Compare commits

...

21 Commits

Author SHA1 Message Date
CJACK
049e40e5f1 fix: drop obsolete release smoke check 2026-05-02 04:19:23 +08:00
CJACK
b3c54fcf3d chore: bump version to 4.3.0
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-02 03:55:36 +08:00
CJACK
1c38709d32 feat: add support for parsing loose JSON lists into arrays in tool call parameters 2026-05-02 03:26:43 +08:00
CJACK
28e800c670 chore: remove obsolete planning and gate target files 2026-05-02 02:40:57 +08:00
CJACK
4389e02b29 feat: implement sync.Pool for tiktoken encoding instances to optimize token counting performance 2026-05-02 02:31:24 +08:00
CJACK
e2756f800d feat: introduce JSON UTF-8 validation middleware and prepend output integrity guard system prompt to messages 2026-05-02 02:22:34 +08:00
CJACK
55abf64717 feat: add model type support for file uploads with automatic resolution and header propagation 2026-05-02 00:55:17 +08:00
CJACK
76ee2faa12 chore: bump version to 4.2.2 and update documentation to reflect improved release workflows, CI dependencies, and project structure 2026-05-01 23:44:07 +08:00
CJACK
0bca6e2cee feat: implement context cancellation handling for chat and response stream runtimes to ensure clean termination without retries 2026-05-01 23:20:46 +08:00
CJACK.
934b40e572 Merge pull request #392 from wyv202011y/fix/timeout-and-context-cancel
fix: increase stream timeout constants for large-context models; guar…
2026-05-01 23:17:31 +08:00
CJACK
dd5a0c5213 refactor: update and standardize current input file continuation prompt instructions 2026-05-01 22:27:59 +08:00
CJACK
43402e7a26 refactor: rename history file constant from HISTORY.txt to DS2API_HISTORY.txt across codebase and tests 2026-05-01 22:05:45 +08:00
CJACK.
6373c001f5 Merge pull request #391 from BigUncle/fix/vercel-admin-history-rewrite
Fix: add missing Vercel rewrite rules for admin API routes
2026-05-01 21:45:14 +08:00
BigUncle
3430322e81 docs: add Vercel chat history read-only filesystem troubleshooting 2026-05-01 21:17:52 +08:00
CJACK
df1cfac9bc refactor: replace history transcript format with numbered sections and rename upload file to HISTORY.txt 2026-05-01 21:15:17 +08:00
706e68de23 fix: increase stream timeout constants for large-context models; guard against context-cancelled double-recording
- Increase StreamIdleTimeout from 90s to 300s and MaxKeepaliveCount from 10 to 40
  to prevent premature stream termination with DeepSeek V4 Pro (~50K token contexts)
- Add r.Context().Err() check after ConsumeSSE in empty_retry_runtime (chat + responses)
  to prevent historySession.error() from overwriting historySession.stopped()
  when the request context is cancelled

References:
- MaxKeepaliveCount=10 creates a 50s no-content timeout that kills the stream
  before DeepSeek V4 Pro can produce its first token with large contexts
- Hermes Agent reports 'No response from provider for 180s' because the
  underlying SSE connection was already terminated by ds2api at 50s
- Context cancellation path: OnContextDone -> stopped(), then finalize()
  with empty output -> retry -> error() overwrites stopped()
2026-05-01 21:11:36 +08:00
BigUncle
83b4c7bcad fix: add missing Vercel rewrite rules for admin API routes
/admin/chat-history, /admin/proxies, /admin/dev/raw-samples,
and /admin/dev/captures were falling through to the SPA fallback
(/admin/index.html), causing "Unexpected token '<'" JSON parse
errors on the frontend.
2026-05-01 20:50:12 +08:00
CJACK.
445c95a4f2 Merge pull request #379 from CJackHwang/dev
Merge pull request #377 from CJackHwang/codex/run-all-tests-and-fix-failures

Fix failing current-input token accounting test
2026-05-01 16:12:17 +08:00
CJACK
0a6ef8e3f2 fix: remove bufio.Scanner 2MiB line limit for SSE; support quasi_status direct patch
Replace bufio.Scanner with bufio.NewReaderSize + ReadBytes('\n') across all
SSE read paths to preserve long single-line data (e.g. write_file content).
Add quasi_status and auto_continue handling as direct path-based patches in
both Go continue observer and Node vercel_stream_impl, mirroring existing
batch-patch logic. Add 2MiB+ line throughput tests at every SSE layer.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-01 15:45:17 +08:00
CJACK
fd0ec29991 refactor: generalize DSML tag parsing to tolerate model noise; split tiktoken by build tags
Replace hardcoded DSML typo variant lists in Go/Node tool call parsers with
generalized prefix consumption that tolerates repeated leading <, repeated DSML
prefix noise, and trailing pipe terminators. Split tiktoken-dependent token
counting into a build-tagged file for non-cgo platform compatibility. Add /data
directory to Dockerfile for bind-mount permissions.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-01 15:17:11 +08:00
CJACK.
94c1acace5 Merge pull request #369 from CJackHwang/dev
Merge pull request #368 from CJackHwang/codex/fix-review-issues-for-pr-#364

Restore thinking fallback for tool-call detection and drop history.txt wrapper tags
2026-04-29 23:42:50 +08:00
83 changed files with 2512 additions and 659 deletions

View File

@@ -47,7 +47,6 @@ jobs:
- name: Release Blocking Gates
run: |
./tests/scripts/check-stage6-manual-smoke.sh
./tests/scripts/check-refactor-line-gate.sh
./tests/scripts/run-unit-all.sh

1
.gitignore vendored
View File

@@ -29,6 +29,7 @@ yarn.lock
pnpm-lock.yaml
# Build artifacts
dist/
*.tsbuildinfo
.cache/
.parcel-cache/

View File

@@ -33,6 +33,8 @@ Docs: [Overview](README.en.md) / [Architecture](docs/ARCHITECTURE.en.md) / [Depl
| Health probes | `GET /healthz`, `GET /readyz` |
| 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) |
- All JSON request bodies must be valid UTF-8; malformed byte sequences are rejected on ingress with `400 invalid json`.
