Merge pull request #266 from CJackHwang/codex/update-documentation-for-accuracy

docs: align API docs with implemented routes and limits
This commit is contained in:
CJACK.
2026-04-19 23:43:28 +08:00
committed by GitHub
7 changed files with 201 additions and 74 deletions

View File

@@ -37,7 +37,7 @@ Docs: [Overview](README.en.md) / [Architecture](docs/ARCHITECTURE.en.md) / [Depl
- 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: structured parsing first (JSON/XML/invoke/markup), plus stream-time anti-leak filtering.
- 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.
- `Admin API` separates static config from runtime policy: `/admin/config*` for configuration state, `/admin/settings*` for runtime behavior.
---
@@ -108,6 +108,7 @@ Gemini-compatible clients can also send `x-goog-api-key`, `?key=`, or `?api_key=
| POST | `/v1/responses` | Business | OpenAI Responses API (stream/non-stream) |
| GET | `/v1/responses/{response_id}` | Business | Query stored response (in-memory TTL) |
| POST | `/v1/embeddings` | Business | OpenAI Embeddings API |
| POST | `/v1/files` | Business | OpenAI Files upload (multipart/form-data) |
| GET | `/anthropic/v1/models` | None | Claude model list |
| POST | `/anthropic/v1/messages` | Business | Claude messages |
| POST | `/anthropic/v1/messages/count_tokens` | Business | Claude token counting |
@@ -131,9 +132,15 @@ Gemini-compatible clients can also send `x-goog-api-key`, `?key=`, or `?api_key=
| GET | `/admin/config/export` | Admin | Export full config (`config`/`json`/`base64`) |
| POST | `/admin/keys` | Admin | Add API key |
| DELETE | `/admin/keys/{key}` | Admin | Delete API key |
| GET | `/admin/proxies` | Admin | List proxies |
| POST | `/admin/proxies` | Admin | Add proxy |
| PUT | `/admin/proxies/{proxyID}` | Admin | Update proxy (empty password keeps old secret) |
| DELETE | `/admin/proxies/{proxyID}` | Admin | Delete proxy (auto-unbind referenced accounts) |
| POST | `/admin/proxies/test` | Admin | Test proxy connectivity |
| GET | `/admin/accounts` | Admin | Paginated account list |
| POST | `/admin/accounts` | Admin | Add account |
| DELETE | `/admin/accounts/{identifier}` | Admin | Delete account |
| PUT | `/admin/accounts/{identifier}/proxy` | Admin | Bind/unbind proxy for an account |
| GET | `/admin/queue/status` | Admin | Account queue status |
| POST | `/admin/accounts/test` | Admin | Test one account |
| POST | `/admin/accounts/test-all` | Admin | Test all accounts |
@@ -208,6 +215,13 @@ 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):
- 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`
### `POST /v1/chat/completions`
**Headers**:
@@ -221,7 +235,7 @@ Content-Type: application/json
| Field | Type | Required | Notes |
| --- | --- | --- | --- |
| `model` | string | ✅ | DeepSeek native models + common aliases (`gpt-4o`, `gpt-5-codex`, `o3`, `claude-sonnet-4-5`, `gemini-2.5-pro`, etc.) |
| `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.) |
| `messages` | array | ✅ | OpenAI-style messages |
| `stream` | boolean | ❌ | Default `false` |
| `tools` | array | ❌ | Function calling schema |
@@ -312,7 +326,12 @@ When `tools` is present, DS2API performs anti-leak handling:
