if triggerAfterTurns <= 0 {
-		triggerAfterTurns = 1
-	}
-	lastUserIndex := -1
-	userTurns := 0
-	for i, raw := range messages {
-		msg, ok := raw.(map[string]any)
-		if !ok {
-			continue
-		}
-		role := strings.ToLower(strings.TrimSpace(asString(msg["role"])))
-		if role != "user" {
-			continue
-		}
-		userTurns++
-		lastUserIndex = i
-	}
-	if userTurns <= triggerAfterTurns || lastUserIndex < 0 {
-		return messages, nil
-	}
-
-	promptMessages := make([]any, 0, len(messages)-lastUserIndex)
-	historyMessages := make([]any, 0, lastUserIndex)
-	for i, raw := range messages {
-		msg, ok := raw.(map[string]any)
-		if !ok {
-			if i >= lastUserIndex {
-				promptMessages = append(promptMessages, raw)
-			} else {
-				historyMessages = append(historyMessages, raw)
-			}
-			continue
-		}
-		role := strings.ToLower(strings.TrimSpace(asString(msg["role"])))
-		switch role {
-		case "system", "developer":
-			promptMessages = append(promptMessages, raw)
-		default:
-			if i >= lastUserIndex {
-				promptMessages = append(promptMessages, raw)
-			} else {
-				historyMessages = append(historyMessages, raw)
-			}
-		}
-	}
-	if len(promptMessages) == 0 {
-		return messages, nil
-	}
-	return promptMessages, historyMessages
-}
-
-func buildOpenAIHistoryTranscript(messages []any) string {
-	var b strings.Builder
-	b.WriteString("# HISTORY.txt\n")
-	b.WriteString("Prior conversation history and tool progress.\n\n")
-
-	entry := 0
-	for _, raw := range messages {
-		msg, ok := raw.(map[string]any)
-		if !ok {
-			continue
-		}
-		role := strings.ToLower(strings.TrimSpace(asString(msg["role"])))
-		content := buildOpenAIHistoryEntry(role, msg)
-		if strings.TrimSpace(content) == "" {
-			continue
-		}
-		entry++
-		fmt.Fprintf(&b, "=== %d. %s ===\n%s\n\n", entry, strings.ToUpper(roleLabelForHistory(role)), content)
-	}
-	return strings.TrimSpace(b.String()) + "\n"
-}
-
-func buildOpenAIHistoryEntry(role string, msg map[string]any) string {
-	switch role {
-	case "assistant":
-		return strings.TrimSpace(buildAssistantHistoryContent(msg))
-	case "tool", "function":
-		return strings.TrimSpace(buildToolHistoryContent(msg))
-	case "user":
-		return strings.TrimSpace(normalizeOpenAIContentForPrompt(msg["content"]))
-	default:
-		return strings.TrimSpace(normalizeOpenAIContentForPrompt(msg["content"]))
-	}
-}
-
-func buildAssistantHistoryContent(msg map[string]any) string {
-	return strings.TrimSpace(buildAssistantContentForPrompt(msg))
-}
-
-func buildToolHistoryContent(msg map[string]any) string {
-	content := strings.TrimSpace(normalizeOpenAIContentForPrompt(msg["content"]))
-	parts := make([]string, 0, 2)
-	if name := strings.TrimSpace(asString(msg["name"])); name != "" {
-		parts = append(parts, "name="+name)
-	}
-	if callID := strings.TrimSpace(asString(msg["tool_call_id"])); callID != "" {
-		parts = append(parts, "tool_call_id="+callID)
-	}
-	header := ""
-	if len(parts) > 0 {
-		header = "[" + strings.Join(parts, " ") + "]"
-	}
-	switch {
-	case header != "" && content != "":
-		return header + "\n" + content
-	case header != "":
-		return header
-	default:
-		return content
-	}
-}
-
-func extractHistorySplitReasoningContent(messages []any) string {
-	for i := len(messages) - 1; i >= 0; i-- {
-		msg, ok := messages[i].(map[string]any)
-		if !ok {
-			continue
-		}
-		role := strings.ToLower(strings.TrimSpace(asString(msg["role"])))
-		if role != "assistant" {
-			continue
-		}
-		reasoning := strings.TrimSpace(normalizeOpenAIReasoningContentForPrompt(msg["reasoning_content"]))
-		if reasoning == "" {
-			reasoning = strings.TrimSpace(extractOpenAIReasoningContentFromMessage(msg["content"]))
-		}
-		if reasoning != "" {
-			return reasoning
-		}
-	}
-	return ""
-}
-
-func injectHistorySplitReasoningMessage(messages []any, reasoningContent string) []any {
-	reasoningContent = strings.TrimSpace(reasoningContent)
-	if reasoningContent == "" {
-		return messages
-	}
-	reasoningMsg := map[string]any{
-		"role":              "assistant",
-		"content":           "",
-		"reasoning_content": reasoningContent,
-	}
-	lastUserIndex := lastOpenAIUserMessageIndex(messages)
-	if lastUserIndex < 0 {
-		out := make([]any, 0, len(messages)+1)
-		out = append(out, reasoningMsg)
-		out = append(out, messages...)
-		return out
-	}
-	out := make([]any, 0, len(messages)+1)
-	for i, raw := range messages {
-		if i == lastUserIndex {
-			out = append(out, reasoningMsg)
-		}
-		out = append(out, raw)
-	}
-	return out
-}
-
-func lastOpenAIUserMessageIndex(messages []any) int {
-	last := -1
-	for i, raw := range messages {
-		msg, ok := raw.(map[string]any)
-		if !ok {
-			continue
-		}
-		if strings.ToLower(strings.TrimSpace(asString(msg["role"]))) == "user" {
-			last = i
-		}
-	}
-	return last
-}
-
-func roleLabelForHistory(role string) string {
-	role = strings.ToLower(strings.TrimSpace(role))
-	switch role {
-	case "function":
-		return "tool"
-	case "":
-		return "unknown"
-	default:
-		return role
-	}
-}
-
-func prependUniqueRefFileID(existing []string, fileID string) []string {
-	fileID = strings.TrimSpace(fileID)
-	if fileID == "" {
-		return existing
-	}
-	out := make([]string, 0, len(existing)+1)
-	out = append(out, fileID)
-	for _, id := range existing {
-		trimmed := strings.TrimSpace(id)
-		if trimmed == "" || strings.EqualFold(trimmed, fileID) {
-			continue
-		}
-		out = append(out, trimmed)
-	}
-	return out
-}

internal/adapter/openai/history_split_test.go

@@ -1,353 +0,0 @@
-package openai
-
-import (
-	"context"
-	"encoding/json"
-	"net/http"
-	"net/http/httptest"
-	"strings"
-	"testing"
-
-	"github.com/go-chi/chi/v5"
-
-	"ds2api/internal/auth"
-	"ds2api/internal/util"
-)
-
-func historySplitTestMessages() []any {
-	toolCalls := []any{
-		map[string]any{
-			"name":      "search",
-			"arguments": map[string]any{"query": "docs"},
-		},
-	}
-	return []any{
-		map[string]any{"role": "system", "content": "system instructions"},
-		map[string]any{"role": "user", "content": "first user turn"},
-		map[string]any{
-			"role":              "assistant",
-			"content":           "",
-			"reasoning_content": "hidden reasoning",
-			"tool_calls":        toolCalls,
-		},
-		map[string]any{
-			"role":         "tool",
-			"name":         "search",
-			"tool_call_id": "call-1",
-			"content":      "tool result",
-		},
-		map[string]any{"role": "user", "content": "latest user turn"},
-	}
-}
-
-func TestBuildOpenAIHistoryTranscriptPreservesOrderAndToolHistory(t *testing.T) {
-	promptMessages, historyMessages := splitOpenAIHistoryMessages(historySplitTestMessages(), 1)
-	if len(promptMessages) != 2 {
-		t.Fatalf("expected 2 prompt messages, got %d", len(promptMessages))
-	}
-	if len(historyMessages) != 3 {
-		t.Fatalf("expected 3 history messages, got %d", len(historyMessages))
-	}
-
-	transcript := buildOpenAIHistoryTranscript(historyMessages)
-	if !strings.Contains(transcript, "first user turn") {
-		t.Fatalf("expected user history in transcript, got %s", transcript)
-	}
-	if !strings.Contains(transcript, "<tool_calls>") {
-		t.Fatalf("expected assistant tool_calls in transcript, got %s", transcript)
-	}
-	if !strings.Contains(transcript, "tool_call_id=call-1") {
-		t.Fatalf("expected tool call id in transcript, got %s", transcript)
-	}
-	if !strings.Contains(transcript, "[reasoning_content]") {
-		t.Fatalf("expected reasoning block in HISTORY.txt, got %s", transcript)
-	}
-	if !strings.Contains(transcript, "hidden reasoning") {
-		t.Fatalf("expected reasoning text in HISTORY.txt, got %s", transcript)
-	}
-
-	userIdx := strings.Index(transcript, "=== 1. USER ===")
-	assistantIdx := strings.Index(transcript, "=== 2. ASSISTANT ===")
-	toolIdx := strings.Index(transcript, "=== 3. TOOL ===")
-	if userIdx < 0 || assistantIdx < 0 || toolIdx < 0 {
-		t.Fatalf("expected ordered role sections, got %s", transcript)
-	}
-	if userIdx >= assistantIdx || assistantIdx >= toolIdx {
-		t.Fatalf("expected USER -> ASSISTANT -> TOOL order, got %s", transcript)
-	}
-	if reasoningIdx := strings.Index(transcript, "[reasoning_content]"); reasoningIdx < 0 || reasoningIdx > strings.Index(transcript, "<tool_calls>") {
-		t.Fatalf("expected reasoning block before tool calls, got %s", transcript)
-	}
-	reasoning := extractHistorySplitReasoningContent(historyMessages)
-	if reasoning != "hidden reasoning" {
-		t.Fatalf("expected latest assistant reasoning to be extracted, got %q", reasoning)
-	}
-
-	finalPrompt, _ := buildHistorySplitPrompt(promptMessages, reasoning, nil, util.DefaultToolChoicePolicy(), false)
-	if !strings.Contains(finalPrompt, "latest user turn") {
-		t.Fatalf("expected latest user turn in final prompt, got %s", finalPrompt)
-	}
-	if strings.Contains(finalPrompt, "first user turn") {
-		t.Fatalf("expected earlier history to be removed from final prompt, got %s", finalPrompt)
-	}
-	if !strings.Contains(finalPrompt, "[reasoning_content]") || !strings.Contains(finalPrompt, "hidden reasoning") {
-		t.Fatalf("expected latest assistant reasoning to be attached to prompt, got %s", finalPrompt)
-	}
-	if !strings.Contains(finalPrompt, "HISTORY.txt") {
-		t.Fatalf("expected history instruction in final prompt, got %s", finalPrompt)
-	}
-	if !strings.Contains(finalPrompt, "Follow the instructions in this prompt first") {
-		t.Fatalf("expected stronger prompt override in final prompt, got %s", finalPrompt)
-	}
-	if strings.Index(finalPrompt, "Follow the instructions in this prompt first") > strings.Index(finalPrompt, "Continue the conversation") {
-		t.Fatalf("expected history split instruction before continuity instructions, got %s", finalPrompt)
-	}
-}
-
-func TestSplitOpenAIHistoryMessagesUsesLatestUserTurn(t *testing.T) {
-	toolCalls := []any{
-		map[string]any{
-			"name":      "search",
-			"arguments": map[string]any{"query": "docs"},
-		},
-	}
-	messages := []any{
-		map[string]any{"role": "system", "content": "system instructions"},
-		map[string]any{"role": "user", "content": "first user turn"},
-		map[string]any{
-			"role":       "assistant",
-			"content":    "",
-			"tool_calls": toolCalls,
-		},
-		map[string]any{
-			"role":         "tool",
-			"name":         "search",
-			"tool_call_id": "call-1",
-			"content":      "tool result",
-		},
-		map[string]any{"role": "user", "content": "middle user turn"},
-		map[string]any{
-			"role":    "assistant",
-			"content": "middle assistant turn",
-		},
-		map[string]any{"role": "user", "content": "latest user turn"},
-	}
-
-	promptMessages, historyMessages := splitOpenAIHistoryMessages(messages, 1)
-	if len(promptMessages) == 0 || len(historyMessages) == 0 {
-		t.Fatalf("expected both prompt and history messages, got prompt=%d history=%d", len(promptMessages), len(historyMessages))
-	}
-	reasoning := extractHistorySplitReasoningContent(historyMessages)
-	if reasoning != "" {
-		t.Fatalf("expected no reasoning in this fixture, got %q", reasoning)
-	}
-
-	promptText, _ := buildHistorySplitPrompt(promptMessages, reasoning, nil, util.DefaultToolChoicePolicy(), false)
-	if !strings.Contains(promptText, "latest user turn") {
-		t.Fatalf("expected latest user turn in prompt, got %s", promptText)
-	}
-	if strings.Contains(promptText, "middle user turn") {
-		t.Fatalf("expected middle user turn to be split into history, got %s", promptText)
-	}
-
-	historyText := buildOpenAIHistoryTranscript(historyMessages)
-	if !strings.Contains(historyText, "middle user turn") {
-		t.Fatalf("expected middle user turn in HISTORY.txt, got %s", historyText)
-	}
-	if strings.Contains(historyText, "latest user turn") {
-		t.Fatalf("expected latest user turn to remain in prompt, got %s", historyText)
-	}
-}
-
-func TestApplyHistorySplitSkipsFirstTurn(t *testing.T) {
-	ds := &inlineUploadDSStub{}
-	h := &Handler{
-		Store: mockOpenAIConfig{
-			wideInput:           true,
-			historySplitEnabled: true,
-			historySplitTurns:   1,
-		},
-		DS: ds,
-	}
-	req := map[string]any{
-		"model": "deepseek-chat",
-		"messages": []any{
-			map[string]any{"role": "user", "content": "hello"},
-		},
-	}
-	stdReq, err := normalizeOpenAIChatRequest(h.Store, req, "")
-	if err != nil {
-		t.Fatalf("normalize failed: %v", err)
-	}
-
-	out, err := h.applyHistorySplit(context.Background(), &auth.RequestAuth{DeepSeekToken: "token"}, stdReq)
-	if err != nil {
-		t.Fatalf("apply history split failed: %v", err)
-	}
-	if len(ds.uploadCalls) != 0 {
-		t.Fatalf("expected no upload on first turn, got %d", len(ds.uploadCalls))
-	}
-	if out.FinalPrompt != stdReq.FinalPrompt {
-		t.Fatalf("expected prompt unchanged on first turn")
-	}
-	if len(out.RefFileIDs) != len(stdReq.RefFileIDs) {
-		t.Fatalf("expected ref files unchanged on first turn")
-	}
-}
-
-func TestChatCompletionsHistorySplitUploadsHistoryAndKeepsLatestPrompt(t *testing.T) {
-	ds := &inlineUploadDSStub{}
-	h := &Handler{
-		Store: mockOpenAIConfig{
-			wideInput:           true,
-			historySplitEnabled: true,
-			historySplitTurns:   1,
-		},
-		Auth: streamStatusAuthStub{},
-		DS:   ds,
-	}
-	reqBody, _ := json.Marshal(map[string]any{
-		"model":    "deepseek-chat",
-		"messages": historySplitTestMessages(),
-		"stream":   false,
-	})
-	req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", strings.NewReader(string(reqBody)))
-	req.Header.Set("Authorization", "Bearer direct-token")
-	req.Header.Set("Content-Type", "application/json")
-	rec := httptest.NewRecorder()
-
-	h.ChatCompletions(rec, req)
-
-	if rec.Code != http.StatusOK {
-		t.Fatalf("expected 200, got %d body=%s", rec.Code, rec.Body.String())
-	}
-	if len(ds.uploadCalls) != 1 {
-		t.Fatalf("expected 1 upload call, got %d", len(ds.uploadCalls))
-	}
-	upload := ds.uploadCalls[0]
-	if upload.Filename != "HISTORY.txt" {
-		t.Fatalf("unexpected upload filename: %q", upload.Filename)
-	}
-	if upload.ContentType != "text/plain; charset=utf-8" {
-		t.Fatalf("unexpected content type: %q", upload.ContentType)
-	}
-	if upload.Purpose != "assistants" {
-		t.Fatalf("unexpected purpose: %q", upload.Purpose)
-	}
-	historyText := string(upload.Data)
-	if !strings.Contains(historyText, "first user turn") || !strings.Contains(historyText, "tool result") {
-		t.Fatalf("expected older turns in HISTORY.txt, got %s", historyText)
-	}
-	if strings.Contains(historyText, "latest user turn") {
-		t.Fatalf("expected latest turn to remain in prompt, got %s", historyText)
-	}
-	if ds.completionReq == nil {
-		t.Fatal("expected completion payload to be captured")
-	}
-	promptText, _ := ds.completionReq["prompt"].(string)
-	if !strings.Contains(promptText, "latest user turn") {
-		t.Fatalf("expected latest turn in completion prompt, got %s", promptText)
-	}
-	if strings.Contains(promptText, "first user turn") {
-		t.Fatalf("expected historical turns removed from completion prompt, got %s", promptText)
-	}
-	if !strings.Contains(promptText, "[reasoning_content]") || !strings.Contains(promptText, "hidden reasoning") {
-		t.Fatalf("expected latest assistant reasoning to be attached to completion prompt, got %s", promptText)
-	}
-	if !strings.Contains(promptText, "HISTORY.txt") {
-		t.Fatalf("expected history instruction in completion prompt, got %s", promptText)
-	}
-	if !strings.Contains(promptText, "Follow the instructions in this prompt first") {
-		t.Fatalf("expected stronger prompt override in completion prompt, got %s", promptText)
-	}
-	if strings.Index(promptText, "Follow the instructions in this prompt first") > strings.Index(promptText, "Continue the conversation") {
-		t.Fatalf("expected history split instruction before continuity instructions, got %s", promptText)
-	}
-	refIDs, _ := ds.completionReq["ref_file_ids"].([]any)
-	if len(refIDs) == 0 || refIDs[0] != "file-inline-1" {
-		t.Fatalf("expected uploaded history file to be first ref_file_id, got %#v", ds.completionReq["ref_file_ids"])
-	}
-}
-
-func TestResponsesHistorySplitUploadsHistoryAndKeepsLatestPrompt(t *testing.T) {
-	ds := &inlineUploadDSStub{}
-	h := &Handler{
-		Store: mockOpenAIConfig{
-			wideInput:           true,
-			historySplitEnabled: true,
-			historySplitTurns:   1,
-		},
-		Auth: streamStatusAuthStub{},
-		DS:   ds,
-	}
-	r := chi.NewRouter()
-	RegisterRoutes(r, h)
-	reqBody, _ := json.Marshal(map[string]any{
-		"model":    "deepseek-chat",
-		"messages": historySplitTestMessages(),
-		"stream":   false,
-	})
-	req := httptest.NewRequest(http.MethodPost, "/v1/responses", strings.NewReader(string(reqBody)))
-	req.Header.Set("Authorization", "Bearer direct-token")
-	req.Header.Set("Content-Type", "application/json")
-	rec := httptest.NewRecorder()
-
-	r.ServeHTTP(rec, req)
-
-	if rec.Code != http.StatusOK {
-		t.Fatalf("expected 200, got %d body=%s", rec.Code, rec.Body.String())
-	}
-	if len(ds.uploadCalls) != 1 {
-		t.Fatalf("expected 1 upload call, got %d", len(ds.uploadCalls))
-	}
-	if ds.completionReq == nil {
-		t.Fatal("expected completion payload to be captured")
-	}
-	promptText, _ := ds.completionReq["prompt"].(string)
-	if !strings.Contains(promptText, "latest user turn") {
-		t.Fatalf("expected latest turn in completion prompt, got %s", promptText)
-	}
-	if strings.Contains(promptText, "first user turn") {
-		t.Fatalf("expected historical turns removed from completion prompt, got %s", promptText)
-	}
-	if !strings.Contains(promptText, "[reasoning_content]") || !strings.Contains(promptText, "hidden reasoning") {
-		t.Fatalf("expected latest assistant reasoning to be attached to completion prompt, got %s", promptText)
-	}
-	if !strings.Contains(promptText, "Follow the instructions in this prompt first") {
-		t.Fatalf("expected stronger prompt override in completion prompt, got %s", promptText)
-	}
-	if strings.Index(promptText, "Follow the instructions in this prompt first") > strings.Index(promptText, "Continue the conversation") {
-		t.Fatalf("expected history split instruction before continuity instructions, got %s", promptText)
-	}
-}
-
-func TestChatCompletionsHistorySplitUploadFailureReturnsInternalServerError(t *testing.T) {
-	ds := &inlineUploadDSStub{uploadErr: context.DeadlineExceeded}
-	h := &Handler{
-		Store: mockOpenAIConfig{
-			wideInput:           true,
-			historySplitEnabled: true,
-			historySplitTurns:   1,
-		},
-		Auth: streamStatusAuthStub{},
-		DS:   ds,
-	}
-	reqBody, _ := json.Marshal(map[string]any{
-		"model":    "deepseek-chat",
-		"messages": historySplitTestMessages(),
-		"stream":   false,
-	})
-	req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", strings.NewReader(string(reqBody)))
-	req.Header.Set("Authorization", "Bearer direct-token")
-	req.Header.Set("Content-Type", "application/json")
-	rec := httptest.NewRecorder()
-
-	h.ChatCompletions(rec, req)
-
-	if rec.Code != http.StatusInternalServerError {
-		t.Fatalf("expected 500, got %d body=%s", rec.Code, rec.Body.String())
-	}
-	if ds.completionReq != nil {
-		t.Fatalf("did not expect completion payload on upload failure")
-	}
-}