### 3.0 Adapter-Layer Notes
- OpenAI / Claude / Gemini protocols are now mounted on one shared `chi` router tree assembled in `internal/server/router.go`.
@@ -81,7 +83,7 @@ Two header formats accepted:
- Token is in `config.keys`**Managed account mode**: DS2API auto-selects an account via rotation
- Token is not in `config.keys`**Direct token mode**: treated as a DeepSeek token directly
**Optional header**: `X-Ds2-Target-Account: <email_or_mobile>` — Pin a specific managed account.
**Optional header**: `X-Ds2-Target-Account: <email_or_mobile>` — Pin a specific managed account; if the target account does not exist or the managed-account queue is exhausted, the request returns `429`, and current responses do not include `Retry-After`. If the account exists but login/refresh fails, the request returns the underlying `401` or upstream error.
Gemini-compatible clients can also send `x-goog-api-key`, `?key=`, or `?api_key=` as the caller credential source.
### Admin Endpoints (`/admin/*`)
@@ -198,10 +200,15 @@ No auth required. Returns the currently supported DeepSeek native model list.
"object": "list",
"data": [
{"id": "deepseek-v4-flash", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
{"id": "deepseek-v4-flash-nothinking", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
{"id": "deepseek-v4-pro", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
{"id": "deepseek-v4-pro-nothinking", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
{"id": "deepseek-v4-flash-search", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
{"id": "deepseek-v4-flash-search-nothinking", "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-pro-search-nothinking", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
{"id": "deepseek-v4-vision", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []},
{"id": "deepseek-v4-vision-nothinking", "object": "model", "created": 1677610602, "owned_by": "deepseek", "permission": []}
]
}
```
@@ -300,7 +307,7 @@ data: [DONE]
- 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
- 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. Failed/interrupted endings (for example `response.failed`) may not include `usage`
#### Tool Calls
@@ -416,7 +423,7 @@ Business auth required. Returns OpenAI-compatible embeddings shape.
| `model` | string | ✅ | Supports native models + alias mapping |
| `input` | string/array | ✅ | Supports string, string array, token array |
> Requires `embeddings.provider`. Current supported values: `mock` / `deterministic` / `builtin`. If missing/unsupported, returns standard error shape with HTTP 501.
> Requires `embeddings.provider`. Current supported values: `mock` / `deterministic` / `builtin` (all three use the same local deterministic implementation). If missing/unsupported, returns standard error shape with HTTP 501.
### `POST /v1/files`
@@ -430,7 +437,7 @@ Business auth required. OpenAI Files-compatible upload endpoint; currently only
Constraints and behavior:
- `Content-Type` must be `multipart/form-data` (otherwise `400`).
- Total request size limit is `100 MiB` (over-limit returns `413`).
- 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.
---
@@ -484,6 +491,13 @@ anthropic-version: 2023-06-01
| `stream` | boolean | ❌ | Default `false` |
| `system` | string | ❌ | Optional system prompt |
| `tools` | array | ❌ | Claude tool schema |
| `thinking` | object | ❌ | Anthropic thinking config; translated into downstream reasoning control, and ignored by `-nothinking` models |
| `temperature` | number | ❌ | Passed through to the downstream bridge; if `temperature` and `top_p` are both present, `temperature` wins |
| `top_p` | number | ❌ | Passed through when `temperature` is absent |
| `stop_sequences` | array | ❌ | Passed through as downstream stop sequences |
| `tool_choice` | string/object | ❌ | Supports `auto` / `none` / `required` / `{"type":"function","name":"..."}` and is translated to downstream tool choice |
> Note: `thinking`, `temperature`, `top_p`, `stop_sequences`, and `tool_choice` are translated through the compatibility bridge. Final behavior still depends on the selected model and upstream support. When both `temperature` and `top_p` are present, `temperature` takes precedence.
#### Non-Stream Response
@@ -1212,7 +1226,7 @@ Clients should handle HTTP status code plus `error` / `detail` fields.