}
```
**Stream**: Once high-confidence toolcall features are matched, DS2API emits `delta.tool_calls` immediately (without waiting for full JSON closure), then keeps sending argument deltas; confirmed raw tool JSON is never forwarded as `delta.content`.
**Stream**: Once high-confidence toolcall features are matched, DS2API emits `delta.tool_calls` immediately (without waiting for full argument closure), then keeps sending argument deltas; confirmed tool-call fragments are not forwarded as `delta.content`.
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.
- `tool_calls` shown inside fenced markdown code blocks (for example, ```json ... ```) are treated as examples, not executable calls.
---
@@ -391,6 +410,21 @@ Business auth required. Returns OpenAI-compatible embeddings shape.
> Requires `embeddings.provider`. Current supported values: `mock` / `deterministic` / `builtin`. If missing/unsupported, returns standard error shape with HTTP 501.
### `POST /v1/files`
Business auth required. OpenAI Files-compatible upload endpoint; currently only `multipart/form-data` is supported.
| Field | Type | Required | Notes |
| --- | --- | --- | --- |
| `file` | file | ✅ | Binary payload |
| `purpose` | string | ❌ | Forwarded purpose field |
Constraints and behavior:
- `Content-Type` must be `multipart/form-data` (otherwise `400`).
- Total request size limit is `100 MiB` (over-limit returns `413`).
- Success returns an OpenAI `file` object (`id/object/bytes/filename/purpose/status`, etc.) and includes `account_id` for source-account tracing.
---
## Claude-Compatible API
@@ -723,6 +757,26 @@ Exports full config in three forms: `config`, `json`, and `base64`.
**Response**: `{"success": true, "total_keys": 2}`
### `GET /admin/proxies`
Lists proxy configs (password is never returned; use `has_password` as a marker).
### `POST /admin/proxies`
Adds a proxy. Request accepts `id` (optional; auto-generated when omitted), `name`, `type` (`http` / `socks5`), `host`, `port`, `username`, `password`.
### `PUT /admin/proxies/{proxyID}`
Updates a proxy. If `password` is an empty string, the existing secret is preserved.
### `DELETE /admin/proxies/{proxyID}`
Deletes a proxy and automatically clears `proxy_id` on all accounts that reference it.
### `POST /admin/proxies/test`
Tests proxy connectivity: provide `proxy_id` to test a saved proxy; omit it to run a one-off test using proxy fields in the request body.
### `GET /admin/accounts`
**Query params**:
@@ -730,7 +784,7 @@ Exports full config in three forms: `config`, `json`, and `base64`.
| Param | Default | Range |
| --- | --- | --- |
| `page` | `1` | ≥ 1 |
| `page_size` | `10` | 1100 |
| `page_size` | `10` | 15000 |
| `q` | empty | Filter by identifier / email / mobile |
**Response**:
@@ -771,6 +825,14 @@ Returned items also include `test_status`, usually `ok` or `failed`.
**Response**: `{"success": true, "total_accounts": 5}`
### `PUT /admin/accounts/{identifier}/proxy`
Updates proxy binding for a specific account.
- Request body: `{"proxy_id":"..."}`.
- Use empty `proxy_id` to unbind proxy.
- `identifier` supports email / mobile / token-only synthetic id.
### `GET /admin/queue/status`
```json