internal/adapter/openai/prompt_build.go

@@ -1,26 +0,0 @@
-package openai
-
-import (
-	"ds2api/internal/deepseek"
-	"ds2api/internal/util"
-)
-
-func buildOpenAIFinalPrompt(messagesRaw []any, toolsRaw any, traceID string, thinkingEnabled bool) (string, []string) {
-	return buildOpenAIFinalPromptWithPolicy(messagesRaw, toolsRaw, traceID, util.DefaultToolChoicePolicy(), thinkingEnabled)
-}
-
-func buildOpenAIFinalPromptWithPolicy(messagesRaw []any, toolsRaw any, traceID string, toolPolicy util.ToolChoicePolicy, thinkingEnabled bool) (string, []string) {
-	messages := normalizeOpenAIMessagesForPrompt(messagesRaw, traceID)
-	toolNames := []string{}
-	if tools, ok := toolsRaw.([]any); ok && len(tools) > 0 {
-		messages, toolNames = injectToolPrompt(messages, tools, toolPolicy)
-	}
-	return deepseek.MessagesPrepareWithThinking(messages, thinkingEnabled), toolNames
-}
-
-// BuildPromptForAdapter exposes the OpenAI-compatible prompt building flow so
-// other protocol adapters (for example Gemini) can reuse the same tool/history
-// normalization logic and remain behavior-compatible with chat/completions.
-func BuildPromptForAdapter(messagesRaw []any, toolsRaw any, traceID string, thinkingEnabled bool) (string, []string) {
-	return buildOpenAIFinalPrompt(messagesRaw, toolsRaw, traceID, thinkingEnabled)
-}

internal/adapter/openai/standard_request_test.go

@@ -1,210 +0,0 @@
-package openai
-
-import (
-	"testing"
-
-	"ds2api/internal/config"
-	"ds2api/internal/util"
-)
-
-func newEmptyStoreForNormalizeTest(t *testing.T) *config.Store {
-	t.Helper()
-	t.Setenv("DS2API_CONFIG_JSON", `{}`)
-	return config.LoadStore()
-}
-
-func TestNormalizeOpenAIChatRequest(t *testing.T) {
-	store := newEmptyStoreForNormalizeTest(t)
-	req := map[string]any{
-		"model": "gpt-5-codex",
-		"messages": []any{
-			map[string]any{"role": "user", "content": "hello"},
-		},
-		"temperature": 0.3,
-		"stream":      true,
-	}
-	n, err := normalizeOpenAIChatRequest(store, req, "")
-	if err != nil {
-		t.Fatalf("normalize failed: %v", err)
-	}
-	if n.ResolvedModel != "deepseek-reasoner" {
-		t.Fatalf("unexpected resolved model: %s", n.ResolvedModel)
-	}
-	if !n.Stream {
-		t.Fatalf("expected stream=true")
-	}
-	if _, ok := n.PassThrough["temperature"]; !ok {
-		t.Fatalf("expected temperature passthrough")
-	}
-	if n.FinalPrompt == "" {
-		t.Fatalf("expected non-empty final prompt")
-	}
-}
-
-func TestNormalizeOpenAIChatRequestCollectsRefFileIDs(t *testing.T) {
-	store := newEmptyStoreForNormalizeTest(t)
-	req := map[string]any{
-		"model": "gpt-5-codex",
-		"messages": []any{
-			map[string]any{
-				"role": "user",
-				"content": []any{
-					map[string]any{"type": "input_text", "text": "hello"},
-					map[string]any{"type": "input_file", "file_id": "file-msg"},
-				},
-			},
-		},
-		"attachments": []any{
-			map[string]any{"file_id": "file-attachment"},
-		},
-		"ref_file_ids": []any{"file-top", "file-attachment"},
-	}
-	n, err := normalizeOpenAIChatRequest(store, req, "")
-	if err != nil {
-		t.Fatalf("normalize failed: %v", err)
-	}
-	if len(n.RefFileIDs) != 3 {
-		t.Fatalf("expected 3 distinct file ids, got %#v", n.RefFileIDs)
-	}
-	if n.RefFileIDs[0] != "file-top" || n.RefFileIDs[1] != "file-attachment" || n.RefFileIDs[2] != "file-msg" {
-		t.Fatalf("unexpected file ids: %#v", n.RefFileIDs)
-	}
-}
-
-func TestNormalizeOpenAIResponsesRequestInput(t *testing.T) {
-	store := newEmptyStoreForNormalizeTest(t)
-	req := map[string]any{
-		"model":        "gpt-4o",
-		"input":        "ping",
-		"instructions": "system",
-	}
-	n, err := normalizeOpenAIResponsesRequest(store, req, "")
-	if err != nil {
-		t.Fatalf("normalize failed: %v", err)
-	}
-	if n.ResolvedModel != "deepseek-chat" {
-		t.Fatalf("unexpected resolved model: %s", n.ResolvedModel)
-	}
-	if len(n.Messages) != 2 {
-		t.Fatalf("expected 2 normalized messages, got %d", len(n.Messages))
-	}
-}
-
-func TestNormalizeOpenAIResponsesRequestToolChoiceRequired(t *testing.T) {
-	store := newEmptyStoreForNormalizeTest(t)
-	req := map[string]any{
-		"model": "gpt-4o",
-		"input": "ping",
-		"tools": []any{
-			map[string]any{
-				"type": "function",
-				"function": map[string]any{
-					"name": "search",
-					"parameters": map[string]any{
-						"type": "object",
-					},
-				},
-			},
-		},
-		"tool_choice": "required",
-	}
-	n, err := normalizeOpenAIResponsesRequest(store, req, "")
-	if err != nil {
-		t.Fatalf("normalize failed: %v", err)
-	}
-	if n.ToolChoice.Mode != util.ToolChoiceRequired {
-		t.Fatalf("expected tool choice mode required, got %q", n.ToolChoice.Mode)
-	}
-	if len(n.ToolNames) != 1 || n.ToolNames[0] != "search" {
-		t.Fatalf("unexpected tool names: %#v", n.ToolNames)
-	}
-}
-
-func TestNormalizeOpenAIResponsesRequestToolChoiceForcedFunction(t *testing.T) {
-	store := newEmptyStoreForNormalizeTest(t)
-	req := map[string]any{
-		"model": "gpt-4o",
-		"input": "ping",
-		"tools": []any{
-			map[string]any{
-				"type": "function",
-				"function": map[string]any{
-					"name": "search",
-				},
-			},
-			map[string]any{
-				"type": "function",
-				"function": map[string]any{
-					"name": "read_file",
-				},
-			},
-		},
-		"tool_choice": map[string]any{
-			"type": "function",
-			"name": "read_file",
-		},
-	}
-	n, err := normalizeOpenAIResponsesRequest(store, req, "")
-	if err != nil {
-		t.Fatalf("normalize failed: %v", err)
-	}
-	if n.ToolChoice.Mode != util.ToolChoiceForced {
-		t.Fatalf("expected tool choice mode forced, got %q", n.ToolChoice.Mode)
-	}
-	if n.ToolChoice.ForcedName != "read_file" {
-		t.Fatalf("expected forced tool name read_file, got %q", n.ToolChoice.ForcedName)
-	}
-	if len(n.ToolNames) != 1 || n.ToolNames[0] != "read_file" {
-		t.Fatalf("expected filtered tool names [read_file], got %#v", n.ToolNames)
-	}
-}
-
-func TestNormalizeOpenAIResponsesRequestToolChoiceForcedUndeclaredFails(t *testing.T) {
-	store := newEmptyStoreForNormalizeTest(t)
-	req := map[string]any{
-		"model": "gpt-4o",
-		"input": "ping",
-		"tools": []any{
-			map[string]any{
-				"type": "function",
-				"function": map[string]any{
-					"name": "search",
-				},
-			},
-		},
-		"tool_choice": map[string]any{
-			"type": "function",
-			"name": "read_file",
-		},
-	}
-	if _, err := normalizeOpenAIResponsesRequest(store, req, ""); err == nil {
-		t.Fatalf("expected forced undeclared tool to fail")
-	}
-}
-
-func TestNormalizeOpenAIResponsesRequestToolChoiceNoneKeepsToolDetectionEnabled(t *testing.T) {
-	store := newEmptyStoreForNormalizeTest(t)
-	req := map[string]any{
-		"model": "gpt-4o",
-		"input": "ping",
-		"tools": []any{
-			map[string]any{
-				"type": "function",
-				"function": map[string]any{
-					"name": "search",
-				},
-			},
-		},
-		"tool_choice": "none",
-	}
-	n, err := normalizeOpenAIResponsesRequest(store, req, "")
-	if err != nil {
-		t.Fatalf("normalize failed: %v", err)
-	}
-	if n.ToolChoice.Mode != util.ToolChoiceNone {
-		t.Fatalf("expected tool choice mode none, got %q", n.ToolChoice.Mode)
-	}
-	if len(n.ToolNames) == 0 {
-		t.Fatalf("expected tool detection sentinel when tool_choice=none, got %#v", n.ToolNames)
-	}
-}

internal/adapter/openai/upstream_empty.go

@@ -1,27 +0,0 @@
-package openai
-
-import "net/http"
-
-func shouldWriteUpstreamEmptyOutputError(text string) bool {
-	return text == ""
-}
-
-func upstreamEmptyOutputDetail(contentFilter bool, text, thinking string) (int, string, string) {
-	_ = text
-	if contentFilter {
-		return http.StatusBadRequest, "Upstream content filtered the response and returned no output.", "content_filter"
-	}
-	if thinking != "" {
-		return http.StatusTooManyRequests, "Upstream model returned reasoning without visible output.", "upstream_empty_output"
-	}
-	return http.StatusTooManyRequests, "Upstream model returned empty output.", "upstream_empty_output"
-}
-
-func writeUpstreamEmptyOutputError(w http.ResponseWriter, text string, contentFilter bool) bool {
-	if !shouldWriteUpstreamEmptyOutputError(text) {
-		return false
-	}
-	status, message, code := upstreamEmptyOutputDetail(contentFilter, text, "")
-	writeOpenAIErrorWithCode(w, status, message, code)
-	return true
-}

internal/adapter/openai/vercel_prepare_test.go

@@ -1,83 +0,0 @@
-package openai
-
-import (
-	"ds2api/internal/auth"
-	"net/http/httptest"
-	"testing"
-	"time"
-)
-
-func TestIsVercelStreamPrepareRequest(t *testing.T) {
-	req := httptest.NewRequest("POST", "/v1/chat/completions?__stream_prepare=1", nil)
-	if !isVercelStreamPrepareRequest(req) {
-		t.Fatalf("expected prepare request to be detected")
-	}
-
-	req2 := httptest.NewRequest("POST", "/v1/chat/completions", nil)
-	if isVercelStreamPrepareRequest(req2) {
-		t.Fatalf("expected non-prepare request")
-	}
-}
-
-func TestIsVercelStreamReleaseRequest(t *testing.T) {
-	req := httptest.NewRequest("POST", "/v1/chat/completions?__stream_release=1", nil)
-	if !isVercelStreamReleaseRequest(req) {
-		t.Fatalf("expected release request to be detected")
-	}
-
-	req2 := httptest.NewRequest("POST", "/v1/chat/completions", nil)
-	if isVercelStreamReleaseRequest(req2) {
-		t.Fatalf("expected non-release request")
-	}
-}
-
-func TestVercelInternalSecret(t *testing.T) {
-	t.Run("prefer explicit secret", func(t *testing.T) {
-		t.Setenv("DS2API_VERCEL_INTERNAL_SECRET", "stream-secret")
-		t.Setenv("DS2API_ADMIN_KEY", "admin-fallback")
-		if got := vercelInternalSecret(); got != "stream-secret" {
-			t.Fatalf("expected explicit secret, got %q", got)
-		}
-	})
-
-	t.Run("fallback to admin key", func(t *testing.T) {
-		t.Setenv("DS2API_VERCEL_INTERNAL_SECRET", "")
-		t.Setenv("DS2API_ADMIN_KEY", "admin-fallback")
-		if got := vercelInternalSecret(); got != "admin-fallback" {
-			t.Fatalf("expected admin key fallback, got %q", got)
-		}
-	})
-
-	t.Run("default admin when env missing", func(t *testing.T) {
-		t.Setenv("DS2API_VERCEL_INTERNAL_SECRET", "")
-		t.Setenv("DS2API_ADMIN_KEY", "")
-		if got := vercelInternalSecret(); got != "admin" {
-			t.Fatalf("expected default admin fallback, got %q", got)
-		}
-	})
-}
-
-func TestStreamLeaseLifecycle(t *testing.T) {
-	h := &Handler{}
-	leaseID := h.holdStreamLease(&auth.RequestAuth{UseConfigToken: false})
-	if leaseID == "" {
-		t.Fatalf("expected non-empty lease id")
-	}
-	if ok := h.releaseStreamLease(leaseID); !ok {
-		t.Fatalf("expected lease release success")
-	}
-	if ok := h.releaseStreamLease(leaseID); ok {
-		t.Fatalf("expected duplicate release to fail")
-	}
-}
-
-func TestStreamLeaseTTL(t *testing.T) {
-	t.Setenv("DS2API_VERCEL_STREAM_LEASE_TTL_SECONDS", "120")
-	if got := streamLeaseTTL(); got != 120*time.Second {
-		t.Fatalf("expected ttl=120s, got %v", got)
-	}
-	t.Setenv("DS2API_VERCEL_STREAM_LEASE_TTL_SECONDS", "invalid")
-	if got := streamLeaseTTL(); got != 15*time.Minute {
-		t.Fatalf("expected default ttl on invalid value, got %v", got)
-	}
-}

internal/admin/handler.go

@@ -1,65 +0,0 @@
-package admin
-
-import (
-	"github.com/go-chi/chi/v5"
-
-	"ds2api/internal/chathistory"
-)
-
-type Handler struct {
-	Store       ConfigStore
-	Pool        PoolController
-	DS          DeepSeekCaller
-	OpenAI      OpenAIChatCaller
-	ChatHistory *chathistory.Store
-}
-
-func RegisterRoutes(r chi.Router, h *Handler) {
-	r.Post("/login", h.login)
-	r.Get("/verify", h.verify)
-	r.Group(func(pr chi.Router) {
-		pr.Use(h.requireAdmin)
-		pr.Get("/vercel/config", h.getVercelConfig)
-		pr.Get("/config", h.getConfig)
-		pr.Post("/config", h.updateConfig)
-		pr.Get("/settings", h.getSettings)
-		pr.Put("/settings", h.updateSettings)
-		pr.Post("/settings/password", h.updateSettingsPassword)
-		pr.Post("/config/import", h.configImport)
-		pr.Get("/config/export", h.configExport)
-		pr.Post("/keys", h.addKey)
-		pr.Put("/keys/{key}", h.updateKey)
-		pr.Delete("/keys/{key}", h.deleteKey)
-		pr.Get("/proxies", h.listProxies)
-		pr.Post("/proxies", h.addProxy)
-		pr.Put("/proxies/{proxyID}", h.updateProxy)
-		pr.Delete("/proxies/{proxyID}", h.deleteProxy)
-		pr.Post("/proxies/test", h.testProxy)
-		pr.Get("/accounts", h.listAccounts)
-		pr.Post("/accounts", h.addAccount)
-		pr.Put("/accounts/{identifier}", h.updateAccount)
-		pr.Delete("/accounts/{identifier}", h.deleteAccount)
-		pr.Put("/accounts/{identifier}/proxy", h.updateAccountProxy)
-		pr.Get("/queue/status", h.queueStatus)
-		pr.Post("/accounts/test", h.testSingleAccount)
-		pr.Post("/accounts/test-all", h.testAllAccounts)
-		pr.Post("/accounts/sessions/delete-all", h.deleteAllSessions)
-		pr.Post("/import", h.batchImport)
-		pr.Post("/test", h.testAPI)
-		pr.Post("/dev/raw-samples/capture", h.captureRawSample)
-		pr.Get("/dev/raw-samples/query", h.queryRawSampleCaptures)
-		pr.Post("/dev/raw-samples/save", h.saveRawSampleFromCaptures)
-		pr.Post("/vercel/sync", h.syncVercel)
-		pr.Get("/vercel/status", h.vercelStatus)
-		pr.Post("/vercel/status", h.vercelStatus)
-		pr.Get("/export", h.exportConfig)
-		pr.Get("/dev/captures", h.getDevCaptures)
-		pr.Delete("/dev/captures", h.clearDevCaptures)
-		pr.Get("/chat-history", h.getChatHistory)
-		pr.Get("/chat-history/{id}", h.getChatHistoryItem)
-		pr.Delete("/chat-history", h.clearChatHistory)
-		pr.Delete("/chat-history/{id}", h.deleteChatHistoryItem)
-		pr.Put("/chat-history/settings", h.updateChatHistorySettings)
-		pr.Get("/version", h.getVersion)
-	})
-}

internal/chathistory/store.go

@@ -46,6 +46,7 @@ type Entry struct {
 	Stream           bool           `json:"stream"`
 	UserInput        string         `json:"user_input,omitempty"`
 	Messages         []Message      `json:"messages,omitempty"`
+	HistoryText      string         `json:"history_text,omitempty"`
 	FinalPrompt      string         `json:"final_prompt,omitempty"`
 	ReasoningContent string         `json:"reasoning_content,omitempty"`
 	Content          string         `json:"content,omitempty"`
@@ -94,6 +95,7 @@ type StartParams struct {
 	Stream      bool
 	UserInput   string
 	Messages    []Message
+	HistoryText string
 	FinalPrompt string
 }
 