| Code | Meaning |
| --- | --- |
| `401` | Authentication failed (invalid key/token, or expired admin JWT) |
| `429` | Too many requests (exceeded inflight + queue capacity) |
| `429` | Too many requests (exceeded inflight + queue capacity; current responses do not include `Retry-After`) |
| `503` | Model unavailable or upstream error |
---

19
API.md
View File

@@ -33,6 +33,8 @@
| 健康检查 | `GET /healthz``GET /readyz` |
| 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) |
- 所有 JSON 请求体都必须是合法 UTF-8非法字节序列会在入站阶段被拒绝为 `400 invalid json`
### 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。
@@ -81,7 +83,7 @@ Vercel 一键部署可先只填 `DS2API_ADMIN_KEY`,部署后在 `/admin` 导
- token 在 `config.keys` 中 → **托管账号模式**,自动轮询选择账号
- token 不在 `config.keys` 中 → **直通 token 模式**,直接作为 DeepSeek token 使用
**可选请求头**`X-Ds2-Target-Account: <email_or_mobile>` — 指定使用某个托管账号。
**可选请求头**`X-Ds2-Target-Account: <email_or_mobile>` — 指定使用某个托管账号;如果目标账号不存在,或管理账号队列已耗尽,相关业务请求会返回 `429`,当前不会附带 `Retry-After` 头。若账号存在但登录/刷新失败,则返回对应的 `401` 或上游错误
Gemini 兼容客户端还可以使用 `x-goog-api-key``?key=``?api_key=` 作为凭据来源。
### Admin 接口(`/admin/*`
@@ -307,7 +309,7 @@ data: [DONE]
- 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``usage`
- token 计数优先透传上游 DeepSeek SSE`accumulated_token_usage` / `token_usage`);仅在上游缺失时回退本地估算
- token 计数优先透传上游 DeepSeek SSE`accumulated_token_usage` / `token_usage`);仅在上游缺失时回退本地估算。失败/中断型结束(例如 `response.failed`)可能不会携带 `usage`
#### Tool Calls
@@ -424,7 +426,7 @@ data: [DONE]
| `model` | string | ✅ | 支持原生模型 + alias 自动映射 |
| `input` | string/array | ✅ | 支持字符串、字符串数组、token 数组 |
> 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
> 需配置 `embeddings.provider`。当前支持:`mock` / `deterministic` / `builtin`(三者都走同一套本地确定性实现)。未配置或不支持时返回标准错误结构HTTP 501
### `POST /v1/files`
@@ -438,7 +440,7 @@ data: [DONE]
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`)。
- 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。
---
@@ -495,6 +497,13 @@ anthropic-version: 2023-06-01
| `stream` | boolean | ❌ | 默认 `false` |
| `system` | string | ❌ | 可选系统提示 |
| `tools` | array | ❌ | Claude tool 定义 |
| `thinking` | object | ❌ | Anthropic thinking 配置;会转译为下游 reasoning 控制,`-nothinking` 模型会忽略 |
| `temperature` | number | ❌ | 透传到下游;若同时提供 `top_p`,以 `temperature` 为准 |
| `top_p` | number | ❌ | 当未提供 `temperature` 时透传到下游 |
| `stop_sequences` | array | ❌ | 透传到下游停用序列 |
| `tool_choice` | string/object | ❌ | 支持 `auto` / `none` / `required` / `{"type":"function","name":"..."}`,并会转译为下游工具选择 |
> 说明:上述 `thinking`、`temperature`、`top_p`、`stop_sequences`、`tool_choice` 都会走兼容层转译;最终是否生效仍取决于当前模型和上游能力。`temperature` 与 `top_p` 同时存在时,`temperature` 优先。
#### 非流式响应
@@ -1226,7 +1235,7 @@ Gemini 路由使用 Google 风格错误结构:
| 状态码 | 说明 |
| --- | --- |
| `401` | 鉴权失败key/token 无效,或 Admin JWT 过期) |
| `429` | 请求过多(超出并发上限 + 等待队列) |
| `429` | 请求过多(超出并发上限 + 等待队列;当前不附带 `Retry-After` |
| `503` | 模型不可用或上游服务异常 |
---

View File

@@ -29,7 +29,7 @@ WORKDIR /app
RUN apt-get update \
&& apt-get install -y --no-install-recommends ca-certificates \
&& groupadd -r ds2api && useradd -r -g ds2api -d /app -s /sbin/nologin ds2api \
&& mkdir -p /app/data && chown -R ds2api:ds2api /app \
&& mkdir -p /app/data /data && chown -R ds2api:ds2api /app /data \
&& rm -rf /var/lib/apt/lists/*
COPY --from=busybox-tools /bin/busybox /usr/local/bin/busybox
EXPOSE 5001

View File

@@ -17,7 +17,7 @@
语言 / Language: [中文](README.MD) | [English](README.en.md)
将 DeepSeek Web 对话能力转换为 OpenAI、Claude 与 Gemini 兼容 API。后端 **Go 全量实现**,前端为 React WebUI 管理台(源码在 `webui/`,部署时自动构建到 `static/admin`)。
将 DeepSeek Web 对话能力转换为 OpenAI、Claude 与 Gemini 兼容 API。核心后端 **Go** 实现Vercel 流式桥接额外使用少量 Node Runtime,前端为 React WebUI 管理台(源码在 `webui/`,部署时自动构建到 `static/admin`)。
文档入口:[文档导航](docs/README.md) / [架构说明](docs/ARCHITECTURE.md) / [接口文档](API.md)
@@ -247,6 +247,7 @@ docker-compose logs -f
默认 `docker-compose.yml` 会把宿主机 `6011` 映射到容器内的 `5001`。如果你希望直接对外暴露 `5001`,请设置 `DS2API_HOST_PORT=5001`(或者手动调整 `ports` 配置)。
同时默认把 `./config.json` 挂载到容器 `/data/config.json`,并设置 `DS2API_CONFIG_PATH=/data/config.json`,用于避免 `/app` 只读导致运行时 token 持久化失败。
镜像会预创建 `/data` 并授权给非 root 的 `ds2api` 用户;如果使用单文件 bind mount请确保宿主机 `config.json` 对容器用户可读写,例如 `chmod 644 config.json`。
更新镜像:`docker-compose up -d --build`