67
API.md
View File

@@ -37,7 +37,7 @@
- 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:优先结构化解析JSON/XML/invoke/markup),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。
- `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。
---
@@ -108,6 +108,7 @@ Gemini 兼容客户端还可以使用 `x-goog-api-key`、`?key=` 或 `?api_key=`
| POST | `/v1/responses` | 业务 | OpenAI Responses 接口(流式/非流式) |
| GET | `/v1/responses/{response_id}` | 业务 | 查询已生成 response内存 TTL |
| POST | `/v1/embeddings` | 业务 | OpenAI Embeddings 接口 |
| POST | `/v1/files` | 业务 | OpenAI Files 上传multipart/form-data |
| GET | `/anthropic/v1/models` | 无 | Claude 模型列表 |
| POST | `/anthropic/v1/messages` | 业务 | Claude 消息接口 |
| POST | `/anthropic/v1/messages/count_tokens` | 业务 | Claude token 计数 |
@@ -131,9 +132,15 @@ Gemini 兼容客户端还可以使用 `x-goog-api-key`、`?key=` 或 `?api_key=`
| GET | `/admin/config/export` | Admin | 导出完整配置(含 `config`/`json`/`base64` |
| POST | `/admin/keys` | Admin | 添加 API key |
| DELETE | `/admin/keys/{key}` | Admin | 删除 API key |
| GET | `/admin/proxies` | Admin | 代理列表 |
| POST | `/admin/proxies` | Admin | 添加代理 |
| PUT | `/admin/proxies/{proxyID}` | Admin | 更新代理(留空 password 表示保留原密码) |
| DELETE | `/admin/proxies/{proxyID}` | Admin | 删除代理(自动解绑引用该代理的账号) |
| POST | `/admin/proxies/test` | Admin | 测试代理连通性 |
| GET | `/admin/accounts` | Admin | 分页账号列表 |
| POST | `/admin/accounts` | Admin | 添加账号 |
| DELETE | `/admin/accounts/{identifier}` | Admin | 删除账号 |
| PUT | `/admin/accounts/{identifier}/proxy` | Admin | 为账号绑定/解绑代理 |
| GET | `/admin/queue/status` | Admin | 账号队列状态 |
| POST | `/admin/accounts/test` | Admin | 测试单个账号 |
| POST | `/admin/accounts/test-all` | Admin | 测试全部账号 |
@@ -208,6 +215,13 @@ 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
- 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`
### `POST /v1/chat/completions`
**请求头**
@@ -221,7 +235,7 @@ Content-Type: application/json
| 字段 | 类型 | 必填 | 说明 |
| --- | --- | --- | --- |
| `model` | string | ✅ | 支持 DeepSeek 原生模型 + 常见 alias`gpt-4o``gpt-5-codex``o3``claude-sonnet-4-5``gemini-2.5-pro` 等) |
| `model` | string | ✅ | 支持 DeepSeek 原生模型 + 常见 alias`gpt-5``gpt-5-mini``gpt-5-codex``o3``claude-opus-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 |