@@ -244,6 +246,7 @@ func (s *Store) Start(params StartParams) (Entry, error) {
 		Stream:      params.Stream,
 		UserInput:   strings.TrimSpace(params.UserInput),
 		Messages:    cloneMessages(params.Messages),
+		HistoryText: params.HistoryText,
 		FinalPrompt: strings.TrimSpace(params.FinalPrompt),
 	}
 	s.details[entry.ID] = entry

internal/chathistory/store_test.go

@@ -46,7 +46,7 @@ func TestStoreCreatesAndPersistsEntries(t *testing.T) {
 	started, err := store.Start(StartParams{
 		CallerID:  "caller:abc",
 		AccountID: "user@example.com",
-		Model:     "deepseek-chat",
+		Model:     "deepseek-v4-flash",
 		Stream:    true,
 		UserInput: "hello",
 	})
@@ -113,7 +113,7 @@ func TestStoreTrimsToConfiguredLimit(t *testing.T) {
 	}
 
 	for i := 0; i < 12; i++ {
-		entry, err := store.Start(StartParams{Model: "deepseek-chat", UserInput: "msg"})
+		entry, err := store.Start(StartParams{Model: "deepseek-v4-flash", UserInput: "msg"})
 		if err != nil {
 			t.Fatalf("start %d failed: %v", i, err)
 		}
@@ -197,7 +197,7 @@ func TestStoreConcurrentUpdatesKeepSplitFilesValid(t *testing.T) {
 			defer wg.Done()
 			entry, err := store.Start(StartParams{
 				CallerID:  "caller:test",
-				Model:     "deepseek-chat",
+				Model:     "deepseek-v4-flash",
 				UserInput: "hello",
 			})
 			if err != nil {
@@ -299,7 +299,7 @@ func TestStoreAutoMigratesMetadataOnlyLegacyMonolith(t *testing.T) {
 			Status:       "error",
 			CallerID:     "caller:test",
 			AccountID:    "acct:test",
-			Model:        "deepseek-chat",
+			Model:        "deepseek-v4-flash",
 			Stream:       true,
 			UserInput:    "hello",
 			Error:        "boom",

internal/claudeconv/convert.go

@@ -1,32 +1,21 @@
 package claudeconv
 
-import "strings"
+import (
+	"strings"
 
-type ClaudeMappingProvider interface {
-	ClaudeMapping() map[string]string
-}
+	"ds2api/internal/config"
+)
 
-func ConvertClaudeToDeepSeek(claudeReq map[string]any, mappingProvider ClaudeMappingProvider, defaultClaudeModel string) map[string]any {
+func ConvertClaudeToDeepSeek(claudeReq map[string]any, aliasProvider config.ModelAliasReader, defaultClaudeModel string) map[string]any {
 	messages, _ := claudeReq["messages"].([]any)
 	model, _ := claudeReq["model"].(string)
 	if model == "" {
 		model = defaultClaudeModel
 	}
 
-	mapping := map[string]string{}
-	if mappingProvider != nil {
-		mapping = mappingProvider.ClaudeMapping()
-	}
-	dsModel := mapping["fast"]
-	if dsModel == "" {
-		dsModel = "deepseek-chat"
-	}
-
-	modelLower := strings.ToLower(model)
-	if strings.Contains(modelLower, "opus") || strings.Contains(modelLower, "reasoner") || strings.Contains(modelLower, "slow") {
-		if slow := mapping["slow"]; slow != "" {
-			dsModel = slow
-		}
+	dsModel, ok := config.ResolveModel(aliasProvider, model)
+	if !ok || strings.TrimSpace(dsModel) == "" {
+		dsModel = "deepseek-v4-flash"
 	}
 
 	convertedMessages := make([]any, 0, len(messages)+1)

internal/config/codec.go

@@ -26,12 +26,6 @@ func (c Config) MarshalJSON() ([]byte, error) {
 	if len(c.Proxies) > 0 {
 		m["proxies"] = c.Proxies
 	}
-	if len(c.ClaudeMapping) > 0 {
-		m["claude_mapping"] = c.ClaudeMapping
-	}
-	if len(c.ClaudeModelMap) > 0 {
-		m["claude_model_mapping"] = c.ClaudeModelMap
-	}
 	if len(c.ModelAliases) > 0 {
 		m["model_aliases"] = c.ModelAliases
 	}
@@ -88,13 +82,8 @@ func (c *Config) UnmarshalJSON(b []byte) error {
 				return fmt.Errorf("invalid field %q: %w", k, err)
 			}
 		case "claude_mapping":
-			if err := json.Unmarshal(v, &c.ClaudeMapping); err != nil {
-				return fmt.Errorf("invalid field %q: %w", k, err)
-			}
 		case "claude_model_mapping":
-			if err := json.Unmarshal(v, &c.ClaudeModelMap); err != nil {
-				return fmt.Errorf("invalid field %q: %w", k, err)
-			}
+			// Removed legacy mapping fields are ignored instead of persisted.
 		case "model_aliases":
 			if err := json.Unmarshal(v, &c.ModelAliases); err != nil {
 				return fmt.Errorf("invalid field %q: %w", k, err)
@@ -150,15 +139,13 @@ func (c *Config) UnmarshalJSON(b []byte) error {
 
 func (c Config) Clone() Config {
 	clone := Config{
-		Keys:           slices.Clone(c.Keys),
-		APIKeys:        slices.Clone(c.APIKeys),
-		Accounts:       slices.Clone(c.Accounts),
-		Proxies:        slices.Clone(c.Proxies),
-		ClaudeMapping:  cloneStringMap(c.ClaudeMapping),
-		ClaudeModelMap: cloneStringMap(c.ClaudeModelMap),
-		ModelAliases:   cloneStringMap(c.ModelAliases),
-		Admin:          c.Admin,
-		Runtime:        c.Runtime,
+		Keys:         slices.Clone(c.Keys),
+		APIKeys:      slices.Clone(c.APIKeys),
+		Accounts:     slices.Clone(c.Accounts),
+		Proxies:      slices.Clone(c.Proxies),
+		ModelAliases: cloneStringMap(c.ModelAliases),
+		Admin:        c.Admin,
+		Runtime:      c.Runtime,
 		Compat: CompatConfig{
 			WideInputStrictOutput: cloneBoolPtr(c.Compat.WideInputStrictOutput),
 			StripReferenceMarkers: cloneBoolPtr(c.Compat.StripReferenceMarkers),

internal/config/config.go

@@ -12,8 +12,6 @@ type Config struct {
 	APIKeys          []APIKey           `json:"api_keys,omitempty"`
 	Accounts         []Account          `json:"accounts,omitempty"`
 	Proxies          []Proxy            `json:"proxies,omitempty"`
-	ClaudeMapping    map[string]string  `json:"claude_mapping,omitempty"`
-	ClaudeModelMap   map[string]string  `json:"claude_model_mapping,omitempty"`
 	ModelAliases     map[string]string  `json:"model_aliases,omitempty"`
 	Admin            AdminConfig        `json:"admin,omitempty"`
 	Runtime          RuntimeConfig      `json:"runtime,omitempty"`
@@ -100,6 +98,9 @@ func (c *Config) NormalizeCredentials() {
 		c.Accounts[i].Name = strings.TrimSpace(c.Accounts[i].Name)
 		c.Accounts[i].Remark = strings.TrimSpace(c.Accounts[i].Remark)
 	}
+
+	c.normalizeModelAliases()
+	c.forceHistorySplitEnabled()
 }
 
 // DropInvalidAccounts removes accounts that cannot be addressed by admin APIs
@@ -119,6 +120,35 @@ func (c *Config) DropInvalidAccounts() {
 	c.Accounts = kept
 }
 
+func (c *Config) normalizeModelAliases() {
+	if c == nil {
+		return
+	}
+
+	aliases := map[string]string{}
+	for k, v := range c.ModelAliases {
+		key := strings.TrimSpace(lower(k))
+		val := strings.TrimSpace(lower(v))
+		if key == "" || val == "" {
+			continue
+		}
+		aliases[key] = val
+	}
+	if len(aliases) == 0 {
+		c.ModelAliases = nil
+	} else {
+		c.ModelAliases = aliases
+	}
+}
+
+func (c *Config) forceHistorySplitEnabled() {
+	if c == nil {
+		return
+	}
+	enabled := true
+	c.HistorySplit.Enabled = &enabled
+}
+
 type CompatConfig struct {
 	WideInputStrictOutput *bool `json:"wide_input_strict_output,omitempty"`
 	StripReferenceMarkers *bool `json:"strip_reference_markers,omitempty"`

internal/config/config_edge_test.go

@@ -10,19 +10,19 @@ import (
 // ─── GetModelConfig edge cases ───────────────────────────────────────
 
 func TestGetModelConfigDeepSeekChat(t *testing.T) {
-	thinking, search, ok := GetModelConfig("deepseek-chat")
+	thinking, search, ok := GetModelConfig("deepseek-v4-flash")
 	if !ok {
-		t.Fatal("expected ok for deepseek-chat")
+		t.Fatal("expected ok for deepseek-v4-flash")
 	}
-	if thinking || search {
-		t.Fatalf("expected no thinking/search for deepseek-chat, got thinking=%v search=%v", thinking, search)
+	if !thinking || search {
+		t.Fatalf("expected thinking=true search=false for deepseek-v4-flash, got thinking=%v search=%v", thinking, search)
 	}
 }
 
 func TestGetModelConfigDeepSeekReasoner(t *testing.T) {
-	thinking, search, ok := GetModelConfig("deepseek-reasoner")
+	thinking, search, ok := GetModelConfig("deepseek-v4-pro")
 	if !ok {
-		t.Fatal("expected ok for deepseek-reasoner")
+		t.Fatal("expected ok for deepseek-v4-pro")
 	}
 	if !thinking || search {
 		t.Fatalf("expected thinking=true search=false, got thinking=%v search=%v", thinking, search)
@@ -30,19 +30,19 @@ func TestGetModelConfigDeepSeekReasoner(t *testing.T) {
 }
 
 func TestGetModelConfigDeepSeekChatSearch(t *testing.T) {
-	thinking, search, ok := GetModelConfig("deepseek-chat-search")
+	thinking, search, ok := GetModelConfig("deepseek-v4-flash-search")
 	if !ok {
-		t.Fatal("expected ok for deepseek-chat-search")
+		t.Fatal("expected ok for deepseek-v4-flash-search")
 	}
-	if thinking || !search {
-		t.Fatalf("expected thinking=false search=true, got thinking=%v search=%v", thinking, search)
+	if !thinking || !search {
+		t.Fatalf("expected thinking=true search=true, got thinking=%v search=%v", thinking, search)
 	}
 }
 
 func TestGetModelConfigDeepSeekReasonerSearch(t *testing.T) {
-	thinking, search, ok := GetModelConfig("deepseek-reasoner-search")
+	thinking, search, ok := GetModelConfig("deepseek-v4-pro-search")
 	if !ok {
-		t.Fatal("expected ok for deepseek-reasoner-search")
+		t.Fatal("expected ok for deepseek-v4-pro-search")
 	}
 	if !thinking || !search {
 		t.Fatalf("expected both true, got thinking=%v search=%v", thinking, search)
@@ -50,19 +50,19 @@ func TestGetModelConfigDeepSeekReasonerSearch(t *testing.T) {
 }
 
 func TestGetModelConfigDeepSeekExpertChat(t *testing.T) {
-	thinking, search, ok := GetModelConfig("deepseek-expert-chat")
+	thinking, search, ok := GetModelConfig("deepseek-v4-pro")
 	if !ok {
-		t.Fatal("expected ok for deepseek-expert-chat")
+		t.Fatal("expected ok for deepseek-v4-pro")
 	}
-	if thinking || search {
-		t.Fatalf("expected no thinking/search for deepseek-expert-chat, got thinking=%v search=%v", thinking, search)
+	if !thinking || search {
+		t.Fatalf("expected thinking=true search=false for deepseek-v4-pro, got thinking=%v search=%v", thinking, search)
 	}
 }
 
 func TestGetModelConfigDeepSeekExpertReasonerSearch(t *testing.T) {
-	thinking, search, ok := GetModelConfig("deepseek-expert-reasoner-search")
+	thinking, search, ok := GetModelConfig("deepseek-v4-pro-search")
 	if !ok {
-		t.Fatal("expected ok for deepseek-expert-reasoner-search")
+		t.Fatal("expected ok for deepseek-v4-pro-search")
 	}
 	if !thinking || !search {
 		t.Fatalf("expected both true, got thinking=%v search=%v", thinking, search)
@@ -70,9 +70,9 @@ func TestGetModelConfigDeepSeekExpertReasonerSearch(t *testing.T) {
 }
 
 func TestGetModelConfigDeepSeekVisionReasonerSearch(t *testing.T) {
-	thinking, search, ok := GetModelConfig("deepseek-vision-reasoner-search")
+	thinking, search, ok := GetModelConfig("deepseek-v4-vision-search")
 	if !ok {
-		t.Fatal("expected ok for deepseek-vision-reasoner-search")
+		t.Fatal("expected ok for deepseek-v4-vision-search")
 	}
 	if !thinking || !search {
 		t.Fatalf("expected both true, got thinking=%v search=%v", thinking, search)
@@ -80,27 +80,27 @@ func TestGetModelConfigDeepSeekVisionReasonerSearch(t *testing.T) {
 }
 
 func TestGetModelTypeDefaultExpertAndVision(t *testing.T) {
-	defaultType, ok := GetModelType("deepseek-chat")
+	defaultType, ok := GetModelType("deepseek-v4-flash")
 	if !ok || defaultType != "default" {
 		t.Fatalf("expected default model_type, got ok=%v model_type=%q", ok, defaultType)
 	}
-	expertType, ok := GetModelType("deepseek-expert-chat")
+	expertType, ok := GetModelType("deepseek-v4-pro")
 	if !ok || expertType != "expert" {
 		t.Fatalf("expected expert model_type, got ok=%v model_type=%q", ok, expertType)
 	}
-	visionType, ok := GetModelType("deepseek-vision-chat")
+	visionType, ok := GetModelType("deepseek-v4-vision")
 	if !ok || visionType != "vision" {
 		t.Fatalf("expected vision model_type, got ok=%v model_type=%q", ok, visionType)
 	}
 }
 
 func TestGetModelConfigCaseInsensitive(t *testing.T) {
-	thinking, search, ok := GetModelConfig("DeepSeek-Chat")
+	thinking, search, ok := GetModelConfig("DeepSeek-V4-Flash")
 	if !ok {
-		t.Fatal("expected ok for case-insensitive deepseek-chat")
+		t.Fatal("expected ok for case-insensitive deepseek-v4-flash")
 	}
-	if thinking || search {
-		t.Fatalf("expected no thinking/search for case-insensitive deepseek-chat")
+	if !thinking || search {
+		t.Fatalf("expected thinking=true search=false for case-insensitive deepseek-v4-flash")
 	}
 }
 
@@ -145,12 +145,9 @@ func TestConfigJSONRoundtrip(t *testing.T) {
 	trueVal := true
 	falseVal := false
 	cfg := Config{
-		Keys:     []string{"key1", "key2"},
-		Accounts: []Account{{Email: "user@example.com", Password: "pass", Token: "tok"}},
-		ClaudeMapping: map[string]string{
-			"fast": "deepseek-chat",
-			"slow": "deepseek-reasoner",
-		},
+		Keys:         []string{"key1", "key2"},
+		Accounts:     []Account{{Email: "user@example.com", Password: "pass", Token: "tok"}},
+		ModelAliases: map[string]string{"Claude-Sonnet-4-6": "DeepSeek-V4-Flash"},
 		AutoDelete: AutoDeleteConfig{
 			Mode: "single",
 		},
@@ -188,8 +185,8 @@ func TestConfigJSONRoundtrip(t *testing.T) {
 	if len(decoded.Accounts) != 1 || decoded.Accounts[0].Email != "user@example.com" {
 		t.Fatalf("unexpected accounts: %#v", decoded.Accounts)
 	}
-	if decoded.ClaudeMapping["fast"] != "deepseek-chat" {
-		t.Fatalf("unexpected claude mapping: %#v", decoded.ClaudeMapping)
+	if decoded.ModelAliases["claude-sonnet-4-6"] != "deepseek-v4-flash" {
+		t.Fatalf("unexpected normalized model aliases: %#v", decoded.ModelAliases)
 	}
 	if decoded.Runtime.TokenRefreshIntervalHours != 12 {
 		t.Fatalf("unexpected runtime refresh interval: %#v", decoded.Runtime.TokenRefreshIntervalHours)
@@ -255,6 +252,23 @@ func TestConfigUnmarshalJSONPreservesUnknownFields(t *testing.T) {
 	}
 }
 
+func TestConfigUnmarshalJSONIgnoresRemovedLegacyModelMappings(t *testing.T) {
+	raw := `{"keys":["k1"],"accounts":[],"claude_mapping":{"fast":"deepseek-v4-pro"},"claude_model_mapping":{"slow":"deepseek-v4-pro"}}`
+	var cfg Config
+	if err := json.Unmarshal([]byte(raw), &cfg); err != nil {
+		t.Fatalf("unmarshal error: %v", err)
+	}
+	if len(cfg.ModelAliases) != 0 {
+		t.Fatalf("expected removed legacy mappings to be ignored, got %#v", cfg.ModelAliases)
+	}
+	if _, ok := cfg.AdditionalFields["claude_mapping"]; ok {
+		t.Fatalf("expected removed legacy field not to persist in additional fields: %#v", cfg.AdditionalFields)
+	}
+	if _, ok := cfg.AdditionalFields["claude_model_mapping"]; ok {
+		t.Fatalf("expected removed legacy field not to persist in additional fields: %#v", cfg.AdditionalFields)
+	}
+}
+
 // ─── Config.Clone ────────────────────────────────────────────────────
 
 func TestConfigCloneIsDeepCopy(t *testing.T) {
@@ -262,11 +276,9 @@ func TestConfigCloneIsDeepCopy(t *testing.T) {
 	trueVal := true
 	turns := 2
 	cfg := Config{
-		Keys:     []string{"key1"},
-		Accounts: []Account{{Email: "user@test.com", Token: "token"}},
-		ClaudeMapping: map[string]string{
-			"fast": "deepseek-chat",
-		},
+		Keys:         []string{"key1"},
+		Accounts:     []Account{{Email: "user@test.com", Token: "token"}},
+		ModelAliases: map[string]string{"claude-sonnet-4-6": "deepseek-v4-flash"},
 		Compat: CompatConfig{
 			StripReferenceMarkers: &falseVal,
 		},
@@ -282,7 +294,7 @@ func TestConfigCloneIsDeepCopy(t *testing.T) {
 	// Modify original
 	cfg.Keys[0] = "modified"
 	cfg.Accounts[0].Email = "modified@test.com"
-	cfg.ClaudeMapping["fast"] = "modified-model"
+	cfg.ModelAliases["claude-sonnet-4-6"] = "modified-model"
 	if cfg.Compat.StripReferenceMarkers != nil {
 		*cfg.Compat.StripReferenceMarkers = true
 	}
@@ -300,8 +312,8 @@ func TestConfigCloneIsDeepCopy(t *testing.T) {
 	if cloned.Accounts[0].Email != "user@test.com" {
 		t.Fatalf("clone accounts was affected: %#v", cloned.Accounts)
 	}
-	if cloned.ClaudeMapping["fast"] != "deepseek-chat" {
-		t.Fatalf("clone claude mapping was affected: %#v", cloned.ClaudeMapping)
+	if cloned.ModelAliases["claude-sonnet-4-6"] != "deepseek-v4-flash" {
+		t.Fatalf("clone model aliases was affected: %#v", cloned.ModelAliases)
 	}
 	if cloned.Compat.StripReferenceMarkers == nil || *cloned.Compat.StripReferenceMarkers {
 		t.Fatalf("clone compat was affected: %#v", cloned.Compat.StripReferenceMarkers)
@@ -652,25 +664,27 @@ func TestNormalizeCredentialsPrefersStructuredAPIKeys(t *testing.T) {
 	}
 }
 