@@ -284,7 +285,7 @@ base64 < config.json | tr -d '\n'
### 方式四:本地源码运行
**前置要求**Go 1.26+Node.js `20.19+` 或 `22.12+`(仅在需要构建 WebUI 时)
**前置要求**Go 1.26+Node.js `20.19+` 或 `22.12+`(仅在需要构建 WebUI 时);同时确保 `npm` 可用,建议 `npm 10+`
```bash
# 1. 克隆仓库
@@ -303,7 +304,7 @@ go run ./cmd/ds2api
服务实际绑定:`0.0.0.0:5001`,因此同一局域网设备通常也可以通过你的内网 IP 访问。
> **WebUI 自动构建**:本地首次启动时,若 `static/admin` 不存在,会自动尝试执行 `npm ci`(仅在缺少依赖时)和 `npm run build -- --outDir static/admin --emptyOutDir`(需要本机有 Node.js。你也可以手动构建`./scripts/build-webui.sh`
> **WebUI 自动构建**:本地首次启动时,若 `static/admin` 不存在,会自动尝试执行 `npm ci`(仅在缺少依赖时)和 `npm run build -- --outDir static/admin --emptyOutDir`(需要本机有 Node.js 和 npm)。你也可以手动构建:`./scripts/build-webui.sh`
## 配置说明
@@ -317,7 +318,7 @@ go run ./cmd/ds2api
- `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`:旧轮次拆分字段,已废弃并忽略,仅保留兼容旧配置。
- `current_input_file`:唯一生效的独立拆分策略;默认开启且阈值为 `0`,触发时将完整上下文合并上传为 `history.txt` 上下文文件。
- `current_input_file`:唯一生效的独立拆分策略;默认开启且阈值为 `0`,触发时将完整上下文合并上传为 `DS2API_HISTORY.txt` 上下文文件。
- 如果关闭 `current_input_file`,请求会直接透传,不上传拆分上下文文件。
- `thinking_injection`:默认开启;在最新 user 消息末尾追加思考增强提示词,提高高强度推理与工具调用前的思考稳定性;`prompt` 留空时使用内置默认提示词。
@@ -333,6 +334,7 @@ go run ./cmd/ds2api
| **直通 token 模式** | 传入 token 不在 `config.keys` 中时,直接作为 DeepSeek token 使用 |
可选请求头 `X-Ds2-Target-Account`:指定使用某个托管账号(值为 email 或 mobile
如果指定账号不存在,或者当前管理账号队列已满,请求会返回 `429`;当前 `429` 不附带 `Retry-After` 头。若账号存在但登录/刷新失败,则返回对应的鉴权错误。
Gemini 路由还可以使用 `x-goog-api-key`,或在没有认证头时使用 `?key=` / `?api_key=` 作为调用方凭据。
## 并发模型
@@ -345,7 +347,7 @@ Gemini 路由还可以使用 `x-goog-api-key`,或在没有认证头时使用 `
```
- 当 in-flight 槽位满时,请求进入等待队列,**不会立即 429**
- 超出总承载上限后才返回 `429 Too Many Requests`
- 超出总承载上限后才返回 `429 Too Many Requests`,当前响应不附带 `Retry-After`
- `GET /admin/queue/status` 返回实时并发状态
## Tool Call 适配
@@ -423,10 +425,10 @@ npm run build --prefix webui
工作流文件:`.github/workflows/release-artifacts.yml`
- **触发条件**:仅在 GitHub Release `published` 时触发(普通 push 不会触发)
- **构建产物**:多平台二进制包(`linux/amd64`、`linux/arm64`、`linux/armv7`、`darwin/amd64`、`darwin/arm64`、`windows/amd64`、`windows/arm64`+ `sha256sums.txt`
- **触发条件**默认仅在 GitHub Release `published` 时自动触发;也支持在 Actions 页面手动 `workflow_dispatch`,并填写 `release_tag` 复跑/补发
- **构建产物**:多平台二进制包(`linux/amd64`、`linux/arm64`、`linux/armv7`、`darwin/amd64`、`darwin/arm64`、`windows/amd64`、`windows/arm64`、Linux Docker 镜像导出包 + `sha256sums.txt`
- **容器镜像发布**:仅推送到 GHCR`ghcr.io/cjackhwang/ds2api`
- **每个压缩包包含**`ds2api` 可执行文件、`static/admin`、WASM 文件(同时支持内置 fallback、`config.example.json` 配置示例、README、LICENSE
- **每个二进制压缩包包含**`ds2api` 可执行文件、`static/admin`、`config.example.json`、`.env.example`、`README.MD`、`README.en.md`、`LICENSE`
## 免责声明

View File

@@ -16,7 +16,7 @@
Language: [中文](README.MD) | [English](README.en.md)
DS2API converts DeepSeek Web chat capability into OpenAI-compatible, Claude-compatible, and Gemini-compatible APIs. The backend is a **pure Go implementation**, with a React WebUI admin panel (source in `webui/`, build output auto-generated to `static/admin` during deployment).
DS2API converts DeepSeek Web chat capability into OpenAI-compatible, Claude-compatible, and Gemini-compatible APIs. The core backend is Go-based, with a small Node Runtime bridge used for Vercel streaming, and the React WebUI admin panel lives in `webui/` (build output auto-generated to `static/admin` during deployment).
Documentation entry: [Docs Index](docs/README.md) / [Architecture](docs/ARCHITECTURE.en.md) / [API Reference](API.en.md)
@@ -306,7 +306,7 @@ Common fields:
- `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`: legacy multi-turn history split field, now ignored and kept only for backward-compatible config loading.
- `current_input_file`: the only active split mode; it is enabled by default and uploads the full context as a `history.txt` context file once the character threshold is reached.
- `current_input_file`: the only active split mode; it is enabled by default and uploads the full context as a `DS2API_HISTORY.txt` context file once the character threshold is reached.
- If you turn off `current_input_file`, requests pass through directly without uploading any split context file.
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).