@@ -312,12 +326,12 @@ data: [DONE]
}
```
**流式**:命中高置信特征后立即输出 `delta.tool_calls`(不等待完整 JSON 闭合),并持续发送 arguments 增量;已确认的 toolcall 原始 JSON 不会回流到 `delta.content`
**流式**:命中高置信特征后立即输出 `delta.tool_calls`(不等待完整工具参数闭合),并持续发送 arguments 增量;已确认的工具调用片段不会回流到 `delta.content`
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 XML/Markup 为最高优先级,并兼容 JSON、ANTML、text-kv 等格式输入;最终按客户端协议转译为对应 tool call 结构OpenAI/Claude/Gemini
- 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
- Markdown fenced code block例如 ```json ... ```)中的 `tool_calls` 仅视为示例文本,不会被执行。
---
@@ -397,6 +411,21 @@ data: [DONE]
> 需配置 `embeddings.provider`。当前支持:`mock` / `deterministic` / `builtin`。未配置或不支持时返回标准错误结构HTTP 501
### `POST /v1/files`
需要业务鉴权。兼容 OpenAI Files 上传接口,当前仅支持 `multipart/form-data`
| 字段 | 类型 | 必填 | 说明 |
| --- | --- | --- | --- |
| `file` | file | ✅ | 上传文件二进制 |
| `purpose` | string | ❌ | 透传到上游用途字段 |
约束与行为:
- 请求必须为 `multipart/form-data`,否则返回 `400`
- 请求体总大小上限 `100 MiB`(超限返回 `413`)。
- 成功返回 OpenAI `file` 对象(`id/object/bytes/filename/purpose/status` 等字段),并附带 `account_id` 便于定位来源账号。
---
## Claude 兼容接口
@@ -729,6 +758,26 @@ data: {"type":"message_stop"}
**响应**`{"success": true, "total_keys": 2}`
### `GET /admin/proxies`
列出代理配置(密码不回传,仅返回 `has_password` 标记)。
### `POST /admin/proxies`
新增代理。请求体支持 `id`(可选,未传则自动生成)、`name``type``http` / `socks5`)、`host``port``username``password`
### `PUT /admin/proxies/{proxyID}`
更新指定代理。若请求中 `password` 为空字符串,则保留原密码。
### `DELETE /admin/proxies/{proxyID}`
删除代理,并自动清空所有引用该代理账号的 `proxy_id`
### `POST /admin/proxies/test`
测试代理连通性:传 `proxy_id` 时测试已保存代理;不传时按请求体代理字段做临时连通性测试。
### `GET /admin/accounts`
**查询参数**
@@ -736,7 +785,7 @@ data: {"type":"message_stop"}
| 参数 | 默认 | 范围 |
| --- | --- | --- |
| `page` | `1` | ≥ 1 |
| `page_size` | `10` | 1100 |
| `page_size` | `10` | 15000 |
| `q` | 空 | 按 identifier / email / mobile 过滤 |
**响应**
@@ -775,6 +824,14 @@ data: {"type":"message_stop"}
**响应**`{"success": true, "total_accounts": 5}`
### `PUT /admin/accounts/{identifier}/proxy`
更新指定账号绑定代理。
- 请求体:`{"proxy_id":"..."}`
- `proxy_id` 传空字符串时表示解绑代理;
- `identifier` 支持 email / mobile / token-only 合成标识。
### `GET /admin/queue/status`
```json