-func TestStoreClaudeMapping(t *testing.T) {
-	t.Setenv("DS2API_CONFIG_JSON", `{"keys":[],"accounts":[],"claude_mapping":{"fast":"deepseek-chat","slow":"deepseek-reasoner"}}`)
+func TestStoreModelAliasesIncludesDefaultsAndOverrides(t *testing.T) {
+	t.Setenv("DS2API_CONFIG_JSON", `{"keys":[],"accounts":[],"model_aliases":{"claude-opus-4-6":"deepseek-v4-pro-search"}}`)
 	store := LoadStore()
-	mapping := store.ClaudeMapping()
-	if mapping["fast"] != "deepseek-chat" {
-		t.Fatalf("unexpected fast mapping: %q", mapping["fast"])
+	aliases := store.ModelAliases()
+	if aliases["claude-sonnet-4-6"] != "deepseek-v4-flash" {
+		t.Fatalf("expected default alias to remain available, got %q", aliases["claude-sonnet-4-6"])
 	}
-	if mapping["slow"] != "deepseek-reasoner" {
-		t.Fatalf("unexpected slow mapping: %q", mapping["slow"])
+	if aliases["claude-opus-4-6"] != "deepseek-v4-pro-search" {
+		t.Fatalf("expected custom alias override, got %q", aliases["claude-opus-4-6"])
 	}
 }
 
-func TestStoreClaudeMappingEmpty(t *testing.T) {
+func TestStoreModelAliasesDefault(t *testing.T) {
 	t.Setenv("DS2API_CONFIG_JSON", `{"keys":[],"accounts":[]}`)
 	store := LoadStore()
-	mapping := store.ClaudeMapping()
-	// Even without config mapping, there are defaults
-	if mapping == nil {
-		t.Fatal("expected non-nil mapping (may contain defaults)")
+	aliases := store.ModelAliases()
+	if aliases == nil {
+		t.Fatal("expected non-nil aliases")
+	}
+	if aliases["claude-sonnet-4-6"] != "deepseek-v4-flash" {
+		t.Fatalf("expected built-in alias, got %q", aliases["claude-sonnet-4-6"])
 	}
 }
 
@@ -720,18 +734,12 @@ func TestOpenAIModelsResponse(t *testing.T) {
 		t.Fatal("expected non-empty models list")
 	}
 	expected := map[string]bool{
-		"deepseek-chat":                   false,
-		"deepseek-reasoner":               false,
-		"deepseek-chat-search":            false,
-		"deepseek-reasoner-search":        false,
-		"deepseek-expert-chat":            false,
-		"deepseek-expert-reasoner":        false,
-		"deepseek-expert-chat-search":     false,
-		"deepseek-expert-reasoner-search": false,
-		"deepseek-vision-chat":            false,
-		"deepseek-vision-reasoner":        false,
-		"deepseek-vision-chat-search":     false,
-		"deepseek-vision-reasoner-search": false,
+		"deepseek-v4-flash":         false,
+		"deepseek-v4-pro":           false,
+		"deepseek-v4-flash-search":  false,
+		"deepseek-v4-pro-search":    false,
+		"deepseek-v4-vision":        false,
+		"deepseek-v4-vision-search": false,
 	}
 	for _, model := range data {
 		if _, ok := expected[model.ID]; ok {

internal/config/model_alias_test.go

@@ -7,22 +7,63 @@ type mockModelAliasReader map[string]string
 func (m mockModelAliasReader) ModelAliases() map[string]string { return m }
 
 func TestResolveModelDirectDeepSeek(t *testing.T) {
-	got, ok := ResolveModel(nil, "deepseek-chat")
-	if !ok || got != "deepseek-chat" {
-		t.Fatalf("expected deepseek-chat, got ok=%v model=%q", ok, got)
+	got, ok := ResolveModel(nil, "deepseek-v4-flash")
+	if !ok || got != "deepseek-v4-flash" {
+		t.Fatalf("expected deepseek-v4-flash, got ok=%v model=%q", ok, got)
 	}
 }
 
 func TestResolveModelAlias(t *testing.T) {
 	got, ok := ResolveModel(nil, "gpt-4.1")
-	if !ok || got != "deepseek-chat" {
-		t.Fatalf("expected alias gpt-4.1 -> deepseek-chat, got ok=%v model=%q", ok, got)
+	if !ok || got != "deepseek-v4-flash" {
+		t.Fatalf("expected alias gpt-4.1 -> deepseek-v4-flash, got ok=%v model=%q", ok, got)
+	}
+}
+
+func TestResolveLatestOpenAIAlias(t *testing.T) {
+	got, ok := ResolveModel(nil, "gpt-5.5")
+	if !ok || got != "deepseek-v4-flash" {
+		t.Fatalf("expected alias gpt-5.5 -> deepseek-v4-flash, got ok=%v model=%q", ok, got)
+	}
+}
+
+func TestResolveLatestClaudeAlias(t *testing.T) {
+	got, ok := ResolveModel(nil, "claude-sonnet-4-6")
+	if !ok || got != "deepseek-v4-flash" {
+		t.Fatalf("expected alias claude-sonnet-4-6 -> deepseek-v4-flash, got ok=%v model=%q", ok, got)
+	}
+}
+
+func TestResolveExpandedHistoricalAliases(t *testing.T) {
+	cases := []struct {
+		name  string
+		model string
+		want  string
+	}{
+		{name: "openai old chatgpt", model: "chatgpt-4o", want: "deepseek-v4-flash"},
+		{name: "openai codex max", model: "gpt-5.1-codex-max", want: "deepseek-v4-pro"},
+		{name: "openai deep research", model: "o3-deep-research", want: "deepseek-v4-pro-search"},
+		{name: "openai historical reasoning", model: "o1-preview", want: "deepseek-v4-pro"},
+		{name: "claude latest historical", model: "claude-3-5-sonnet-latest", want: "deepseek-v4-flash"},
+		{name: "claude historical opus", model: "claude-3-opus-20240229", want: "deepseek-v4-pro"},
+		{name: "claude historical haiku", model: "claude-3-haiku-20240307", want: "deepseek-v4-flash"},
+		{name: "gemini latest alias", model: "gemini-flash-latest", want: "deepseek-v4-flash"},
+		{name: "gemini historical pro", model: "gemini-1.5-pro", want: "deepseek-v4-pro"},
+		{name: "gemini vision legacy", model: "gemini-pro-vision", want: "deepseek-v4-vision"},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			got, ok := ResolveModel(nil, tc.model)
+			if !ok || got != tc.want {
+				t.Fatalf("expected alias %s -> %s, got ok=%v model=%q", tc.model, tc.want, ok, got)
+			}
+		})
 	}
 }
 
 func TestResolveModelHeuristicReasoner(t *testing.T) {
 	got, ok := ResolveModel(nil, "o3-super")
-	if !ok || got != "deepseek-reasoner" {
+	if !ok || got != "deepseek-v4-pro" {
 		t.Fatalf("expected heuristic reasoner, got ok=%v model=%q", ok, got)
 	}
 }
@@ -34,28 +75,58 @@ func TestResolveModelUnknown(t *testing.T) {
 	}
 }
 
+func TestResolveModelRejectsLegacyDeepSeekIDs(t *testing.T) {
+	legacyModels := []string{
+		"deepseek-chat",
+		"deepseek-reasoner",
+		"deepseek-chat-search",
+		"deepseek-reasoner-search",
+		"deepseek-expert-chat",
+		"deepseek-expert-reasoner",
+		"deepseek-vision-chat",
+	}
+	for _, model := range legacyModels {
+		if got, ok := ResolveModel(nil, model); ok {
+			t.Fatalf("expected legacy model %q to be rejected, got %q", model, got)
+		}
+	}
+}
+
+func TestResolveModelRejectsRetiredHistoricalModels(t *testing.T) {
+	retiredModels := []string{
+		"claude-2.1",
+		"claude-instant-1.2",
+		"gpt-3.5-turbo",
+	}
+	for _, model := range retiredModels {
+		if got, ok := ResolveModel(nil, model); ok {
+			t.Fatalf("expected retired model %q to be rejected, got %q", model, got)
+		}
+	}
+}
+
 func TestResolveModelDirectDeepSeekExpert(t *testing.T) {
-	got, ok := ResolveModel(nil, "deepseek-expert-chat")
-	if !ok || got != "deepseek-expert-chat" {
-		t.Fatalf("expected deepseek-expert-chat, got ok=%v model=%q", ok, got)
+	got, ok := ResolveModel(nil, "deepseek-v4-pro")
+	if !ok || got != "deepseek-v4-pro" {
+		t.Fatalf("expected deepseek-v4-pro, got ok=%v model=%q", ok, got)
 	}
 }
 
 func TestResolveModelCustomAliasToExpert(t *testing.T) {
 	got, ok := ResolveModel(mockModelAliasReader{
-		"my-expert-model": "deepseek-expert-reasoner-search",
+		"my-expert-model": "deepseek-v4-pro-search",
 	}, "my-expert-model")
-	if !ok || got != "deepseek-expert-reasoner-search" {
-		t.Fatalf("expected alias -> deepseek-expert-reasoner-search, got ok=%v model=%q", ok, got)
+	if !ok || got != "deepseek-v4-pro-search" {
+		t.Fatalf("expected alias -> deepseek-v4-pro-search, got ok=%v model=%q", ok, got)
 	}
 }
 
 func TestResolveModelCustomAliasToVision(t *testing.T) {
 	got, ok := ResolveModel(mockModelAliasReader{
-		"my-vision-model": "deepseek-vision-chat-search",
+		"my-vision-model": "deepseek-v4-vision-search",
 	}, "my-vision-model")
-	if !ok || got != "deepseek-vision-chat-search" {
-		t.Fatalf("expected alias -> deepseek-vision-chat-search, got ok=%v model=%q", ok, got)
+	if !ok || got != "deepseek-v4-vision-search" {
+		t.Fatalf("expected alias -> deepseek-v4-vision-search, got ok=%v model=%q", ok, got)
 	}
 }
 

internal/config/models.go

@@ -15,28 +15,22 @@ type ModelAliasReader interface {
 }
 
 var DeepSeekModels = []ModelInfo{
-	{ID: "deepseek-chat", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
-	{ID: "deepseek-reasoner", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
-	{ID: "deepseek-chat-search", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
-	{ID: "deepseek-reasoner-search", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
-	{ID: "deepseek-expert-chat", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
-	{ID: "deepseek-expert-reasoner", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
-	{ID: "deepseek-expert-chat-search", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
-	{ID: "deepseek-expert-reasoner-search", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
-	{ID: "deepseek-vision-chat", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
-	{ID: "deepseek-vision-reasoner", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
-	{ID: "deepseek-vision-chat-search", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
-	{ID: "deepseek-vision-reasoner-search", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
+	{ID: "deepseek-v4-flash", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
+	{ID: "deepseek-v4-pro", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
+	{ID: "deepseek-v4-flash-search", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
+	{ID: "deepseek-v4-pro-search", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
+	{ID: "deepseek-v4-vision", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
+	{ID: "deepseek-v4-vision-search", Object: "model", Created: 1677610602, OwnedBy: "deepseek", Permission: []any{}},
 }
 
 var ClaudeModels = []ModelInfo{
 	// Current aliases
 	{ID: "claude-opus-4-6", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
-	{ID: "claude-sonnet-4-5", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
+	{ID: "claude-sonnet-4-6", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
 	{ID: "claude-haiku-4-5", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
 
-	// Current snapshots
-	{ID: "claude-opus-4-5-20251101", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
+	// Claude 4.x snapshots and prior aliases kept for compatibility
+	{ID: "claude-sonnet-4-5", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
 	{ID: "claude-opus-4-1", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
 	{ID: "claude-opus-4-1-20250805", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
 	{ID: "claude-opus-4-0", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
@@ -57,44 +51,13 @@ var ClaudeModels = []ModelInfo{
 	{ID: "claude-3-5-haiku-latest", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
 	{ID: "claude-3-5-haiku-20241022", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
 	{ID: "claude-3-haiku-20240307", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
-
-	// Claude 2.x and 1.x (retired but accepted for compatibility)
-	{ID: "claude-2.1", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
-	{ID: "claude-2.0", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
-	{ID: "claude-1.3", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
-	{ID: "claude-1.2", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
-	{ID: "claude-1.1", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
-	{ID: "claude-1.0", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
-	{ID: "claude-instant-1.2", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
-	{ID: "claude-instant-1.1", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
-	{ID: "claude-instant-1.0", Object: "model", Created: 1715635200, OwnedBy: "anthropic"},
 }
 
 func GetModelConfig(model string) (thinking bool, search bool, ok bool) {
 	switch lower(model) {
-	case "deepseek-chat":
-		return false, false, true
-	case "deepseek-reasoner":
+	case "deepseek-v4-flash", "deepseek-v4-pro", "deepseek-v4-vision":
 		return true, false, true
-	case "deepseek-chat-search":
-		return false, true, true
-	case "deepseek-reasoner-search":
-		return true, true, true
-	case "deepseek-expert-chat":
-		return false, false, true
-	case "deepseek-expert-reasoner":
-		return true, false, true
-	case "deepseek-expert-chat-search":
-		return false, true, true
-	case "deepseek-expert-reasoner-search":
-		return true, true, true
-	case "deepseek-vision-chat":
-		return false, false, true
-	case "deepseek-vision-reasoner":
-		return true, false, true
-	case "deepseek-vision-chat-search":
-		return false, true, true
-	case "deepseek-vision-reasoner-search":
+	case "deepseek-v4-flash-search", "deepseek-v4-pro-search", "deepseek-v4-vision-search":
 		return true, true, true
 	default:
 		return false, false, false
@@ -103,11 +66,11 @@ func GetModelConfig(model string) (thinking bool, search bool, ok bool) {
 
 func GetModelType(model string) (modelType string, ok bool) {
 	switch lower(model) {
-	case "deepseek-chat", "deepseek-reasoner", "deepseek-chat-search", "deepseek-reasoner-search":
+	case "deepseek-v4-flash", "deepseek-v4-flash-search":
 		return "default", true
-	case "deepseek-expert-chat", "deepseek-expert-reasoner", "deepseek-expert-chat-search", "deepseek-expert-reasoner-search":
+	case "deepseek-v4-pro", "deepseek-v4-pro-search":
 		return "expert", true
-	case "deepseek-vision-chat", "deepseek-vision-reasoner", "deepseek-vision-chat-search", "deepseek-vision-reasoner-search":
+	case "deepseek-v4-vision", "deepseek-v4-vision-search":
 		return "vision", true
 	default:
 		return "", false
@@ -121,27 +84,105 @@ func IsSupportedDeepSeekModel(model string) bool {
 
 func DefaultModelAliases() map[string]string {
 	return map[string]string{
-		"gpt-4o":                 "deepseek-chat",
-		"gpt-4.1":                "deepseek-chat",
-		"gpt-4.1-mini":           "deepseek-chat",
-		"gpt-4.1-nano":           "deepseek-chat",
-		"gpt-5":                  "deepseek-chat",
-		"gpt-5-mini":             "deepseek-chat",
-		"gpt-5-codex":            "deepseek-reasoner",
-		"o1":                     "deepseek-reasoner",
-		"o1-mini":                "deepseek-reasoner",
-		"o3":                     "deepseek-reasoner",
-		"o3-mini":                "deepseek-reasoner",
-		"claude-sonnet-4-5":      "deepseek-chat",
-		"claude-haiku-4-5":       "deepseek-chat",
-		"claude-opus-4-6":        "deepseek-reasoner",
-		"claude-3-5-sonnet":      "deepseek-chat",
-		"claude-3-5-haiku":       "deepseek-chat",
-		"claude-3-opus":          "deepseek-reasoner",
-		"gemini-2.5-pro":         "deepseek-chat",
-		"gemini-2.5-flash":       "deepseek-chat",
-		"llama-3.1-70b-instruct": "deepseek-chat",
-		"qwen-max":               "deepseek-chat",
+		// OpenAI GPT / ChatGPT families
+		"chatgpt-4o":          "deepseek-v4-flash",
+		"gpt-4":               "deepseek-v4-flash",
+		"gpt-4-turbo":         "deepseek-v4-flash",
+		"gpt-4-turbo-preview": "deepseek-v4-flash",
+		"gpt-4.5-preview":     "deepseek-v4-flash",
+		"gpt-4o":              "deepseek-v4-flash",
+		"gpt-4o-mini":         "deepseek-v4-flash",
+		"gpt-4.1":             "deepseek-v4-flash",
+		"gpt-4.1-mini":        "deepseek-v4-flash",
+		"gpt-4.1-nano":        "deepseek-v4-flash",
+		"gpt-5":               "deepseek-v4-flash",
+		"gpt-5-chat":          "deepseek-v4-flash",
+		"gpt-5.1":             "deepseek-v4-flash",
+		"gpt-5.1-chat":        "deepseek-v4-flash",
+		"gpt-5.2":             "deepseek-v4-flash",
+		"gpt-5.2-chat":        "deepseek-v4-flash",
+		"gpt-5.3-chat":        "deepseek-v4-flash",
+		"gpt-5.4":             "deepseek-v4-flash",
+		"gpt-5.5":             "deepseek-v4-flash",
+		"gpt-5-mini":          "deepseek-v4-flash",
+		"gpt-5-nano":          "deepseek-v4-flash",
+		"gpt-5.4-mini":        "deepseek-v4-flash",
+		"gpt-5.4-nano":        "deepseek-v4-flash",
+		"gpt-5-pro":           "deepseek-v4-pro",
+		"gpt-5.2-pro":         "deepseek-v4-pro",
+		"gpt-5.4-pro":         "deepseek-v4-pro",
+		"gpt-5.5-pro":         "deepseek-v4-pro",
+		"gpt-5-codex":         "deepseek-v4-pro",
+		"gpt-5.1-codex":       "deepseek-v4-pro",
+		"gpt-5.1-codex-mini":  "deepseek-v4-pro",
+		"gpt-5.1-codex-max":   "deepseek-v4-pro",
+		"gpt-5.2-codex":       "deepseek-v4-pro",
+		"gpt-5.3-codex":       "deepseek-v4-pro",
+		"codex-mini-latest":   "deepseek-v4-pro",
+
+		// OpenAI reasoning / research families
+		"o1":                    "deepseek-v4-pro",
+		"o1-preview":            "deepseek-v4-pro",
+		"o1-mini":               "deepseek-v4-pro",
+		"o1-pro":                "deepseek-v4-pro",
+		"o3":                    "deepseek-v4-pro",
+		"o3-mini":               "deepseek-v4-pro",
+		"o3-pro":                "deepseek-v4-pro",
+		"o3-deep-research":      "deepseek-v4-pro-search",
+		"o4-mini":               "deepseek-v4-pro",
+		"o4-mini-deep-research": "deepseek-v4-pro-search",
+
+		// Claude current and historical aliases
+		"claude-opus-4-6":            "deepseek-v4-pro",
+		"claude-opus-4-1":            "deepseek-v4-pro",
+		"claude-opus-4-1-20250805":   "deepseek-v4-pro",
+		"claude-opus-4-0":            "deepseek-v4-pro",
+		"claude-opus-4-20250514":     "deepseek-v4-pro",
+		"claude-sonnet-4-6":          "deepseek-v4-flash",
+		"claude-sonnet-4-5":          "deepseek-v4-flash",
+		"claude-sonnet-4-5-20250929": "deepseek-v4-flash",
+		"claude-sonnet-4-0":          "deepseek-v4-flash",
+		"claude-sonnet-4-20250514":   "deepseek-v4-flash",
+		"claude-haiku-4-5":           "deepseek-v4-flash",
+		"claude-haiku-4-5-20251001":  "deepseek-v4-flash",
+		"claude-3-7-sonnet":          "deepseek-v4-flash",
+		"claude-3-7-sonnet-latest":   "deepseek-v4-flash",
+		"claude-3-7-sonnet-20250219": "deepseek-v4-flash",
+		"claude-3-5-sonnet":          "deepseek-v4-flash",
+		"claude-3-5-sonnet-latest":   "deepseek-v4-flash",
+		"claude-3-5-sonnet-20240620": "deepseek-v4-flash",
+		"claude-3-5-sonnet-20241022": "deepseek-v4-flash",
+		"claude-3-5-haiku":           "deepseek-v4-flash",
+		"claude-3-5-haiku-latest":    "deepseek-v4-flash",
+		"claude-3-5-haiku-20241022":  "deepseek-v4-flash",
+		"claude-3-opus":              "deepseek-v4-pro",
+		"claude-3-opus-20240229":     "deepseek-v4-pro",
+		"claude-3-sonnet":            "deepseek-v4-flash",
+		"claude-3-sonnet-20240229":   "deepseek-v4-flash",
+		"claude-3-haiku":             "deepseek-v4-flash",
+		"claude-3-haiku-20240307":    "deepseek-v4-flash",
+
+		// Gemini current and historical text / multimodal models
+		"gemini-pro":            "deepseek-v4-pro",
+		"gemini-pro-vision":     "deepseek-v4-vision",
+		"gemini-pro-latest":     "deepseek-v4-pro",
+		"gemini-flash-latest":   "deepseek-v4-flash",
+		"gemini-1.5-pro":        "deepseek-v4-pro",
+		"gemini-1.5-flash":      "deepseek-v4-flash",
+		"gemini-1.5-flash-8b":   "deepseek-v4-flash",
+		"gemini-2.0-flash":      "deepseek-v4-flash",
+		"gemini-2.0-flash-lite": "deepseek-v4-flash",
+		"gemini-2.5-pro":        "deepseek-v4-pro",
+		"gemini-2.5-flash":      "deepseek-v4-flash",
+		"gemini-2.5-flash-lite": "deepseek-v4-flash",
+		"gemini-3.1-pro":        "deepseek-v4-pro",
+		"gemini-3-pro":          "deepseek-v4-pro",
+		"gemini-3-flash":        "deepseek-v4-flash",
+		"gemini-3.1-flash":      "deepseek-v4-flash",
+		"gemini-3.1-flash-lite": "deepseek-v4-flash",
+
+		"llama-3.1-70b-instruct": "deepseek-v4-flash",
+		"qwen-max":               "deepseek-v4-flash",
 	}
 }
 