@@ -409,10 +409,10 @@ npm run build --prefix webui
Workflow: `.github/workflows/release-artifacts.yml`
- **Trigger**: only on GitHub Release `published` (normal pushes do not trigger builds)
- **Outputs**: multi-platform archives (`linux/amd64`, `linux/arm64`, `linux/armv7`, `darwin/amd64`, `darwin/arm64`, `windows/amd64`, `windows/arm64`) + `sha256sums.txt`
- **Trigger**: by default only on GitHub Release `published`; you can also run it manually via `workflow_dispatch` and pass `release_tag` to rerun / backfill
- **Outputs**: multi-platform binary archives (`linux/amd64`, `linux/arm64`, `linux/armv7`, `darwin/amd64`, `darwin/arm64`, `windows/amd64`, `windows/arm64`), Linux Docker image export tarballs, and `sha256sums.txt`
- **Container publishing**: GHCR only (`ghcr.io/cjackhwang/ds2api`)
- **Each archive includes**: `ds2api` executable, `static/admin`, WASM file (with embedded fallback support), `config.example.json`-based config template, README, LICENSE
- **Each binary archive includes**: the `ds2api` executable, `static/admin`, `config.example.json`, `.env.example`, `README.MD`, `README.en.md`, and `LICENSE`
## Disclaimer

View File

@@ -1 +1 @@
4.2.1
4.3.0

View File

@@ -36,7 +36,7 @@ go run ./cmd/ds2api
cd webui
# 2. Install dependencies
npm install
npm ci
# 3. Start dev server (hot reload)
npm run dev

View File

@@ -36,7 +36,7 @@ go run ./cmd/ds2api
cd webui
# 2. 安装依赖
npm install
npm ci
# 3. 启动开发服务器(热更新)
npm run dev

View File

@@ -64,8 +64,8 @@ Use `config.json` as the single source of truth:
Built-in GitHub Actions workflow: `.github/workflows/release-artifacts.yml`
- **Trigger**: only on Release `published` (no build on normal push)
- **Outputs**: multi-platform binary archives + `sha256sums.txt`
- **Trigger**: by default only on Release `published`; you can also run it manually via `workflow_dispatch` and pass `release_tag` to rerun / backfill
- **Outputs**: multi-platform binary archives, Linux Docker image export tarballs, and `sha256sums.txt`
- **Container publishing**: GHCR only (`ghcr.io/cjackhwang/ds2api`)
| Platform | Architecture | Format |
@@ -131,6 +131,7 @@ docker-compose logs -f
The default `docker-compose.yml` directly uses `ghcr.io/cjackhwang/ds2api:latest` and maps host port `6011` to container port `5001`. If you want `5001` exposed directly, set `DS2API_HOST_PORT=5001` (or adjust the `ports` mapping).
The compose template also defaults to `DS2API_CONFIG_PATH=/data/config.json` with `./config.json:/data/config.json` mounted, so deployments avoid read-only `/app` persistence issues by default.
The image pre-creates `/data` and grants it to the non-root `ds2api` user. If you bind-mount a single host file, make sure `config.json` is readable/writable by the container user, for example with `chmod 644 config.json`; otherwise Linux UID/GID mismatches can still cause `open /data/config.json: permission denied`.
Compatibility note: when `DS2API_CONFIG_PATH` is unset and runtime base dir is `/app`, newer versions prefer `/data/config.json`; if that file is missing but legacy `/app/config.json` exists, DS2API automatically falls back to the legacy path to avoid post-upgrade config loss.
If you want a pinned version instead of `latest`, you can also pull a specific tag directly:
@@ -270,6 +271,7 @@ VERCEL_TEAM_ID=team_xxxxxxxxxxxx # optional for personal accounts
| `VERCEL_TOKEN` | Vercel sync token | — |
| `VERCEL_PROJECT_ID` | Vercel project ID | — |
| `VERCEL_TEAM_ID` | Vercel team ID | — |
| `DS2API_CHAT_HISTORY_PATH` | Chat history storage path (must be set to `/tmp/chat_history.json` on Vercel, otherwise unavailable due to read-only filesystem) | `data/chat_history.json` |
| `DS2API_VERCEL_PROTECTION_BYPASS` | Deployment protection bypass for internal Node→Go calls | — |
### 3.4 Vercel Architecture
@@ -359,6 +361,22 @@ 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
#### Chat History Unavailable (read-only file system)
```text
create chat history dir: mkdir /var/task/data: read-only file system
```
**Cause**: Vercel Serverless functions have a read-only filesystem (`/var/task`). Chat history fails because it cannot create directories there.
**Fix**: Add the following in Vercel Project Settings → Environment Variables:
```text
DS2API_CHAT_HISTORY_PATH=/tmp/chat_history.json
```
`/tmp` is the only writable directory in Vercel Serverless. Data is ephemeral (not persisted across cold starts), but the feature works within a single instance lifetime.