View File

@@ -87,7 +87,7 @@ flowchart LR
- **统一路由内核**:所有协议入口统一汇聚到 `internal/server/router.go`,并在同一路由树中注册 OpenAI / Claude / Gemini / Admin / WebUI 路由,避免多入口行为漂移。
- **统一执行链路**Claude / Gemini 入口先经 `internal/translatorcliproxy` 做协议转换,再进入 `openai.ChatCompletions` 统一处理工具调用与流式语义,最后再转换回原协议响应。
- **适配器分层更清晰**`internal/adapter/{claude,gemini}` 负责入口/出口协议封装,`internal/adapter/openai` 负责核心执行DeepSeek 侧调用只保留在 OpenAI 内核中。
- **Tool Calling 双运行时对齐**Go 侧(`internal/toolcall`)与 Vercel Node 侧(`internal/js/helpers/stream-tool-sieve`)保持一致的解析/防泄漏语义,覆盖 JSON / XML / invoke / text-kv 多风格输入
- **Tool Calling 双运行时对齐**Go 侧(`internal/toolcall`)与 Vercel Node 侧(`internal/js/helpers/stream-tool-sieve`)保持一致的解析/防泄漏语义;当前以 XML/Markup 家族为主(`<tool_call>` / `<function_call>` / `<invoke>` / `tool_use` / antml 变体)
- **配置与运行时设置解耦**:静态配置(`config`)与运行时策略(`settings`)通过 Admin API 分离管理,支持热更新和密码轮换失效旧 JWT。
- **流式能力升级**`/v1/responses` 与 `/v1/chat/completions` 共享更一致的工具调用增量输出策略,降低不同 SDK 下的行为差异。
- **可观测与可运维增强**`/healthz`、`/readyz`、`/admin/version`、`/admin/dev/captures` 形成排障闭环,便于发布后验证。
@@ -96,14 +96,14 @@ flowchart LR
| 能力 | 说明 |
| --- | --- |
| 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` |
| 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) |
| 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 |
| Tool Calling | 防泄漏处理:非代码块高置信特征识别、`delta.tool_calls` 早发、结构化增量输出 |
| Admin API | 配置管理、运行时设置热更新、账号测试 / 批量测试、会话清理、导入导出、Vercel 同步、版本检查 |
| Admin API | 配置管理、运行时设置热更新、代理管理、账号测试 / 批量测试、会话清理、导入导出、Vercel 同步、版本检查 |
| WebUI 管理台 | `/admin` 单页应用(中英文双语、深色模式) |
| 运维探针 | `GET /healthz`(存活)、`GET /readyz`(就绪) |
@@ -137,7 +137,7 @@ flowchart LR
| vision | `deepseek-vision-chat-search` | ❌ | ✅ |
| vision | `deepseek-vision-reasoner-search` | ✅ | ✅ |
除原生模型外,也支持常见 alias 输入(如 `gpt-4o`、`gpt-5-codex`、`o3`、`claude-sonnet-4-5`、`gemini-2.5-pro` 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-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。
### Claude 接口(`GET /anthropic/v1/models`
@@ -155,7 +155,7 @@ flowchart LR
- `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,请升级到包含 Claude 工具调用多格式解析JSON/XML/ANTML/invoke的版本
- 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
### Gemini 接口
@@ -293,8 +293,12 @@ go run ./cmd/ds2api
],
"model_aliases": {
"gpt-4o": "deepseek-chat",
"gpt-5": "deepseek-chat",
"gpt-5-mini": "deepseek-chat",
"gpt-5-codex": "deepseek-reasoner",
"o3": "deepseek-reasoner"
"o3": "deepseek-reasoner",
"claude-opus-4-6": "deepseek-reasoner",
"gemini-2.5-flash": "deepseek-chat"
},
"compat": {
"wide_input_strict_output": true,
@@ -398,7 +402,7 @@ 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. 解析层以 XML/Markup 为最高优先级,同时兼容 JSON / ANTML / invoke / text-kv并统一归一到内部工具调用结构
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
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