@@ -150,6 +191,9 @@ func ResolveModel(store ModelAliasReader, requested string) (string, bool) {
 	if model == "" {
 		return "", false
 	}
+	if isRetiredHistoricalModel(model) {
+		return "", false
+	}
 	if IsSupportedDeepSeekModel(model) {
 		return model, true
 	}
@@ -179,23 +223,44 @@ func ResolveModel(store ModelAliasReader, requested string) (string, bool) {
 		return "", false
 	}
 
+	useVision := strings.Contains(model, "vision")
 	useReasoner := strings.Contains(model, "reason") ||
 		strings.Contains(model, "reasoner") ||
 		strings.HasPrefix(model, "o1") ||
 		strings.HasPrefix(model, "o3") ||
 		strings.Contains(model, "opus") ||
+		strings.Contains(model, "slow") ||
 		strings.Contains(model, "r1")
 	useSearch := strings.Contains(model, "search")
 
 	switch {
+	case useVision && useSearch:
+		return "deepseek-v4-vision-search", true
+	case useVision:
+		return "deepseek-v4-vision", true
 	case useReasoner && useSearch:
-		return "deepseek-reasoner-search", true
+		return "deepseek-v4-pro-search", true
 	case useReasoner:
-		return "deepseek-reasoner", true
+		return "deepseek-v4-pro", true
 	case useSearch:
-		return "deepseek-chat-search", true
+		return "deepseek-v4-flash-search", true
 	default:
-		return "deepseek-chat", true
+		return "deepseek-v4-flash", true
+	}
+}
+
+func isRetiredHistoricalModel(model string) bool {
+	switch {
+	case strings.HasPrefix(model, "claude-1."):
+		return true
+	case strings.HasPrefix(model, "claude-2."):
+		return true
+	case strings.HasPrefix(model, "claude-instant-"):
+		return true
+	case strings.HasPrefix(model, "gpt-3.5"):
+		return true
+	default:
+		return false
 	}
 }
 

internal/config/store_accessors.go

@@ -6,18 +6,6 @@ import (
 	"strings"
 )
 
-func (s *Store) ClaudeMapping() map[string]string {
-	s.mu.RLock()
-	defer s.mu.RUnlock()
-	if len(s.cfg.ClaudeModelMap) > 0 {
-		return cloneStringMap(s.cfg.ClaudeModelMap)
-	}
-	if len(s.cfg.ClaudeMapping) > 0 {
-		return cloneStringMap(s.cfg.ClaudeMapping)
-	}
-	return map[string]string{"fast": "deepseek-chat", "slow": "deepseek-reasoner"}
-}
-
 func (s *Store) ModelAliases() map[string]string {
 	s.mu.RLock()
 	defer s.mu.RUnlock()
@@ -176,12 +164,7 @@ func (s *Store) AutoDeleteSessions() bool {
 }
 
 func (s *Store) HistorySplitEnabled() bool {
-	s.mu.RLock()
-	defer s.mu.RUnlock()
-	if s.cfg.HistorySplit.Enabled == nil {
-		return true
-	}
-	return *s.cfg.HistorySplit.Enabled
+	return true
 }
 
 func (s *Store) HistorySplitTriggerAfterTurns() int {

internal/config/store_accessors_test.go

@@ -18,10 +18,25 @@ func TestStoreHistorySplitAccessors(t *testing.T) {
 		TriggerAfterTurns: &turns,
 	}
 
-	if store.HistorySplitEnabled() {
-		t.Fatal("expected history split disabled after override")
+	if !store.HistorySplitEnabled() {
+		t.Fatal("expected history split to stay enabled after legacy disabled override")
 	}
 	if got := store.HistorySplitTriggerAfterTurns(); got != 3 {
 		t.Fatalf("history split trigger_after_turns=%d want=3", got)
 	}
 }
+
+func TestStoreHistorySplitLegacyDisabledConfigNormalizesToEnabled(t *testing.T) {
+	t.Setenv("DS2API_CONFIG_JSON", `{"keys":["k1"],"history_split":{"enabled":false,"trigger_after_turns":2}}`)
+	store := LoadStore()
+	if !store.HistorySplitEnabled() {
+		t.Fatal("expected history split enabled when legacy config disables it")
+	}
+	snap := store.Snapshot()
+	if snap.HistorySplit.Enabled == nil || !*snap.HistorySplit.Enabled {
+		t.Fatalf("expected normalized history_split.enabled=true, got %#v", snap.HistorySplit.Enabled)
+	}
+	if got := store.HistorySplitTriggerAfterTurns(); got != 2 {
+		t.Fatalf("history split trigger_after_turns=%d want=2", got)
+	}
+}

internal/deepseek/client/client_auth.go

@@ -1,7 +1,8 @@
-package deepseek
+package client
 
 import (
 	"context"
+	dsprotocol "ds2api/internal/deepseek/protocol"
 	"errors"
 	"fmt"
 	"net/http"
@@ -28,7 +29,7 @@ func (c *Client) Login(ctx context.Context, acc config.Account) (string, error)
 	} else {
 		return "", errors.New("missing email/mobile")
 	}
-	resp, err := c.postJSON(ctx, clients.regular, clients.fallback, DeepSeekLoginURL, BaseHeaders, payload)
+	resp, err := c.postJSON(ctx, clients.regular, clients.fallback, dsprotocol.DeepSeekLoginURL, dsprotocol.BaseHeaders, payload)
 	if err != nil {
 		return "", err
 	}
@@ -58,7 +59,7 @@ func (c *Client) CreateSession(ctx context.Context, a *auth.RequestAuth, maxAtte
 	refreshed := false
 	for attempts < maxAttempts {
 		headers := c.authHeaders(a.DeepSeekToken)
-		resp, status, err := c.postJSONWithStatus(ctx, clients.regular, clients.fallback, DeepSeekCreateSessionURL, headers, map[string]any{"agent": "chat"})
+		resp, status, err := c.postJSONWithStatus(ctx, clients.regular, clients.fallback, dsprotocol.DeepSeekCreateSessionURL, headers, map[string]any{"agent": "chat"})
 		if err != nil {
 			config.Logger.Warn("[create_session] request error", "error", err, "account", a.AccountID)
 			attempts++
@@ -91,7 +92,7 @@ func (c *Client) CreateSession(ctx context.Context, a *auth.RequestAuth, maxAtte
 }
 
 func (c *Client) GetPow(ctx context.Context, a *auth.RequestAuth, maxAttempts int) (string, error) {
-	return c.GetPowForTarget(ctx, a, DeepSeekCompletionTargetPath, maxAttempts)
+	return c.GetPowForTarget(ctx, a, dsprotocol.DeepSeekCompletionTargetPath, maxAttempts)
 }
 
 func (c *Client) GetPowForTarget(ctx context.Context, a *auth.RequestAuth, targetPath string, maxAttempts int) (string, error) {
@@ -100,16 +101,20 @@ func (c *Client) GetPowForTarget(ctx context.Context, a *auth.RequestAuth, targe
 	}
 	targetPath = strings.TrimSpace(targetPath)
 	if targetPath == "" {
-		targetPath = DeepSeekCompletionTargetPath
+		targetPath = dsprotocol.DeepSeekCompletionTargetPath
 	}
 	clients := c.requestClientsForAuth(ctx, a)
 	attempts := 0
 	refreshed := false
+	lastFailureKind := FailureUnknown
+	lastFailureMessage := ""
 	for attempts < maxAttempts {
 		headers := c.authHeaders(a.DeepSeekToken)
-		resp, status, err := c.postJSONWithStatus(ctx, clients.regular, clients.fallback, DeepSeekCreatePowURL, headers, map[string]any{"target_path": targetPath})
+		resp, status, err := c.postJSONWithStatus(ctx, clients.regular, clients.fallback, dsprotocol.DeepSeekCreatePowURL, headers, map[string]any{"target_path": targetPath})
 		if err != nil {
 			config.Logger.Warn("[get_pow] request error", "error", err, "account", a.AccountID, "target_path", targetPath)
+			lastFailureKind = FailureUnknown
+			lastFailureMessage = err.Error()
 			attempts++
 			continue
 		}
@@ -126,6 +131,12 @@ func (c *Client) GetPowForTarget(ctx context.Context, a *auth.RequestAuth, targe
 			return BuildPowHeader(challenge, answer)
 		}
 		config.Logger.Warn("[get_pow] failed", "status", status, "code", code, "biz_code", bizCode, "msg", msg, "biz_msg", bizMsg, "use_config_token", a.UseConfigToken, "account", a.AccountID, "target_path", targetPath)
+		lastFailureMessage = failureMessage(msg, bizMsg, "get pow failed")
+		if isTokenInvalid(status, code, bizCode, msg, bizMsg) || isAuthIndicativeBizFailure(msg, bizMsg) {
+			lastFailureKind = authFailureKind(a.UseConfigToken)
+		} else {
+			lastFailureKind = FailureUnknown
+		}
 		if a.UseConfigToken {
 			if !refreshed && shouldAttemptRefresh(status, code, bizCode, msg, bizMsg) {
 				if c.Auth.RefreshToken(ctx, a) {
@@ -141,12 +152,15 @@ func (c *Client) GetPowForTarget(ctx context.Context, a *auth.RequestAuth, targe
 		}
 		attempts++
 	}
+	if lastFailureKind != FailureUnknown {
+		return "", &RequestFailure{Op: "get pow", Kind: lastFailureKind, Message: lastFailureMessage}
+	}
 	return "", errors.New("get pow failed")
 }
 
 func (c *Client) authHeaders(token string) map[string]string {
-	headers := make(map[string]string, len(BaseHeaders)+1)
-	for k, v := range BaseHeaders {
+	headers := make(map[string]string, len(dsprotocol.BaseHeaders)+1)
+	for k, v := range dsprotocol.BaseHeaders {
 		headers[k] = v
 	}
 	headers["authorization"] = "Bearer " + token
@@ -210,6 +224,23 @@ func isAuthIndicativeBizFailure(msg string, bizMsg string) bool {
 	return false
 }
 
+func authFailureKind(useConfigToken bool) FailureKind {
+	if useConfigToken {
+		return FailureManagedUnauthorized
+	}
+	return FailureDirectUnauthorized
+}
+
+func failureMessage(msg string, bizMsg string, fallback string) string {
+	if trimmed := strings.TrimSpace(bizMsg); trimmed != "" {
+		return trimmed
+	}
+	if trimmed := strings.TrimSpace(msg); trimmed != "" {
+		return trimmed
+	}
+	return strings.TrimSpace(fallback)
+}
+
 // DeepSeek has returned create-session ids in both biz_data.id and
 // biz_data.chat_session.id across observed response variants; accept either.
 func extractCreateSessionID(resp map[string]any) string {

internal/deepseek/client/client_auth_mobile_test.go

@@ -1,4 +1,4 @@
-package deepseek
+package client
 
 import "testing"
 

internal/deepseek/client/client_auth_refresh_test.go

@@ -1,4 +1,4 @@
-package deepseek
+package client
 
 import "testing"
 

internal/deepseek/client/client_auth_test.go

@@ -1,4 +1,4 @@
-package deepseek
+package client
 
 import "testing"
 

internal/deepseek/client/client_completion.go

@@ -1,8 +1,9 @@
-package deepseek
+package client
 
 import (
 	"bytes"
 	"context"
+	dsprotocol "ds2api/internal/deepseek/protocol"
 	"encoding/json"
 	"errors"
 	"net/http"
@@ -20,10 +21,10 @@ func (c *Client) CallCompletion(ctx context.Context, a *auth.RequestAuth, payloa
 	clients := c.requestClientsForAuth(ctx, a)
 	headers := c.authHeaders(a.DeepSeekToken)
 	headers["x-ds-pow-response"] = powResp
-	captureSession := c.capture.Start("deepseek_completion", DeepSeekCompletionURL, a.AccountID, payload)
+	captureSession := c.capture.Start("deepseek_completion", dsprotocol.DeepSeekCompletionURL, a.AccountID, payload)
 	attempts := 0
 	for attempts < maxAttempts {
-		resp, err := c.streamPost(ctx, clients.stream, DeepSeekCompletionURL, headers, payload)
+		resp, err := c.streamPost(ctx, clients.stream, dsprotocol.DeepSeekCompletionURL, headers, payload)
 		if err != nil {
 			attempts++
 			time.Sleep(time.Second)

internal/deepseek/client/client_continue.go

@@ -1,9 +1,10 @@
-package deepseek
+package client
 
 import (
 	"bufio"
 	"bytes"
 	"context"
+	dsprotocol "ds2api/internal/deepseek/protocol"
 	"encoding/json"
 	"errors"
 	"io"
@@ -60,8 +61,8 @@ func (c *Client) callContinue(ctx context.Context, a *auth.RequestAuth, sessionI
 		"fallback_to_resume": true,
 	}
 	config.Logger.Info("[auto_continue] calling continue", "session_id", sessionID, "message_id", responseMessageID)
-	captureSession := c.capture.Start("deepseek_continue", DeepSeekContinueURL, a.AccountID, payload)
-	resp, err := c.streamPost(ctx, clients.stream, DeepSeekContinueURL, headers, payload)
+	captureSession := c.capture.Start("deepseek_continue", dsprotocol.DeepSeekContinueURL, a.AccountID, payload)
+	resp, err := c.streamPost(ctx, clients.stream, dsprotocol.DeepSeekContinueURL, headers, payload)
 	if err != nil {
 		return nil, err
 	}

internal/deepseek/client/client_continue_test.go

@@ -1,8 +1,9 @@
-package deepseek
+package client
 
 import (
 	"bytes"
 	"context"
+	dsprotocol "ds2api/internal/deepseek/protocol"
 	"errors"
 	"io"
 	"net/http"
@@ -58,8 +59,8 @@ func TestCallContinuePropagatesPowHeaderToFallbackRequest(t *testing.T) {
 	if seenPow != "pow-response-abc" {
 		t.Fatalf("continue request pow header=%q want=%q", seenPow, "pow-response-abc")
 	}
-	if seenURL != DeepSeekContinueURL {
-		t.Fatalf("continue request url=%q want=%q", seenURL, DeepSeekContinueURL)
+	if seenURL != dsprotocol.DeepSeekContinueURL {
+		t.Fatalf("continue request url=%q want=%q", seenURL, dsprotocol.DeepSeekContinueURL)
 	}
 }
 
@@ -112,8 +113,8 @@ func TestCallCompletionAutoContinueThreadsPowHeader(t *testing.T) {
 	if seenPow != "pow-response-xyz" {
 		t.Fatalf("threaded continue pow header=%q want=%q", seenPow, "pow-response-xyz")
 	}
-	if seenContinueURL != DeepSeekContinueURL {
-		t.Fatalf("continue url=%q want=%q", seenContinueURL, DeepSeekContinueURL)
+	if seenContinueURL != dsprotocol.DeepSeekContinueURL {
+		t.Fatalf("continue url=%q want=%q", seenContinueURL, dsprotocol.DeepSeekContinueURL)
 	}
 	if !bytes.Contains(out, []byte(`"status":"WIP"`)) {
 		t.Fatalf("expected initial stream content in body, got=%s", string(out))

internal/deepseek/client/client_core.go

@@ -1,4 +1,4 @@
-package deepseek
+package client
 
 import (
 	"context"

internal/deepseek/client/client_file_status.go

@@ -1,7 +1,8 @@
-package deepseek
+package client
 
 import (
 	"context"
+	dsprotocol "ds2api/internal/deepseek/protocol"
 	"errors"
 	"fmt"
 	"net/http"
@@ -70,7 +71,7 @@ func (c *Client) fetchUploadedFile(ctx context.Context, a *auth.RequestAuth, fil
 		return nil, errors.New("file id is required")
 	}
 	clients := c.requestClientsForAuth(ctx, a)
-	reqURL := DeepSeekFetchFilesURL + "?file_ids=" + url.QueryEscape(fileID)
+	reqURL := dsprotocol.DeepSeekFetchFilesURL + "?file_ids=" + url.QueryEscape(fileID)
 	headers := c.authHeaders(a.DeepSeekToken)
 
 	resp, status, err := c.getJSONWithStatus(ctx, clients.regular, reqURL, headers)

internal/deepseek/client/client_http_helpers.go

@@ -1,7 +1,6 @@
-package deepseek
+package client
 
 import (
-	"bufio"
 	"compress/gzip"
 	"io"
 	"net/http"
@@ -41,17 +40,10 @@ func (c *Client) jsonHeaders(headers map[string]string) map[string]string {
 	return out
 }
 