### 3.6 Build Artifacts Not Committed
- `static/admin` directory is not in Git
@@ -401,7 +419,7 @@ Or step by step:
```bash
cd webui
npm install
npm ci
npm run build
# Output goes to static/admin/
```

View File

@@ -64,8 +64,8 @@ cp config.example.json config.json
仓库内置 GitHub Actions 工作流:`.github/workflows/release-artifacts.yml`
- **触发条件**:仅在 Release `published`触发(普通 push 不会构建)
- **构建产物**:多平台二进制压缩包 + `sha256sums.txt`
- **触发条件**默认仅在 Release `published`自动触发;也支持在 Actions 页面手动 `workflow_dispatch`,并填写 `release_tag` 复跑/补发
- **构建产物**:多平台二进制压缩包、Linux Docker 镜像导出包 + `sha256sums.txt`
- **容器镜像发布**:仅发布到 GHCR`ghcr.io/cjackhwang/ds2api`
| 平台 | 架构 | 文件格式 |
@@ -131,6 +131,7 @@ docker-compose logs -f
默认 `docker-compose.yml` 直接使用 `ghcr.io/cjackhwang/ds2api:latest`,并把宿主机 `6011` 映射到容器内的 `5001`。如果你希望直接对外暴露 `5001`,请设置 `DS2API_HOST_PORT=5001`(或者手动调整 `ports` 配置)。
Compose 模板还会默认设置 `DS2API_CONFIG_PATH=/data/config.json` 并挂载 `./config.json:/data/config.json`,优先避免 `/app` 只读带来的配置持久化问题。
镜像内会预创建 `/data` 并授权给非 root 的 `ds2api` 用户;如果你使用 bind mount 单文件,请确保宿主机 `config.json` 至少可被容器用户读取/写入,例如 `chmod 644 config.json`,否则 Linux UID/GID 不一致时仍可能出现 `open /data/config.json: permission denied`
兼容说明:若未设置 `DS2API_CONFIG_PATH` 且运行目录是 `/app`,新版本会优先使用 `/data/config.json`;当该文件不存在但检测到历史 `/app/config.json` 时,会自动回退读取旧路径,避免升级后“配置丢失”。
如需固定版本,也可以直接拉取指定 tag
@@ -270,6 +271,7 @@ VERCEL_TEAM_ID=team_xxxxxxxxxxxx # 个人账号可留空
| `VERCEL_TOKEN` | Vercel 同步 token | — |
| `VERCEL_PROJECT_ID` | Vercel 项目 ID | — |
| `VERCEL_TEAM_ID` | Vercel 团队 ID | — |
| `DS2API_CHAT_HISTORY_PATH` | Chat history 存储路径Vercel 上必须设为 `/tmp/chat_history.json`,否则因文件系统只读而不可用) | `data/chat_history.json` |
| `DS2API_VERCEL_PROTECTION_BYPASS` | 部署保护绕过密钥(内部 Node→Go 调用) | — |
### 3.3 运行时行为配置(通过 Admin API 设置)
@@ -369,6 +371,22 @@ No Output Directory named "public" found after the Build completed.
- **方案 B**:请求中添加 `x-vercel-protection-bypass`
- **方案 C**:设置 `VERCEL_AUTOMATION_BYPASS_SECRET`(或 `DS2API_VERCEL_PROTECTION_BYPASS`),仅影响内部 Node→Go 调用
#### Chat History 不可用read-only file system
```text
create chat history dir: mkdir /var/task/data: read-only file system
```
**原因**Vercel Serverless 函数的文件系统(`/var/task`为只读chat history 尝试在该路径下创建目录失败。
**解决**:在 Vercel Project Settings → Environment Variables 中添加:
```text
DS2API_CHAT_HISTORY_PATH=/tmp/chat_history.json
```
`/tmp` 是 Vercel Serverless 环境中唯一可写的目录。数据在函数冷启动之间不会持久化ephemeral但在单个实例生命周期内功能正常。
### 3.6 仓库不提交构建产物
- `static/admin` 目录不在 Git 中
@@ -411,7 +429,7 @@ go run ./cmd/ds2api
```bash
cd webui
npm install
npm ci
npm run build
# 产物输出到 static/admin/
```

View File

@@ -22,7 +22,7 @@
### 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``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。
- 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/**/handler*.go``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。
@@ -53,7 +53,7 @@ Recommended reading order:
### 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`.
- Documentation updates must be grounded in the actual implementation: root routing lives in `internal/server/router.go`, protocol/resource routes live in `internal/httpapi/**/handler*.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.

View File

@@ -60,11 +60,10 @@ npm run build --prefix webui
./tests/scripts/check-refactor-line-gate.sh
./tests/scripts/check-node-split-syntax.sh
./tests/scripts/check-cross-build.sh
# 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
```
说明:`plans/stage6-manual-smoke.md` 已移除,阶段 6 手工烟测不再作为当前 CI 或发布门禁。
### 端到端测试 | End-to-End Tests
```bash

View File

@@ -117,6 +117,11 @@ OpenAI Chat / Responses 在标准化后、current input file 之前,会默认