View File

@@ -85,7 +85,7 @@ For the full module-by-module architecture and directory responsibilities, see [
- **Unified routing core**: all protocol entries are now centralized through `internal/server/router.go`, with OpenAI / Claude / Gemini / Admin / WebUI routes registered in one tree to avoid multi-entry drift.
- **Unified execution chain**: Claude/Gemini entries are translated by `internal/translatorcliproxy`, then executed through `openai.ChatCompletions` for shared tool-calling and stream semantics, then translated back to the client protocol.
- **Cleaner adapter boundaries**: `internal/adapter/{claude,gemini}` handles protocol wrappers, while `internal/adapter/openai` remains the execution core; upstream DeepSeek calls are retained only in the OpenAI core.
- **Tool-calling parity across runtimes**: Go (`internal/toolcall`) and Vercel Node (`internal/js/helpers/stream-tool-sieve`) follow aligned parsing/anti-leak semantics across JSON / XML / invoke / text-kv inputs.
- **Tool-calling parity across runtimes**: Go (`internal/toolcall`) and Vercel Node (`internal/js/helpers/stream-tool-sieve`) share aligned parsing/anti-leak semantics, now centered on XML/Markup-family payloads (`<tool_call>` / `<function_call>` / `<invoke>` / `tool_use` / antml variants).
- **Config/runtime separation**: static config (`config`) and runtime policy (`settings`) are managed independently via Admin APIs, enabling hot updates and password rotation with JWT invalidation.
- **Streaming behavior upgrade**: `/v1/responses` and `/v1/chat/completions` now share a more consistent incremental tool-call emission strategy across SDK ecosystems.
- **Improved operability**: `/healthz`, `/readyz`, `/admin/version`, and `/admin/dev/captures` form a tighter post-deploy diagnostics loop.
@@ -94,14 +94,14 @@ For the full module-by-module architecture and directory responsibilities, see [
| Capability | Details |
| --- | --- |
| OpenAI compatible | `GET /v1/models`, `GET /v1/models/{id}`, `POST /v1/chat/completions`, `POST /v1/responses`, `GET /v1/responses/{response_id}`, `POST /v1/embeddings` |
| 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) |
| 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 |
| Tool Calling | Anti-leak handling: non-code-block feature match, early `delta.tool_calls`, structured incremental output |
| Admin API | Config management, runtime settings hot-reload, account testing/batch test, session cleanup, import/export, Vercel sync, version check |
| Admin API | Config management, runtime settings hot-reload, proxy management, account testing/batch test, session cleanup, import/export, Vercel sync, version check |
| WebUI Admin Panel | SPA at `/admin` (bilingual Chinese/English, dark mode) |
| Health Probes | `GET /healthz` (liveness), `GET /readyz` (readiness) |
@@ -135,7 +135,7 @@ For the full module-by-module architecture and directory responsibilities, see [
| vision | `deepseek-vision-chat-search` | ❌ | ✅ |
| vision | `deepseek-vision-reasoner-search` | ✅ | ✅ |
Besides native IDs, DS2API also accepts common aliases as input (for example `gpt-4o`, `gpt-5-codex`, `o3`, `claude-sonnet-4-5`, `gemini-2.5-pro`), but `/v1/models` returns normalized DeepSeek native model IDs.
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.
### Claude Endpoint (`GET /anthropic/v1/models`)
@@ -153,7 +153,7 @@ Besides the current primary aliases above, `/anthropic/v1/models` also returns C
- 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, upgrade to a build that includes multi-format Claude tool-call parsing (JSON/XML/ANTML/invoke).
- 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`.
### Gemini Endpoint
@@ -291,8 +291,12 @@ The server actually binds to `0.0.0.0:5001`, so devices on the same LAN can usua
],
"model_aliases": {
"gpt-4o": "deepseek-chat",
"gpt-5": "deepseek-chat",
"gpt-5-mini": "deepseek-chat",
"gpt-5-codex": "deepseek-reasoner",
"o3": "deepseek-reasoner"
"o3": "deepseek-reasoner",
"claude-opus-4-6": "deepseek-reasoner",
"gemini-2.5-flash": "deepseek-chat"
},
"compat": {
"wide_input_strict_output": true,
@@ -396,7 +400,7 @@ 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 prioritizes XML/Markup, while also accepting JSON / ANTML / invoke / text-kv, and normalizes everything into the internal tool-call structure
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
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