-func ScanSSELines(resp *http.Response, onLine func([]byte) bool) error {
-	scanner := bufio.NewScanner(resp.Body)
-	buf := make([]byte, 0, 64*1024)
-	scanner.Buffer(buf, 2*1024*1024)
-	for scanner.Scan() {
-		if !onLine(scanner.Bytes()) {
-			break
-		}
+func cloneStringMap(in map[string]string) map[string]string {
+	out := make(map[string]string, len(in))
+	for k, v := range in {
+		out[k] = v
 	}
-	if err := scanner.Err(); err != nil {
-		return err
-	}
-	return nil
+	return out
 }

internal/deepseek/client/client_http_json.go

@@ -1,4 +1,4 @@
-package deepseek
+package client
 
 import (
 	"bytes"

internal/deepseek/client/client_http_json_test.go

@@ -1,4 +1,4 @@
-package deepseek
+package client
 
 import (
 	"context"

internal/deepseek/client/client_session.go

@@ -1,7 +1,8 @@
-package deepseek
+package client
 
 import (
 	"context"
+	dsprotocol "ds2api/internal/deepseek/protocol"
 	"errors"
 	"fmt"
 	"net/http"
@@ -49,7 +50,7 @@ func (c *Client) GetSessionCount(ctx context.Context, a *auth.RequestAuth, maxAt
 		headers := c.authHeaders(a.DeepSeekToken)
 
 		// 构建请求 URL
-		reqURL := DeepSeekFetchSessionURL + "?lte_cursor.pinned=false"
+		reqURL := dsprotocol.DeepSeekFetchSessionURL + "?lte_cursor.pinned=false"
 
 		resp, status, err := c.getJSONWithStatus(ctx, clients.regular, reqURL, headers)
 		if err != nil {
@@ -109,7 +110,7 @@ func (c *Client) GetSessionCount(ctx context.Context, a *auth.RequestAuth, maxAt
 func (c *Client) GetSessionCountForToken(ctx context.Context, token string) (*SessionStats, error) {
 	clients := c.requestClientsFromContext(ctx)
 	headers := c.authHeaders(token)
-	reqURL := DeepSeekFetchSessionURL + "?lte_cursor.pinned=false"
+	reqURL := dsprotocol.DeepSeekFetchSessionURL + "?lte_cursor.pinned=false"
 
 	resp, status, err := c.getJSONWithStatus(ctx, clients.regular, reqURL, headers)
 	if err != nil {
@@ -202,7 +203,7 @@ func (c *Client) FetchSessionPage(ctx context.Context, a *auth.RequestAuth, curs
 	if cursor != "" {
 		params.Set("lte_cursor", cursor)
 	}
-	reqURL := DeepSeekFetchSessionURL + "?" + params.Encode()
+	reqURL := dsprotocol.DeepSeekFetchSessionURL + "?" + params.Encode()
 
 	resp, status, err := c.getJSONWithStatus(ctx, clients.regular, reqURL, headers)
 	if err != nil {

internal/deepseek/client/client_session_delete.go

@@ -1,7 +1,8 @@
-package deepseek
+package client
 
 import (
 	"context"
+	dsprotocol "ds2api/internal/deepseek/protocol"
 	"errors"
 	"fmt"
 	"net/http"
@@ -43,7 +44,7 @@ func (c *Client) DeleteSession(ctx context.Context, a *auth.RequestAuth, session
 			"chat_session_id": sessionID,
 		}
 
-		resp, status, err := c.postJSONWithStatus(ctx, clients.regular, clients.fallback, DeepSeekDeleteSessionURL, headers, payload)
+		resp, status, err := c.postJSONWithStatus(ctx, clients.regular, clients.fallback, dsprotocol.DeepSeekDeleteSessionURL, headers, payload)
 		if err != nil {
 			config.Logger.Warn("[delete_session] request error", "error", err, "session_id", sessionID)
 			attempts++
@@ -97,7 +98,7 @@ func (c *Client) DeleteSessionForToken(ctx context.Context, token string, sessio
 		"chat_session_id": sessionID,
 	}
 
-	resp, status, err := c.postJSONWithStatus(ctx, clients.regular, clients.fallback, DeepSeekDeleteSessionURL, headers, payload)
+	resp, status, err := c.postJSONWithStatus(ctx, clients.regular, clients.fallback, dsprotocol.DeepSeekDeleteSessionURL, headers, payload)
 	if err != nil {
 		result.ErrorMessage = err.Error()
 		return result, err
@@ -120,7 +121,7 @@ func (c *Client) DeleteAllSessions(ctx context.Context, a *auth.RequestAuth) err
 	headers := c.authHeaders(a.DeepSeekToken)
 	payload := map[string]any{}
 
-	resp, status, err := c.postJSONWithStatus(ctx, clients.regular, clients.fallback, DeepSeekDeleteAllSessionsURL, headers, payload)
+	resp, status, err := c.postJSONWithStatus(ctx, clients.regular, clients.fallback, dsprotocol.DeepSeekDeleteAllSessionsURL, headers, payload)
 	if err != nil {
 		config.Logger.Warn("[delete_all_sessions] request error", "error", err)
 		return err
@@ -142,7 +143,7 @@ func (c *Client) DeleteAllSessionsForToken(ctx context.Context, token string) er
 	headers := c.authHeaders(token)
 	payload := map[string]any{}
 
-	resp, status, err := c.postJSONWithStatus(ctx, clients.regular, clients.fallback, DeepSeekDeleteAllSessionsURL, headers, payload)
+	resp, status, err := c.postJSONWithStatus(ctx, clients.regular, clients.fallback, dsprotocol.DeepSeekDeleteAllSessionsURL, headers, payload)
 	if err != nil {
 		config.Logger.Warn("[delete_all_sessions_for_token] request error", "error", err)
 		return err

internal/deepseek/client/client_upload.go

@@ -1,8 +1,9 @@
-package deepseek
+package client
 
 import (
 	"bytes"
 	"context"
+	dsprotocol "ds2api/internal/deepseek/protocol"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -63,14 +64,16 @@ func (c *Client) UploadFile(ctx context.Context, a *auth.RequestAuth, req Upload
 		"purpose":      purpose,
 		"bytes":        len(req.Data),
 	}
-	captureSession := c.capture.Start("deepseek_upload_file", DeepSeekUploadFileURL, a.AccountID, capturePayload)
+	captureSession := c.capture.Start("deepseek_upload_file", dsprotocol.DeepSeekUploadFileURL, a.AccountID, capturePayload)
 	attempts := 0
 	refreshed := false
 	powHeader := ""
+	lastFailureKind := FailureUnknown
+	lastFailureMessage := ""
 	for attempts < maxAttempts {
 		clients := c.requestClientsForAuth(ctx, a)
 		if strings.TrimSpace(powHeader) == "" {
-			powHeader, err = c.GetPowForTarget(ctx, a, DeepSeekUploadTargetPath, maxAttempts)
+			powHeader, err = c.GetPowForTarget(ctx, a, dsprotocol.DeepSeekUploadTargetPath, maxAttempts)
 			if err != nil {
 				return nil, err
 			}
@@ -81,10 +84,12 @@ func (c *Client) UploadFile(ctx context.Context, a *auth.RequestAuth, req Upload
 		headers["x-ds-pow-response"] = powHeader
 		headers["x-file-size"] = strconv.Itoa(len(req.Data))
 		headers["x-thinking-enabled"] = "1"
-		resp, err := c.doUpload(ctx, clients.regular, clients.fallback, DeepSeekUploadFileURL, headers, body)
+		resp, err := c.doUpload(ctx, clients.regular, clients.fallback, dsprotocol.DeepSeekUploadFileURL, headers, body)
 		if err != nil {
 			config.Logger.Warn("[upload_file] request error", "error", err, "account", a.AccountID, "filename", filename)
 			powHeader = ""
+			lastFailureKind = FailureUnknown
+			lastFailureMessage = err.Error()
 			attempts++
 			continue
 		}
@@ -131,6 +136,12 @@ func (c *Client) UploadFile(ctx context.Context, a *auth.RequestAuth, req Upload
 		}
 		config.Logger.Warn("[upload_file] failed", "status", resp.StatusCode, "code", code, "biz_code", bizCode, "msg", msg, "biz_msg", bizMsg, "account", a.AccountID, "filename", filename)
 		powHeader = ""
+		lastFailureMessage = failureMessage(msg, bizMsg, "upload file failed")
+		if isTokenInvalid(resp.StatusCode, code, bizCode, msg, bizMsg) || isAuthIndicativeBizFailure(msg, bizMsg) {
+			lastFailureKind = authFailureKind(a.UseConfigToken)
+		} else {
+			lastFailureKind = FailureUnknown
+		}
 		if a.UseConfigToken {
 			if !refreshed && shouldAttemptRefresh(resp.StatusCode, code, bizCode, msg, bizMsg) {
 				if c.Auth.RefreshToken(ctx, a) {
@@ -147,6 +158,9 @@ func (c *Client) UploadFile(ctx context.Context, a *auth.RequestAuth, req Upload
 		}
 		attempts++
 	}
+	if lastFailureKind != FailureUnknown {
+		return nil, &RequestFailure{Op: "upload file", Kind: lastFailureKind, Message: lastFailureMessage}
+	}
 	return nil, errors.New("upload file failed")
 }
 

internal/deepseek/client/client_upload_test.go

@@ -1,7 +1,8 @@
-package deepseek
+package client
 
 import (
 	"context"
+	dsprotocol "ds2api/internal/deepseek/protocol"
 	"encoding/base64"
 	"encoding/hex"
 	"encoding/json"
@@ -75,7 +76,7 @@ func TestExtractUploadFileResultSupportsNestedShapes(t *testing.T) {
 
 func TestUploadFileUsesUploadTargetPowAndMultipartHeaders(t *testing.T) {
 	challengeHash := powpkg.DeepSeekHashV1([]byte(powpkg.BuildPrefix("salt", 1712345678) + "42"))
-	powResponse := `{"code":0,"msg":"ok","data":{"biz_code":0,"biz_data":{"challenge":{"algorithm":"DeepSeekHashV1","challenge":"` + hex.EncodeToString(challengeHash[:]) + `","salt":"salt","expire_at":1712345678,"difficulty":1000,"signature":"sig","target_path":"` + DeepSeekUploadTargetPath + `"}}}}`
+	powResponse := `{"code":0,"msg":"ok","data":{"biz_code":0,"biz_data":{"challenge":{"algorithm":"DeepSeekHashV1","challenge":"` + hex.EncodeToString(challengeHash[:]) + `","salt":"salt","expire_at":1712345678,"difficulty":1000,"signature":"sig","target_path":"` + dsprotocol.DeepSeekUploadTargetPath + `"}}}}`
 	uploadResponse := `{"code":0,"msg":"ok","data":{"biz_code":0,"biz_data":{"file":{"file_id":"file_789","filename":"demo.txt","bytes":5,"status":"processed","purpose":"assistants","is_image":false}}}}`
 	var seenPow string
 	var seenTargetPath string
@@ -119,7 +120,7 @@ func TestUploadFileUsesUploadTargetPowAndMultipartHeaders(t *testing.T) {
 	if result.ID != "file_789" {
 		t.Fatalf("expected uploaded file id file_789, got %#v", result)
 	}
-	if !strings.Contains(seenTargetPath, `"target_path":"`+DeepSeekUploadTargetPath+`"`) {
+	if !strings.Contains(seenTargetPath, `"target_path":"`+dsprotocol.DeepSeekUploadTargetPath+`"`) {
 		t.Fatalf("expected upload target_path in pow request, got %q", seenTargetPath)
 	}
 	if strings.TrimSpace(seenPow) == "" {
@@ -133,8 +134,8 @@ func TestUploadFileUsesUploadTargetPowAndMultipartHeaders(t *testing.T) {
 	if err := json.Unmarshal(rawPow, &powHeader); err != nil {
 		t.Fatalf("unmarshal pow header failed: %v", err)
 	}
-	if powHeader["target_path"] != DeepSeekUploadTargetPath {
-		t.Fatalf("expected pow target_path %q, got %#v", DeepSeekUploadTargetPath, powHeader["target_path"])
+	if powHeader["target_path"] != dsprotocol.DeepSeekUploadTargetPath {
+		t.Fatalf("expected pow target_path %q, got %#v", dsprotocol.DeepSeekUploadTargetPath, powHeader["target_path"])
 	}
 	if seenFileSize != "5" {
 		t.Fatalf("expected x-file-size=5, got %q", seenFileSize)
@@ -153,7 +154,7 @@ func TestUploadFileWaitsForProcessedFetchFiles(t *testing.T) {
 	defer func() { fileReadySleep = oldSleep }()
 
 	challengeHash := powpkg.DeepSeekHashV1([]byte(powpkg.BuildPrefix("salt", 1712345678) + "42"))
-	powResponse := `{"code":0,"msg":"ok","data":{"biz_code":0,"biz_data":{"challenge":{"algorithm":"DeepSeekHashV1","challenge":"` + hex.EncodeToString(challengeHash[:]) + `","salt":"salt","expire_at":1712345678,"difficulty":1000,"signature":"sig","target_path":"` + DeepSeekUploadTargetPath + `"}}}}`
+	powResponse := `{"code":0,"msg":"ok","data":{"biz_code":0,"biz_data":{"challenge":{"algorithm":"DeepSeekHashV1","challenge":"` + hex.EncodeToString(challengeHash[:]) + `","salt":"salt","expire_at":1712345678,"difficulty":1000,"signature":"sig","target_path":"` + dsprotocol.DeepSeekUploadTargetPath + `"}}}}`
 	uploadResponse := `{"code":0,"msg":"ok","data":{"biz_code":0,"biz_data":{"file":{"file_id":"file_789","filename":"demo.txt","bytes":5,"status":"PENDING","purpose":"assistants","is_image":false}}}}`
 	pendingFetchResponse := `{"code":0,"msg":"ok","data":{"biz_code":0,"biz_data":{"files":[{"file_id":"file_789","filename":"demo.txt","bytes":5,"status":"PENDING","purpose":"assistants","is_image":false}]}}}`
 	processedFetchResponse := `{"code":0,"msg":"ok","data":{"biz_code":0,"biz_data":{"files":[{"file_id":"file_789","filename":"demo.txt","bytes":5,"status":"processed","purpose":"assistants","is_image":true}]}}}`
@@ -165,7 +166,7 @@ func TestUploadFileWaitsForProcessedFetchFiles(t *testing.T) {
 			switch call {
 			case 1:
 				bodyBytes, _ := io.ReadAll(req.Body)
-				if !strings.Contains(string(bodyBytes), `"target_path":"`+DeepSeekUploadTargetPath+`"`) {
+				if !strings.Contains(string(bodyBytes), `"target_path":"`+dsprotocol.DeepSeekUploadTargetPath+`"`) {
 					t.Fatalf("expected pow target path request, got %s", string(bodyBytes))
 				}
 				return &http.Response{StatusCode: http.StatusOK, Header: make(http.Header), Body: io.NopCloser(strings.NewReader(powResponse)), Request: req}, nil

internal/deepseek/client/deepseek_edge_test.go

@@ -1,4 +1,4 @@
-package deepseek
+package client
 
 import (
 	"context"

internal/deepseek/client/errors.go

@@ -0,0 +1,46 @@
+package client
+
+import (
+	"errors"
+	"fmt"
+)
+
+type FailureKind string
+
+const (
+	FailureUnknown             FailureKind = ""
+	FailureDirectUnauthorized  FailureKind = "direct_unauthorized"
+	FailureManagedUnauthorized FailureKind = "managed_unauthorized"
+)
+
+type RequestFailure struct {
+	Op      string
+	Kind    FailureKind
+	Message string
+}
+
+func (e *RequestFailure) Error() string {
+	if e == nil {
+		return ""
+	}
+	switch {
+	case e.Op != "" && e.Message != "":
+		return fmt.Sprintf("%s: %s", e.Op, e.Message)
+	case e.Op != "":
+		return e.Op + " failed"
+	case e.Message != "":
+		return e.Message
+	default:
+		return "request failed"
+	}
+}
+
+func IsManagedUnauthorizedError(err error) bool {
+	var failure *RequestFailure
+	return errors.As(err, &failure) && failure.Kind == FailureManagedUnauthorized
+}
+
+func IsDirectUnauthorizedError(err error) bool {
+	var failure *RequestFailure
+	return errors.As(err, &failure) && failure.Kind == FailureDirectUnauthorized
+}

internal/deepseek/client/pow.go

@@ -1,4 +1,4 @@
-package deepseek
+package client
 
 import (
 	"context"

internal/deepseek/client/pow_test.go

@@ -1,4 +1,4 @@
-package deepseek
+package client
 
 import (
 	"context"

internal/deepseek/client/proxy.go

@@ -1,7 +1,8 @@
-package deepseek
+package client
 
 import (
 	"context"
+	dsprotocol "ds2api/internal/deepseek/protocol"
 	"fmt"
 	"net"
 	"net/http"
@@ -172,7 +173,7 @@ func applyProxyConnectivityHeaders(req *http.Request) {
 	if req == nil {
 		return
 	}
-	for key, value := range BaseHeaders {
+	for key, value := range dsprotocol.BaseHeaders {
 		key = strings.TrimSpace(key)
 		value = strings.TrimSpace(value)
 		if key == "" || value == "" {

internal/deepseek/client/proxy_test.go

@@ -1,7 +1,8 @@
-package deepseek
+package client
 
 import (
 	"context"
+	dsprotocol "ds2api/internal/deepseek/protocol"
 	"net/http"
 	"strings"
 	"testing"
@@ -52,7 +53,7 @@ func TestApplyProxyConnectivityHeadersUsesBaseHeaders(t *testing.T) {
 
 	applyProxyConnectivityHeaders(req)
 
-	for key, want := range BaseHeaders {
+	for key, want := range dsprotocol.BaseHeaders {
 		if got := req.Header.Get(key); got != want {
 			t.Fatalf("expected header %q=%q, got %q", key, want, got)
 		}

internal/deepseek/prompt.go

@@ -1,11 +0,0 @@
-package deepseek
-
-import "ds2api/internal/prompt"
-
-func MessagesPrepare(messages []map[string]any) string {
-	return prompt.MessagesPrepare(messages)
-}
-
-func MessagesPrepareWithThinking(messages []map[string]any, thinkingEnabled bool) string {
-	return prompt.MessagesPrepareWithThinking(messages, thinkingEnabled)
-}

internal/deepseek/protocol/constants.go

@@ -1,4 +1,4 @@
-package deepseek
+package protocol
 
 import (
 	_ "embed"

internal/deepseek/protocol/constants_test.go

@@ -1,4 +1,4 @@
-package deepseek
+package protocol
 
 import "testing"
 

internal/deepseek/protocol/sse.go

@@ -0,0 +1,21 @@
+package protocol
+
+import (
+	"bufio"
+	"net/http"
+)
+
+func ScanSSELines(resp *http.Response, onLine func([]byte) bool) error {
+	scanner := bufio.NewScanner(resp.Body)
+	buf := make([]byte, 0, 64*1024)
+	scanner.Buffer(buf, 2*1024*1024)
+	for scanner.Scan() {
+		if !onLine(scanner.Bytes()) {
+			break
+		}
+	}
+	if err := scanner.Err(); err != nil {
+		return err
+	}
+	return nil
+}

internal/format/openai/render_stream_events.go

@@ -117,7 +117,7 @@ func BuildResponsesFunctionCallArgumentsDonePayload(responseID, itemID string, o
 	}
 }
 