- 普通请求会直接出现在最终 `prompt` 的最新 user block 末尾。
- 如果触发 current input file它会进入完整上下文文件中。
另外,`MessagesPrepareWithThinking` 还会在最终 prompt 的最前面预置一段固定的 system 级“输出完整性约束Output integrity guard
- 如果上游上下文、工具输出或解析后的文本出现乱码、损坏、部分解析、重复或其他畸形片段,不要模仿、不要回显,只输出给用户的正确内容。
- 这段约束位于普通 system / tool prompt 之前,因此是当前最终 prompt 里的最高优先级前置指令。
### 5.1 角色标记
最终 prompt 使用 DeepSeek 风格角色标记:
@@ -155,7 +160,8 @@ OpenAI Chat / Responses 在标准化后、current input file 之前,会默认
工具调用正例现在优先示范官方 DSML 风格:`<|DSML|tool_calls>``<|DSML|invoke name="...">``<|DSML|parameter name="...">`
兼容层仍接受旧式纯 `<tool_calls>` wrapper但提示词会优先要求模型输出官方 DSML 标签,并强调不能只输出 closing wrapper 而漏掉 opening tag。需要注意这是“兼容 DSML 外壳,内部仍以 XML 解析语义为准”,不是原生 DSML 全链路实现DSML 标签会在解析入口归一化回现有 XML 标签后继续走同一套 parser。
数组参数使用 `<item>...</item>` 子节点表示;当某个参数体只包含 item 子节点时Go / Node 解析器会把它还原成数组,避免 `questions` / `options` 这类 schema 中要求 array 的参数被误解析成 `{ "item": ... }` 对象。若模型把完整结构化 XML fragment 误包进 CDATA兼容层会在保护 `content` / `command` 等原文字段的前提下,尝试把非原文字段中的 CDATA XML fragment 还原成 object / array。不过如果 CDATA 只是单个平面的 XML/HTML 标签,例如 `<b>urgent</b>` 这种行内标记,兼容层会保留原始字符串,不会强行升成 object / array只有明显表示结构的 CDATA 片段,例如多兄弟节点、嵌套子节点或 `item` 列表,才会触发结构化恢复。
数组参数使用 `<item>...</item>` 子节点表示;当某个参数体只包含 item 子节点时Go / Node 解析器会把它还原成数组,避免 `questions` / `options` 这类 schema 中要求 array 的参数被误解析成 `{ "item": ... }` 对象。除此之外,解析器还会回收一些更松散的列表写法,例如 JSON array 字面量或逗号分隔的 JSON 项序列,只要它们足够明确;但 `<item>` 仍然是首选形态。若模型把完整结构化 XML fragment 误包进 CDATA兼容层会在保护 `content` / `command` 等原文字段的前提下,尝试把非原文字段中的 CDATA XML fragment 还原成 object / array。不过如果 CDATA 只是单个平面的 XML/HTML 标签,例如 `<b>urgent</b>` 这种行内标记,兼容层会保留原始字符串,不会强行升成 object / array只有明显表示结构的 CDATA 片段,例如多兄弟节点、嵌套子节点或 `item` 列表,才会触发结构化恢复。
Go 侧读取 DeepSeek SSE 时不再依赖 `bufio.Scanner` 的固定 2MiB 单行上限;当写文件类工具把很长的 `content` 放在单个 `data:` 行里返回时,非流式收集、流式解析和 auto-continue 透传都会保留完整行,再进入同一套工具解析与序列化流程。
在 assistant 最终回包阶段,如果某个 tool 参数在声明 schema 中明确是 `string`,兼容层会在把解析后的 `tool_calls` / `function_call` 重新序列化成 OpenAI / Responses / Claude 可见参数前,递归把该路径上的 number / bool / object / array 统一转成字符串;其中 object / array 会压成紧凑 JSON 字符串。这个保护只对 schema 明确声明为 string 的路径生效,不会改写本来就是 `number` / `boolean` / `object` / `array` 的参数。这样可以兼容 DeepSeek 输出了结构化片段、但上游客户端工具 schema 又严格要求字符串参数的场景(例如 `content``prompt``path``taskId` 等)。
工具 schema 的权威来源始终是**当前请求实际携带的 schema**,而不是同名工具在其他 runtimeClaude Code / OpenCode / Codex 等)里的默认印象。兼容层现在会同时兼容 OpenAI 风格 `function.parameters`、直接工具对象上的 `parameters` / `input_schema`、以及 camelCase 的 `inputSchema` / `schema`,并在最终输出阶段按这份请求内 schema 决定是保留 array/object还是仅对明确声明为 `string` 的路径做字符串化。该规则同样适用于 Claude 的流式收尾和 Vercel Node 流式 tool-call formatter避免不同 runtime 因 schema shape 差异而出现同名工具参数类型漂移。
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。
@@ -237,6 +243,14 @@ OpenAI 文件相关实现:
- 文件 ID 收集:
[internal/promptcompat/file_refs.go](../internal/promptcompat/file_refs.go)
OpenAI 的文件上传现在不再是“只传文件本体”的通用路径,而是会先根据请求里的 `model` 解析出 DeepSeek 的上传类型,并把它透传到上传接口的 `x-model-type`。当前可见的上传类型就是 `default` / `expert` / `vision`,其中 vision 请求上传图片时必须带上 `vision`,否则下游容易退回到仅文本或 OCR 语义。这个模型类型会同时用于:
- `/v1/files` 这类独立文件上传入口
- Chat / Responses 的 inline 图片、附件上传
- current input file 触发时生成的 `DS2API_HISTORY.txt` 上下文文件
也就是说,文件上传和完成请求的 `model_type` 现在是一致的:完成 payload 里仍然是 `model_type`,上传文件则会在 DeepSeek 上传阶段携带同样的模型类型信息。
结论:
- “systemprompt 文字”在 prompt 里
@@ -248,7 +262,7 @@ OpenAI 文件相关实现:
兼容层现在只保留 `current_input_file` 这一种拆分方式;旧的 `history_split` 已废弃,只保留为兼容旧配置的字段,不再参与请求处理。
- `current_input_file` 默认开启;它用于把“完整上下文”合并进 `history.txt` 上下文文件。当最新 user turn 的纯文本长度达到 `current_input_file.min_chars`(默认 `0`)时,兼容层会上传一个文件名为 `history.txt` 的上下文文件,并在 live prompt 中只保留一个中性的 user 消息要求模型直接回答最新请求,不再暴露文件名或要求模型读取本地文件