View File

@@ -116,7 +116,7 @@ flowchart LR
- `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`: JSON/XML/invoke/text-kv tool-call parsing + anti-leak sieve.
- `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/config`: config loading/validation + runtime settings hot-reload.
- `internal/account`: managed account pool, inflight slots, waiting queue.

View File

@@ -116,7 +116,7 @@ flowchart LR
- `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`JSON/XML/invoke/text-kv 工具调用解析防泄漏筛分。
- `internal/toolcall`以 XML/Markup 家族为核心的工具调用解析防泄漏筛分`<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/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。

View File

@@ -1,74 +1,74 @@
# Tool call parsing semanticsGo/Node 统一语义)
本文档描述当前代码中 `ParseToolCallsDetailed` / `parseToolCallsDetailed` 的**实际行为**,用于对齐 Go 与 Node Runtime
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` `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) 输出结构(当前实现)
## 1) 当前输出结构
- `calls`:解析得到的工具调用列表(`name` + `input`)。
- `sawToolCallSyntax`:检测到工具调用语法特征时为 `true`(例如 `tool_calls``<tool_call>``<function_call>``<invoke>``function.name:`)。
- `rejectedByPolicy`:当前实现固定为 `false`(预留字段,尚未启用 allow-list 拒绝)。
`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)。
> 说明:`filterToolCallsDetailed` 当前仅做结构清洗,不做工具名策略拒绝。
> 当前 `filterToolCallsDetailed` 仅做结构清洗,不做 allow-list 工具名拒绝。
## 2) 解析管线
## 2) 解析范围(重点)
1. **示例保护**:若判定为 fenced code block 示例上下文,则跳过执行型解析。
2. **候选片段构建**:从完整文本中构建候选(原文、围绕 `tool_calls` 的 JSON 片段、首尾大括号切片等)。
3. **按序尝试解析(命中即停)**
- 对“明显 JSON 工具载荷候选”(以 `{`/`[` 开头且包含 `tool_calls`/`\"function\"`)先走 JSON 解析,避免 JSON 字符串内偶发 XML 片段误命中;
- 其余候选优先 XML 解析(`<tool_call>` / `<function_call>` / `<invoke>` / `tool_use` / `antml:function_call` 等);
- JSON 解析(`{"tool_calls": [...]}`、列表、单对象);
- Markup 解析;
- Text-KV 回退(如 `function.name:` + `function.arguments:`)。
4. **兜底**:候选全部失败后,再对全文做 XML / Text-KV 回退。
当前版本的可执行解析以 **XML/Markup 家族**为主:
## 3) XML 能力边界(当前)
- `<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`
当前已支持输入端的“多 XML/标记风格”解析,包括但不限于
并支持在这些标记块内部解析
- `<tool_call><tool_name>...</tool_name><parameters>...</parameters></tool_call>`
- `<function_call>tool</function_call><function parameter name="x">...</function parameter>`
- `<invoke name="tool"><parameter name="x">...</parameter></invoke>`
- `antml:function_call` / `antml:argument` / `antml:parameters`
- `tool_use` 家族标签
- JSON 参数字符串
- 标签参数(`<parameter name="...">...`
- key/value 风格子标签
但**输出端仍统一转换为 OpenAI 兼容 JSON 事件/对象**`message.tool_calls``delta.tool_calls``response.function_call_arguments.*`)。
## 3) 不应再假设的行为
## 4) 关于“是否可以封装成 XML 再喂给模型”
以下说法在当前实现中已不成立:
结论:**可以做,而且当前解析器已经能兼容 XML 作为输入格式之一**,但代码里并没有 `toolcall.prefer_xml_output` 这个开关。现有可调配置只有:
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”。
- `toolcall.mode``feature_match` / `off`
- `toolcall.early_emit_confidence``high` / `low` / `off`
当前策略在代码中固定为:
推荐思路仍然是“输入兼容层 + 输出按客户端协议渲染”:
- 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
1. **Prompt 约束层**:如果你要尝试 XML-first可以在系统提示词里约束模型输出规范 XML tool block例如 `<tool_calls><tool_call>...</tool_call></tool_calls>`)。
2. **解析兼容层**:继续在 parser 中同时接受 JSON / XML / ANTML / invoke / text-kv。
3. **协议归一层**:无论模型输出什么格式,统一落到内部 `ParsedToolCall`
4. **对外渲染层**根据客户端请求协议渲染OpenAI / Claude / Gemini 各自格式)。
## 4) 流式与防泄漏语义
这样可以同时获得
在流式链路中OpenAI / Claude / Gemini 统一内核)
- 减少模型端 JSON 转义/引号错误
- 不破坏现有 SDK / 客户端生态
- 逐步灰度(按模型、按租户、按请求开关)
- 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(低风险迭代
## 5) 落地建议(按当前实现
- 继续使用现有的 `toolcall.mode=feature_match``toolcall.early_emit_confidence=high` 作为默认策略
- 如果要试 XML-first把它放在 prompt 层或上游模板层,不要假设代码里已有专门的 XML 输出开关
- 增加观测指标:
- `toolcall_parse_source`json/xml/markup/textkv
- `toolcall_parse_success_rate`
- `toolcall_malformed_rate`
- `toolcall_repair_rate`
- 先在 `responses` 链路灰度,再扩展 `chat.completions`
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) 兼容性提醒
## 6) 回归验证建议
- 上游模型若输出混合文本 + XML仍可能出现“半结构化”噪声需要依赖现有 sieve 增量消费策略。
- XML 不等于安全:仍需做 tool 名、参数 schema、执行权限的服务端校验。
可直接运行:
```bash
go test -v -run 'TestParseToolCalls|TestRepair' ./internal/toolcall/
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