-func BuildResponsesFailedPayload(responseID, model, message, code string) map[string]any {
+func BuildResponsesFailedPayload(responseID, model string, status int, message, code string) map[string]any {
 	code = strings.TrimSpace(code)
 	if code == "" {
 		code = "api_error"
@@ -129,15 +129,36 @@ func BuildResponsesFailedPayload(responseID, model, message, code string) map[st
 		"object":      "response",
 		"model":       model,
 		"status":      "failed",
+		"status_code": status,
 		"error": map[string]any{
 			"message": message,
-			"type":    "invalid_request_error",
+			"type":    responsesErrorType(status),
 			"code":    code,
 			"param":   nil,
 		},
 	}
 }
 
+func responsesErrorType(status int) string {
+	switch status {
+	case 400, 404, 422:
+		return "invalid_request_error"
+	case 401:
+		return "authentication_error"
+	case 403:
+		return "permission_error"
+	case 429:
+		return "rate_limit_error"
+	case 503:
+		return "service_unavailable_error"
+	default:
+		if status >= 500 {
+			return "api_error"
+		}
+		return "invalid_request_error"
+	}
+}
+
 func BuildResponsesCompletedPayload(response map[string]any) map[string]any {
 	responseID, _ := response["id"].(string)
 	return map[string]any{

internal/httpapi/admin/accounts/deps.go

@@ -0,0 +1,46 @@
+package accounts
+
+import (
+	"net/http"
+
+	"ds2api/internal/chathistory"
+	"ds2api/internal/config"
+	adminshared "ds2api/internal/httpapi/admin/shared"
+)
+
+type Handler struct {
+	Store       adminshared.ConfigStore
+	Pool        adminshared.PoolController
+	DS          adminshared.DeepSeekCaller
+	OpenAI      adminshared.OpenAIChatCaller
+	ChatHistory *chathistory.Store
+}
+
+var writeJSON = adminshared.WriteJSON
+
+func reverseAccounts(a []config.Account) { adminshared.ReverseAccounts(a) }
+func intFromQuery(r *http.Request, key string, d int) int {
+	return adminshared.IntFromQuery(r, key, d)
+}
+func maskSecretPreview(secret string) string {
+	return adminshared.MaskSecretPreview(secret)
+}
+func toAccount(m map[string]any) config.Account {
+	return adminshared.ToAccount(m)
+}
+func fieldStringOptional(m map[string]any, key string) (string, bool) {
+	return adminshared.FieldStringOptional(m, key)
+}
+func accountMatchesIdentifier(acc config.Account, identifier string) bool {
+	return adminshared.AccountMatchesIdentifier(acc, identifier)
+}
+func findProxyByID(c config.Config, proxyID string) (config.Proxy, bool) {
+	return adminshared.FindProxyByID(c, proxyID)
+}
+func findAccountByIdentifier(store adminshared.ConfigStore, identifier string) (config.Account, bool) {
+	return adminshared.FindAccountByIdentifier(store, identifier)
+}
+func newRequestError(detail string) error { return adminshared.NewRequestError(detail) }
+func requestErrorDetail(err error) (string, bool) {
+	return adminshared.RequestErrorDetail(err)
+}

internal/httpapi/admin/accounts/handler_accounts_crud.go

@@ -1,4 +1,4 @@
-package admin
+package accounts
 
 import (
 	"encoding/json"
@@ -58,14 +58,6 @@ func (h *Handler) listAccounts(w http.ResponseWriter, r *http.Request) {
 	for _, acc := range accounts[start:end] {
 		testStatus, _ := h.Store.AccountTestStatus(acc.Identifier())
 		token := strings.TrimSpace(acc.Token)
-		preview := ""
-		if token != "" {
-			if len(token) > 20 {
-				preview = token[:20] + "..."
-			} else {
-				preview = token
-			}
-		}
 		items = append(items, map[string]any{
 			"identifier":    acc.Identifier(),
 			"name":          acc.Name,
@@ -75,7 +67,7 @@ func (h *Handler) listAccounts(w http.ResponseWriter, r *http.Request) {
 			"proxy_id":      acc.ProxyID,
 			"has_password":  acc.Password != "",
 			"has_token":     token != "",
-			"token_preview": preview,
+			"token_preview": maskSecretPreview(token),
 			"test_status":   testStatus,
 		})
 	}

internal/httpapi/admin/accounts/handler_accounts_crud_test.go

@@ -1,4 +1,4 @@
-package admin
+package accounts
 
 import (
 	"encoding/json"
@@ -86,3 +86,33 @@ func TestUpdateAccountMetadataPreservesCredentials(t *testing.T) {
 		t.Fatalf("password should be preserved, got %#v", acc)
 	}
 }
+
+func TestListAccountsMasksTokenPreview(t *testing.T) {
+	h := newAdminTestHandler(t, `{
+		"accounts":[{"email":"u@example.com","password":"pwd"}]
+	}`)
+	if err := h.Store.UpdateAccountToken("u@example.com", "abcdefgh"); err != nil {
+		t.Fatalf("seed runtime token: %v", err)
+	}
+
+	req := httptest.NewRequest(http.MethodGet, "/admin/accounts?page=1&page_size=10", nil)
+	rec := httptest.NewRecorder()
+	h.listAccounts(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("unexpected status: %d body=%s", rec.Code, rec.Body.String())
+	}
+
+	var payload map[string]any
+	if err := json.Unmarshal(rec.Body.Bytes(), &payload); err != nil {
+		t.Fatalf("decode response failed: %v", err)
+	}
+	items, _ := payload["items"].([]any)
+	if len(items) != 1 {
+		t.Fatalf("expected 1 item, got %d", len(items))
+	}
+	first, _ := items[0].(map[string]any)
+	if got, _ := first["token_preview"].(string); got != "ab****gh" {
+		t.Fatalf("expected masked token preview, got %q", got)
+	}
+}

internal/httpapi/admin/accounts/handler_accounts_identifier_test.go

@@ -1,4 +1,4 @@
-package admin
+package accounts
 
 import (
 	"bytes"

internal/httpapi/admin/accounts/handler_accounts_queue.go

@@ -1,4 +1,4 @@
-package admin
+package accounts
 
 import "net/http"
 

internal/httpapi/admin/accounts/handler_accounts_testing.go

@@ -1,4 +1,4 @@
-package admin
+package accounts
 
 import (
 	"bytes"
@@ -13,9 +13,9 @@ import (
 
 	authn "ds2api/internal/auth"
 	"ds2api/internal/config"
-	"ds2api/internal/deepseek"
+	"ds2api/internal/prompt"
+	"ds2api/internal/promptcompat"
 	"ds2api/internal/sse"
-	"ds2api/internal/util"
 )
 
 type modelAliasSnapshotReader struct {
@@ -41,7 +41,7 @@ func (h *Handler) testSingleAccount(w http.ResponseWriter, r *http.Request) {
 	}
 	model, _ := req["model"].(string)
 	if model == "" {
-		model = "deepseek-chat"
+		model = "deepseek-v4-flash"
 	}
 	message, _ := req["message"].(string)
 	result := h.testAccount(r.Context(), acc, model, message)
@@ -53,7 +53,7 @@ func (h *Handler) testAllAccounts(w http.ResponseWriter, r *http.Request) {
 	_ = json.NewDecoder(r.Body).Decode(&req)
 	model, _ := req["model"].(string)
 	if model == "" {
-		model = "deepseek-chat"
+		model = "deepseek-v4-flash"
 	}
 	accounts := h.Store.Snapshot().Accounts
 	if len(accounts) == 0 {
@@ -174,9 +174,9 @@ func (h *Handler) testAccount(ctx context.Context, acc config.Account, model, me
 		result["message"] = "获取 PoW 失败: " + err.Error()
 		return result
 	}
-	payload := util.StandardRequest{
+	payload := promptcompat.StandardRequest{
 		ResolvedModel: model,
-		FinalPrompt:   deepseek.MessagesPrepare([]map[string]any{{"role": "user", "content": message}}),
+		FinalPrompt:   prompt.MessagesPrepare([]map[string]any{{"role": "user", "content": message}}),
 		Thinking:      thinking,
 		Search:        search,
 	}.CompletionPayload(sessionID)
@@ -211,7 +211,7 @@ func (h *Handler) testAPI(w http.ResponseWriter, r *http.Request) {
 	message, _ := req["message"].(string)
 	apiKey, _ := req["api_key"].(string)
 	if model == "" {
-		model = "deepseek-chat"
+		model = "deepseek-v4-flash"
 	}
 	if message == "" {
 		message = "你好"

internal/httpapi/admin/accounts/handler_accounts_testing_test.go

@@ -1,4 +1,4 @@
-package admin
+package accounts
 
 import (
 	"bytes"
@@ -13,7 +13,7 @@ import (
 
 	"ds2api/internal/auth"
 	"ds2api/internal/config"
-	"ds2api/internal/deepseek"
+	dsclient "ds2api/internal/deepseek/client"
 )
 
 type testingDSMock struct {
@@ -58,8 +58,8 @@ func (m *testingDSMock) DeleteAllSessionsForToken(_ context.Context, _ string) e
 	return nil
 }
 
-func (m *testingDSMock) GetSessionCountForToken(_ context.Context, _ string) (*deepseek.SessionStats, error) {
-	return &deepseek.SessionStats{Success: true}, nil
+func (m *testingDSMock) GetSessionCountForToken(_ context.Context, _ string) (*dsclient.SessionStats, error) {
+	return &dsclient.SessionStats{Success: true}, nil
 }
 
 func TestTestAccount_BatchModeOnlyCreatesSession(t *testing.T) {
@@ -72,7 +72,7 @@ func TestTestAccount_BatchModeOnlyCreatesSession(t *testing.T) {
 		t.Fatal("expected test account")
 	}
 
-	result := h.testAccount(context.Background(), acc, "deepseek-chat", "")
+	result := h.testAccount(context.Background(), acc, "deepseek-v4-flash", "")
 
 	if ok, _ := result["success"].(bool); !ok {
 		t.Fatalf("expected success=true, got %#v", result)
@@ -163,8 +163,8 @@ func (m *completionPayloadDSMock) DeleteAllSessionsForToken(_ context.Context, _
 	return nil
 }
 
-func (m *completionPayloadDSMock) GetSessionCountForToken(_ context.Context, _ string) (*deepseek.SessionStats, error) {
-	return &deepseek.SessionStats{Success: true}, nil
+func (m *completionPayloadDSMock) GetSessionCountForToken(_ context.Context, _ string) (*dsclient.SessionStats, error) {
+	return &dsclient.SessionStats{Success: true}, nil
 }
 
 func TestTestAccount_MessageModeUsesExpertModelTypeForExpertModel(t *testing.T) {
@@ -177,7 +177,7 @@ func TestTestAccount_MessageModeUsesExpertModelTypeForExpertModel(t *testing.T)
 		t.Fatal("expected test account")
 	}
 
-	result := h.testAccount(context.Background(), acc, "deepseek-expert-chat", "hello")
+	result := h.testAccount(context.Background(), acc, "deepseek-v4-pro", "hello")
 
 	if ok, _ := result["success"].(bool); !ok {
 		t.Fatalf("expected success=true, got %#v", result)
@@ -200,7 +200,7 @@ func TestTestAccount_MessageModeUsesVisionModelTypeForVisionModel(t *testing.T)
 		t.Fatal("expected test account")
 	}
 
-	result := h.testAccount(context.Background(), acc, "deepseek-vision-chat", "hello")
+	result := h.testAccount(context.Background(), acc, "deepseek-v4-vision", "hello")
 
 	if ok, _ := result["success"].(bool); !ok {
 		t.Fatalf("expected success=true, got %#v", result)

internal/httpapi/admin/accounts/routes.go

@@ -0,0 +1,38 @@
+package accounts
+
+import (
+	"context"
+	"net/http"
+
+	"github.com/go-chi/chi/v5"
+
+	"ds2api/internal/config"
+)
+
+func RegisterRoutes(r chi.Router, h *Handler) {
+	r.Get("/accounts", h.listAccounts)
+	r.Post("/accounts", h.addAccount)
+	r.Put("/accounts/{identifier}", h.updateAccount)
+	r.Delete("/accounts/{identifier}", h.deleteAccount)
+	r.Get("/queue/status", h.queueStatus)
+	r.Post("/accounts/test", h.testSingleAccount)
+	r.Post("/accounts/test-all", h.testAllAccounts)
+	r.Post("/accounts/sessions/delete-all", h.deleteAllSessions)
+	r.Post("/test", h.testAPI)
+}
+
+func RunAccountTestsConcurrently(accounts []config.Account, maxConcurrency int, testFn func(int, config.Account) map[string]any) []map[string]any {
+	return runAccountTestsConcurrently(accounts, maxConcurrency, testFn)
+}
+
+func (h *Handler) TestAccount(ctx context.Context, acc config.Account, model, message string) map[string]any {
+	return h.testAccount(ctx, acc, model, message)
+}
+
+func (h *Handler) ListAccounts(w http.ResponseWriter, r *http.Request)  { h.listAccounts(w, r) }
+func (h *Handler) AddAccount(w http.ResponseWriter, r *http.Request)    { h.addAccount(w, r) }
+func (h *Handler) UpdateAccount(w http.ResponseWriter, r *http.Request) { h.updateAccount(w, r) }
+func (h *Handler) DeleteAccount(w http.ResponseWriter, r *http.Request) { h.deleteAccount(w, r) }
+func (h *Handler) DeleteAllSessions(w http.ResponseWriter, r *http.Request) {
+	h.deleteAllSessions(w, r)
+}

internal/httpapi/admin/accounts/test_http_helpers_test.go

@@ -0,0 +1,35 @@
+package accounts
+
+import (
+	"bytes"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"github.com/go-chi/chi/v5"
+
+	"ds2api/internal/account"
+	"ds2api/internal/config"
+	adminshared "ds2api/internal/httpapi/admin/shared"
+)
+
+func newHTTPAdminHarness(t *testing.T, rawConfig string, ds adminshared.DeepSeekCaller) http.Handler {
+	t.Helper()
+	t.Setenv("DS2API_CONFIG_JSON", rawConfig)
+	store := config.LoadStore()
+	h := &Handler{
+		Store: store,
+		Pool:  account.NewPool(store),
+		DS:    ds,
+	}
+	r := chi.NewRouter()
+	RegisterRoutes(r, h)
+	return r
+}
+
+func adminReq(method, path string, body []byte) *http.Request {
+	req := httptest.NewRequest(method, path, bytes.NewReader(body))
+	req.Header.Set("Authorization", "Bearer admin")
+	req.Header.Set("Content-Type", "application/json")
+	return req
+}

internal/httpapi/admin/auth/deps.go

@@ -0,0 +1,19 @@
+package auth
+
+import (
+	"ds2api/internal/chathistory"
+	adminshared "ds2api/internal/httpapi/admin/shared"
+)
+
+type Handler struct {
+	Store       adminshared.ConfigStore
+	Pool        adminshared.PoolController
+	DS          adminshared.DeepSeekCaller
+	OpenAI      adminshared.OpenAIChatCaller
+	ChatHistory *chathistory.Store
+}
+
+var writeJSON = adminshared.WriteJSON
+var intFrom = adminshared.IntFrom
+
+func nilIfEmpty(s string) any { return adminshared.NilIfEmpty(s) }

internal/httpapi/admin/auth/handler_auth.go

@@ -1,4 +1,4 @@
-package admin
+package auth
 
 import (
 	"encoding/json"

internal/httpapi/admin/auth/routes.go

@@ -0,0 +1,20 @@
+package auth
+
+import (
+	"net/http"
+
+	"github.com/go-chi/chi/v5"
+)
+
+func (h *Handler) RequireAdmin(next http.Handler) http.Handler {
+	return h.requireAdmin(next)
+}
+
+func RegisterPublicRoutes(r chi.Router, h *Handler) {
+	r.Post("/login", h.login)
+	r.Get("/verify", h.verify)
+}
+
+func RegisterProtectedRoutes(r chi.Router, h *Handler) {
+	r.Get("/vercel/config", h.getVercelConfig)
+}

internal/httpapi/admin/configmgmt/deps.go

@@ -0,0 +1,50 @@
+package configmgmt
+
+import (
+	"ds2api/internal/chathistory"
+	"ds2api/internal/config"
+	adminshared "ds2api/internal/httpapi/admin/shared"
+)
+
+type Handler struct {
+	Store       adminshared.ConfigStore
+	Pool        adminshared.PoolController
+	DS          adminshared.DeepSeekCaller
+	OpenAI      adminshared.OpenAIChatCaller
+	ChatHistory *chathistory.Store
+}
+
+var writeJSON = adminshared.WriteJSON
+
+func maskSecretPreview(secret string) string {
+	return adminshared.MaskSecretPreview(secret)
+}
+func toStringSlice(v any) ([]string, bool) { return adminshared.ToStringSlice(v) }
+func toAccount(m map[string]any) config.Account {
+	return adminshared.ToAccount(m)
+}
+func toAPIKeys(v any) ([]config.APIKey, bool) { return adminshared.ToAPIKeys(v) }
+func mergeAPIKeysPreferStructured(existing, incoming []config.APIKey) ([]config.APIKey, int) {
+	return adminshared.MergeAPIKeysPreferStructured(existing, incoming)
+}
+func fieldString(m map[string]any, key string) string {
+	return adminshared.FieldString(m, key)
+}
+func fieldStringOptional(m map[string]any, key string) (string, bool) {
+	return adminshared.FieldStringOptional(m, key)
+}
+func normalizeAccountForStorage(acc config.Account) config.Account {
+	return adminshared.NormalizeAccountForStorage(acc)
+}
+func accountDedupeKey(acc config.Account) string { return adminshared.AccountDedupeKey(acc) }
+func normalizeAndDedupeAccounts(accounts []config.Account) []config.Account {
+	return adminshared.NormalizeAndDedupeAccounts(accounts)
+}
+func newRequestError(detail string) error { return adminshared.NewRequestError(detail) }
+func requestErrorDetail(err error) (string, bool) {
+	return adminshared.RequestErrorDetail(err)
+}
+func normalizeSettingsConfig(c *config.Config) { adminshared.NormalizeSettingsConfig(c) }
+func validateSettingsConfig(c config.Config) error {
+	return adminshared.ValidateSettingsConfig(c)
+}

internal/httpapi/admin/configmgmt/handler_config_import.go

@@ -1,9 +1,7 @@
-package admin
+package configmgmt
 
 import (
-	"crypto/md5"
 	"encoding/json"
-	"fmt"
 	"net/http"
 	"strings"
 
@@ -82,23 +80,6 @@ func (h *Handler) configImport(w http.ResponseWriter, r *http.Request) {
 				importedAccounts++
 			}
 
-			if len(incoming.ClaudeMapping) > 0 {
-				if next.ClaudeMapping == nil {
-					next.ClaudeMapping = map[string]string{}
-				}
-				for k, v := range incoming.ClaudeMapping {
-					next.ClaudeMapping[k] = v
-				}
-			}
-			if len(incoming.ClaudeModelMap) > 0 {
-				if next.ClaudeModelMap == nil {
-					next.ClaudeModelMap = map[string]string{}
-				}
-				for k, v := range incoming.ClaudeModelMap {
-					next.ClaudeModelMap[k] = v
-				}
-			}
-
 			if len(incoming.ModelAliases) > 0 {
 				if next.ModelAliases == nil {
 					next.ModelAliases = map[string]string{}
@@ -162,13 +143,3 @@ func (h *Handler) configImport(w http.ResponseWriter, r *http.Request) {
 		"message":           "config imported",
 	})
 }
-
-func (h *Handler) computeSyncHash() string {
-	snap := h.Store.Snapshot().Clone()
-	snap.ClearAccountTokens()
-	snap.VercelSyncHash = ""
-	snap.VercelSyncTime = 0
-	b, _ := json.Marshal(snap)
-	sum := md5.Sum(b)
-	return fmt.Sprintf("%x", sum)
-}

internal/httpapi/admin/configmgmt/handler_config_read.go

@@ -1,4 +1,4 @@
-package admin
+package configmgmt
 
 import (
 	"net/http"
@@ -18,24 +18,11 @@ func (h *Handler) getConfig(w http.ResponseWriter, _ *http.Request) {
 		"env_source_present":    h.Store.HasEnvConfigSource(),
 		"env_writeback_enabled": h.Store.IsEnvWritebackEnabled(),
 		"config_path":           h.Store.ConfigPath(),
-		"claude_mapping": func() map[string]string {
-			if len(snap.ClaudeMapping) > 0 {
-				return snap.ClaudeMapping
-			}
-			return snap.ClaudeModelMap
-		}(),
+		"model_aliases":         snap.ModelAliases,
 	}
 	accounts := make([]map[string]any, 0, len(snap.Accounts))
 	for _, acc := range snap.Accounts {
 		token := strings.TrimSpace(acc.Token)
-		preview := ""
-		if token != "" {
-			if len(token) > 20 {
-				preview = token[:20] + "..."
-			} else {
-				preview = token
-			}
-		}
 		accounts = append(accounts, map[string]any{
 			"identifier":    acc.Identifier(),
 			"name":          acc.Name,
@@ -45,7 +32,7 @@ func (h *Handler) getConfig(w http.ResponseWriter, _ *http.Request) {
 			"proxy_id":      acc.ProxyID,
 			"has_password":  strings.TrimSpace(acc.Password) != "",
 			"has_token":     token != "",
-			"token_preview": preview,
+			"token_preview": maskSecretPreview(token),
 		})
 	}
 	safe["accounts"] = accounts

internal/httpapi/admin/configmgmt/handler_config_write.go

@@ -1,4 +1,4 @@
-package admin
+package configmgmt
 
 import (
 	"encoding/json"
@@ -58,12 +58,12 @@ func (h *Handler) updateConfig(w http.ResponseWriter, r *http.Request) {
 			}
 			c.Accounts = accounts
 		}
-		if m, ok := req["claude_mapping"].(map[string]any); ok {
-			newMap := map[string]string{}
+		if m, ok := req["model_aliases"].(map[string]any); ok {
+			aliases := make(map[string]string, len(m))
 			for k, v := range m {
-				newMap[k] = fmt.Sprintf("%v", v)
+				aliases[k] = fmt.Sprintf("%v", v)
 			}
-			c.ClaudeMapping = newMap
+			c.ModelAliases = aliases
 		}
 		return nil
 	})

internal/httpapi/admin/configmgmt/handler_keys_test.go

@@ -1,4 +1,4 @@
-package admin
+package configmgmt
 
 import (
 	"bytes"

internal/httpapi/admin/configmgmt/routes.go

@@ -0,0 +1,27 @@
+package configmgmt
+
+import (
+	"net/http"
+
+	"github.com/go-chi/chi/v5"
+)
+
+func RegisterRoutes(r chi.Router, h *Handler) {
+	r.Get("/config", h.getConfig)
+	r.Post("/config", h.updateConfig)
+	r.Post("/config/import", h.configImport)
+	r.Get("/config/export", h.configExport)
+	r.Get("/export", h.exportConfig)
+	r.Post("/keys", h.addKey)
+	r.Put("/keys/{key}", h.updateKey)
+	r.Delete("/keys/{key}", h.deleteKey)
+	r.Post("/import", h.batchImport)
+}
+
+func (h *Handler) GetConfig(w http.ResponseWriter, r *http.Request)    { h.getConfig(w, r) }
+func (h *Handler) UpdateConfig(w http.ResponseWriter, r *http.Request) { h.updateConfig(w, r) }
+func (h *Handler) ConfigImport(w http.ResponseWriter, r *http.Request) { h.configImport(w, r) }
+func (h *Handler) BatchImport(w http.ResponseWriter, r *http.Request)  { h.batchImport(w, r) }
+func (h *Handler) AddKey(w http.ResponseWriter, r *http.Request)       { h.addKey(w, r) }
+func (h *Handler) UpdateKey(w http.ResponseWriter, r *http.Request)    { h.updateKey(w, r) }
+func (h *Handler) DeleteKey(w http.ResponseWriter, r *http.Request)    { h.deleteKey(w, r) }

internal/httpapi/admin/configmgmt/test_helpers_test.go

@@ -0,0 +1,18 @@
+package configmgmt
+
+import (
+	"testing"
+
+	"ds2api/internal/account"
+	"ds2api/internal/config"
+)
+
+func newAdminTestHandler(t *testing.T, raw string) *Handler {
+	t.Helper()
+	t.Setenv("DS2API_CONFIG_JSON", raw)
+	store := config.LoadStore()
+	return &Handler{
+		Store: store,
+		Pool:  account.NewPool(store),
+	}
+}

internal/httpapi/admin/devcapture/deps.go

@@ -0,0 +1,16 @@
+package devcapture
+
+import (
+	"ds2api/internal/chathistory"
+	adminshared "ds2api/internal/httpapi/admin/shared"
+)
+
+type Handler struct {
+	Store       adminshared.ConfigStore
+	Pool        adminshared.PoolController
+	DS          adminshared.DeepSeekCaller
+	OpenAI      adminshared.OpenAIChatCaller
+	ChatHistory *chathistory.Store
+}
+
+var writeJSON = adminshared.WriteJSON

internal/httpapi/admin/devcapture/handler_dev_capture.go

@@ -1,4 +1,4 @@
-package admin
+package devcapture
 
 import (
 	"net/http"

internal/httpapi/admin/devcapture/handler_dev_capture_test.go

@@ -1,4 +1,4 @@
-package admin
+package devcapture
 
 import (
 	"encoding/json"

internal/httpapi/admin/devcapture/routes.go

@@ -0,0 +1,8 @@
+package devcapture
+
+import "github.com/go-chi/chi/v5"
+
+func RegisterRoutes(r chi.Router, h *Handler) {
+	r.Get("/dev/captures", h.getDevCaptures)
+	r.Delete("/dev/captures", h.clearDevCaptures)
+}

internal/httpapi/admin/handler.go

@@ -0,0 +1,70 @@
+package admin
+
+import (
+	"github.com/go-chi/chi/v5"
+
+	"ds2api/internal/chathistory"
+	adminaccounts "ds2api/internal/httpapi/admin/accounts"
+	adminauth "ds2api/internal/httpapi/admin/auth"
+	adminconfig "ds2api/internal/httpapi/admin/configmgmt"
+	admindevcapture "ds2api/internal/httpapi/admin/devcapture"
+	adminhistory "ds2api/internal/httpapi/admin/history"
+	adminproxies "ds2api/internal/httpapi/admin/proxies"
+	adminrawsamples "ds2api/internal/httpapi/admin/rawsamples"
+	adminsettings "ds2api/internal/httpapi/admin/settings"
+	adminshared "ds2api/internal/httpapi/admin/shared"
+	adminvercel "ds2api/internal/httpapi/admin/vercel"
+	adminversion "ds2api/internal/httpapi/admin/version"
+)
+
+type Handler struct {
+	Store       adminshared.ConfigStore
+	Pool        adminshared.PoolController
+	DS          adminshared.DeepSeekCaller
+	OpenAI      adminshared.OpenAIChatCaller
+	ChatHistory *chathistory.Store
+}
+
+func RegisterRoutes(r chi.Router, h *Handler) {
+	deps := adminsharedDeps(h)
+	authHandler := &adminauth.Handler{Store: deps.Store, Pool: deps.Pool, DS: deps.DS, OpenAI: deps.OpenAI, ChatHistory: deps.ChatHistory}
+	accountsHandler := &adminaccounts.Handler{Store: deps.Store, Pool: deps.Pool, DS: deps.DS, OpenAI: deps.OpenAI, ChatHistory: deps.ChatHistory}
+	configHandler := &adminconfig.Handler{Store: deps.Store, Pool: deps.Pool, DS: deps.DS, OpenAI: deps.OpenAI, ChatHistory: deps.ChatHistory}
+	settingsHandler := &adminsettings.Handler{Store: deps.Store, Pool: deps.Pool, DS: deps.DS, OpenAI: deps.OpenAI, ChatHistory: deps.ChatHistory}
+	proxiesHandler := &adminproxies.Handler{Store: deps.Store, Pool: deps.Pool, DS: deps.DS, OpenAI: deps.OpenAI, ChatHistory: deps.ChatHistory}
+	rawSamplesHandler := &adminrawsamples.Handler{Store: deps.Store, Pool: deps.Pool, DS: deps.DS, OpenAI: deps.OpenAI, ChatHistory: deps.ChatHistory}
+	vercelHandler := &adminvercel.Handler{Store: deps.Store, Pool: deps.Pool, DS: deps.DS, OpenAI: deps.OpenAI, ChatHistory: deps.ChatHistory}
+	historyHandler := &adminhistory.Handler{Store: deps.Store, Pool: deps.Pool, DS: deps.DS, OpenAI: deps.OpenAI, ChatHistory: deps.ChatHistory}
+	devCaptureHandler := &admindevcapture.Handler{Store: deps.Store, Pool: deps.Pool, DS: deps.DS, OpenAI: deps.OpenAI, ChatHistory: deps.ChatHistory}
+	versionHandler := &adminversion.Handler{Store: deps.Store, Pool: deps.Pool, DS: deps.DS, OpenAI: deps.OpenAI, ChatHistory: deps.ChatHistory}
+
+	adminauth.RegisterPublicRoutes(r, authHandler)
+	r.Group(func(pr chi.Router) {
+		pr.Use(authHandler.RequireAdmin)
+		adminauth.RegisterProtectedRoutes(pr, authHandler)
+		adminconfig.RegisterRoutes(pr, configHandler)
+		adminsettings.RegisterRoutes(pr, settingsHandler)
+		adminproxies.RegisterRoutes(pr, proxiesHandler)
+		adminaccounts.RegisterRoutes(pr, accountsHandler)
+		adminrawsamples.RegisterRoutes(pr, rawSamplesHandler)
+		adminvercel.RegisterRoutes(pr, vercelHandler)
+		admindevcapture.RegisterRoutes(pr, devCaptureHandler)
+		adminhistory.RegisterRoutes(pr, historyHandler)
+		adminversion.RegisterRoutes(pr, versionHandler)
+	})
+}
+
+func adminsharedDeps(h *Handler) adminsharedDepsValue {
+	if h == nil {
+		return adminsharedDepsValue{}
+	}
+	return adminsharedDepsValue{Store: h.Store, Pool: h.Pool, DS: h.DS, OpenAI: h.OpenAI, ChatHistory: h.ChatHistory}
+}
+
+type adminsharedDepsValue struct {
+	Store       adminshared.ConfigStore
+	Pool        adminshared.PoolController
+	DS          adminshared.DeepSeekCaller
+	OpenAI      adminshared.OpenAIChatCaller
+	ChatHistory *chathistory.Store
+}

internal/httpapi/admin/handler_settings_test.go

@@ -189,8 +189,8 @@ func TestUpdateSettingsHistorySplit(t *testing.T) {
 		t.Fatalf("expected 200, got %d body=%s", rec.Code, rec.Body.String())
 	}
 	snap := h.Store.Snapshot()
-	if snap.HistorySplit.Enabled == nil || *snap.HistorySplit.Enabled {
-		t.Fatalf("expected history_split.enabled=false, got %#v", snap.HistorySplit.Enabled)
+	if snap.HistorySplit.Enabled == nil || !*snap.HistorySplit.Enabled {
+		t.Fatalf("expected history_split.enabled to be forced true, got %#v", snap.HistorySplit.Enabled)
 	}
 	if snap.HistorySplit.TriggerAfterTurns == nil || *snap.HistorySplit.TriggerAfterTurns != 3 {
 		t.Fatalf("expected history_split.trigger_after_turns=3, got %#v", snap.HistorySplit.TriggerAfterTurns)
@@ -346,6 +346,34 @@ func TestUpdateConfigLegacyKeysPreserveStructuredMetadata(t *testing.T) {
 	}
 }
 
+func TestUpdateConfigReplacesModelAliases(t *testing.T) {
+	h := newAdminTestHandler(t, `{
+		"keys":["k1"],
+		"model_aliases":{"claude-sonnet-4-6":"deepseek-v4-flash"}
+	}`)
+
+	payload := map[string]any{
+		"model_aliases": map[string]any{
+			"gpt-5.5": "deepseek-v4-pro",
+		},
+	}
+	b, _ := json.Marshal(payload)
+	req := httptest.NewRequest(http.MethodPost, "/admin/config", bytes.NewReader(b))
+	rec := httptest.NewRecorder()
+	h.updateConfig(rec, req)
+	if rec.Code != http.StatusOK {
+		t.Fatalf("status=%d body=%s", rec.Code, rec.Body.String())
+	}
+
+	snap := h.Store.Snapshot()
+	if len(snap.ModelAliases) != 1 {
+		t.Fatalf("expected aliases to be replaced, got %#v", snap.ModelAliases)
+	}
+	if snap.ModelAliases["gpt-5.5"] != "deepseek-v4-pro" {
+		t.Fatalf("expected updated alias, got %#v", snap.ModelAliases)
+	}
+}
+
 func TestUpdateSettingsPasswordInvalidatesOldJWT(t *testing.T) {
 	hash := authn.HashAdminPassword("old-password")
 	h := newAdminTestHandler(t, `{"admin":{"password_hash":"`+hash+`"}}`)

internal/httpapi/admin/handler_test.go

@@ -1,6 +1,9 @@
 package admin
 
 import (
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
 	"sync/atomic"
 	"testing"
 	"time"
@@ -33,6 +36,53 @@ func TestFieldStringNilToEmpty(t *testing.T) {
 	}
 }
 
+func TestMaskSecretPreviewKeepsOnlyFirstAndLastTwoChars(t *testing.T) {
+	cases := map[string]string{
+		"":         "",
+		"a":        "*",
+		"ab":       "**",
+		"abcd":     "****",
+		"abcdef":   "ab****ef",
+		"abc12345": "ab****45",
+	}
+
+	for input, want := range cases {
+		if got := maskSecretPreview(input); got != want {
+			t.Fatalf("maskSecretPreview(%q)=%q want %q", input, got, want)
+		}
+	}
+}
+
+func TestGetConfigMasksAccountTokenPreview(t *testing.T) {
+	h := newAdminTestHandler(t, `{
+		"accounts":[{"email":"u@example.com","password":"pwd"}]
+	}`)
+	if err := h.Store.UpdateAccountToken("u@example.com", "abcdefgh"); err != nil {
+		t.Fatalf("seed runtime token: %v", err)
+	}
+
+	req := httptest.NewRequest(http.MethodGet, "/admin/config", nil)
+	rec := httptest.NewRecorder()
+	h.getConfig(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("unexpected status: %d body=%s", rec.Code, rec.Body.String())
+	}
+
+	var payload map[string]any
+	if err := json.Unmarshal(rec.Body.Bytes(), &payload); err != nil {
+		t.Fatalf("decode response failed: %v", err)
+	}
+	accounts, _ := payload["accounts"].([]any)
+	if len(accounts) != 1 {
+		t.Fatalf("expected 1 account, got %d", len(accounts))
+	}
+	first, _ := accounts[0].(map[string]any)
+	if got, _ := first["token_preview"].(string); got != "ab****gh" {
+		t.Fatalf("expected masked token preview, got %q", got)
+	}
+}
+
 func TestRunAccountTestsConcurrentlyKeepsInputOrder(t *testing.T) {
 	accounts := []config.Account{
 		{Email: "a@example.com"},

internal/httpapi/admin/history/deps.go

@@ -0,0 +1,16 @@
+package history
+
+import (
+	"ds2api/internal/chathistory"
+	adminshared "ds2api/internal/httpapi/admin/shared"
+)
+
+type Handler struct {
+	Store       adminshared.ConfigStore
+	Pool        adminshared.PoolController
+	DS          adminshared.DeepSeekCaller
+	OpenAI      adminshared.OpenAIChatCaller
+	ChatHistory *chathistory.Store
+}
+
+var writeJSON = adminshared.WriteJSON

internal/httpapi/admin/history/handler_chat_history.go

@@ -1,4 +1,4 @@
-package admin
+package history
 
 import (
 	"encoding/json"

internal/httpapi/admin/history/handler_chat_history_test.go

@@ -1,4 +1,4 @@
-package admin
+package history
 
 import (
 	"bytes"
@@ -38,7 +38,7 @@ func TestGetChatHistoryAndUpdateSettings(t *testing.T) {
 	entry, err := historyStore.Start(chathistory.StartParams{
 		CallerID:  "caller:test",
 		AccountID: "user@example.com",
-		Model:     "deepseek-chat",
+		Model:     "deepseek-v4-flash",
 		UserInput: "hello",
 	})
 	if err != nil {

internal/httpapi/admin/history/routes.go

@@ -0,0 +1,11 @@
+package history
+
+import "github.com/go-chi/chi/v5"
+
+func RegisterRoutes(r chi.Router, h *Handler) {
+	r.Get("/chat-history", h.getChatHistory)
+	r.Get("/chat-history/{id}", h.getChatHistoryItem)
+	r.Delete("/chat-history", h.clearChatHistory)
+	r.Delete("/chat-history/{id}", h.deleteChatHistoryItem)
+	r.Put("/chat-history/settings", h.updateChatHistorySettings)
+}

internal/httpapi/admin/proxies/deps.go

@@ -0,0 +1,32 @@
+package proxies
+
+import (
+	"ds2api/internal/chathistory"
+	"ds2api/internal/config"
+	adminshared "ds2api/internal/httpapi/admin/shared"
+)
+
+type Handler struct {
+	Store       adminshared.ConfigStore
+	Pool        adminshared.PoolController
+	DS          adminshared.DeepSeekCaller
+	OpenAI      adminshared.OpenAIChatCaller
+	ChatHistory *chathistory.Store
+}
+
+var writeJSON = adminshared.WriteJSON
+
+func fieldString(m map[string]any, key string) string {
+	return adminshared.FieldString(m, key)
+}
+func accountMatchesIdentifier(acc config.Account, identifier string) bool {
+	return adminshared.AccountMatchesIdentifier(acc, identifier)
+}
+func toProxy(m map[string]any) config.Proxy { return adminshared.ToProxy(m) }
+func findProxyByID(c config.Config, proxyID string) (config.Proxy, bool) {
+	return adminshared.FindProxyByID(c, proxyID)
+}
+func newRequestError(detail string) error { return adminshared.NewRequestError(detail) }
+func requestErrorDetail(err error) (string, bool) {
+	return adminshared.RequestErrorDetail(err)
+}