- `current_input_file` 默认开启;它用于把“完整上下文”合并进 `DS2API_HISTORY.txt` 上下文文件。当最新 user turn 的纯文本长度达到 `current_input_file.min_chars`(默认 `0`)时,兼容层会上传一个文件名为 `DS2API_HISTORY.txt` 的上下文文件。文件内容会先做 OpenAI 消息标准化,再序列化成按轮次编号的 `DS2API_HISTORY.txt` 风格 transcript带有 `# DS2API_HISTORY.txt` 标题和 `=== N. ROLE ===` 分段live prompt 中则会给出一个 continuation 语气的 user 消息,引导模型从 `DS2API_HISTORY.txt` 的最新状态继续推进,并直接回答最新请求,避免把任务拉回起点
- 如果 `current_input_file.enabled=false`,请求会直接透传,不上传任何拆分上下文文件。
- 旧的 `history_split.enabled` / `history_split.trigger_after_turns` 会被读取进配置对象以保持兼容,但不会触发拆分上传,也不会影响 `current_input_file` 的默认开启。
- 即使触发 `current_input_file` 后 live prompt 被缩短,对客户端回包里的上下文 token 统计,仍会沿用**拆分前的完整 prompt 语义**做计数,而不是按缩短后的占位 prompt 计算;否则会把真实上下文显著算小。
@@ -262,11 +276,24 @@ OpenAI 文件相关实现:
- 旧历史拆分兼容壳:
[internal/httpapi/openai/history/history_split.go](../internal/httpapi/openai/history/history_split.go)
当前输入转文件启用并触发时,上传文件的真实文件名是 `history.txt`,文件内容是完整 `messages` 上下文;它仍会先用 OpenAI 消息标准化和 DeepSeek 角色标记序列化,并直接作为 `history.txt` 的纯文本内容上传(不再注入文件边界标签):
当前输入转文件启用并触发时,上传文件的真实文件名是 `DS2API_HISTORY.txt`,文件内容是完整 `messages` 上下文;它仍会先用 OpenAI 消息标准化和 DeepSeek 角色标记序列化,再按轮次编号成 `DS2API_HISTORY.txt` 风格的 transcript(不再注入文件边界标签):
```text
[uploaded filename]: history.txt
<begin▁of▁sentence><System>...<User>...<Assistant>...<Tool>...<User>...
[uploaded filename]: DS2API_HISTORY.txt
# DS2API_HISTORY.txt
Prior conversation history and tool progress.
=== 1. SYSTEM ===
...
=== 2. USER ===
...
=== 3. ASSISTANT ===
...
=== 4. TOOL ===
...
```
开启后,请求的 live prompt 不再直接内联完整上下文,而是保留一个 user role 的短提示,提示模型基于已提供上下文直接回答最新请求;上传后的 `file_id` 会进入 `ref_file_ids`
@@ -318,7 +345,7 @@ OpenAI 文件相关实现:
```json
{
"prompt": "<begin▁of▁sentence><System>原 system / developer\n\nYou have access to these tools: ...<end▁of▁instructions><User>The current request and prior conversation context have already been provided. Answer the latest user request directly.<Assistant>",
"prompt": "<begin▁of▁sentence><System>原 system / developer\n\nYou have access to these tools: ...<end▁of▁instructions><User>Continue from the latest state in the attached DS2API_HISTORY.txt context. Treat it as the current working state and answer the latest user request directly.<Assistant>",
"ref_file_ids": [
"file-current-input-ignore",
"file-systemprompt",
@@ -333,7 +360,7 @@ OpenAI 文件相关实现:
- 大部分结构化语义被压进 `prompt`
- 文件保持文件
- 需要时把完整上下文拆进 `history.txt` 上下文文件
- 需要时把完整上下文拆进 `DS2API_HISTORY.txt` 上下文文件,并按轮次编号成 transcript
## 12. 修改时必须同步本文档的场景
@@ -346,7 +373,7 @@ OpenAI 文件相关实现:
- 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
- current input file 触发条件、上传格式、`history.txt` 包装格式变更
- current input file 触发条件、上传格式、`DS2API_HISTORY.txt` transcript 结构变更
-`history_split` 兼容逻辑的读取、忽略或退化行为变更
- completion payload 字段语义变更
- Claude / Gemini 对这套统一语义的复用关系变更

View File

@@ -26,7 +26,7 @@
</tool_calls>
```
这不是原生 DSML 全链路实现。DSML 只作为 prompt 外壳和解析入口别名;进入 parser 前会归一化成 `<tool_calls>` / `<invoke>` / `<parameter>`,内部仍以现有 XML 解析语义为准。
这不是原生 DSML 全链路实现。DSML 主要用于让模型有意识地输出协议标识,隔离普通 XML 语义;进入 parser 前会按固定本地标签名归一化成 `<tool_calls>` / `<invoke>` / `<parameter>`,内部仍以现有 XML 解析语义为准。
约束:
@@ -39,7 +39,8 @@
兼容修复:
- 如果模型漏掉 opening wrapper但后面仍输出了一个或多个 invoke 并以 closing wrapper 收尾Go 解析链路会在解析前补回缺失的 opening wrapper。
- 如果模型把 DSML 标签里的分隔符 `|` 写漏成空格(例如 `<|DSML tool_calls>` / `<|DSML invoke>` / `<|DSML parameter>`,或无 leading pipe 的 `<DSML tool_calls>` 形态),或把 `DSML` 与工具标签名直接黏连(例如 `<DSMLtool_calls>` / `<DSMLinvoke>` / `<DSMLparameter>`),或把最前面的 pipe 误写成全宽竖线(例如 `<DSML|tool_calls>` / `<DSML|invoke>` / `<DSML|parameter>`Go / Node 会在固定工具标签名范围内归一化;相似但非工具标签名(如 `tool_calls_extra`)仍按普通文本处理。
- Go / Node 解析层不再枚举每一种 DSML typo。它会把工具标签名前的 `DSML`、管道符 `|` / ``、空白、重复 leading `<` 视为可容忍的协议噪声,然后只匹配固定本地标签名 `tool_calls` / `invoke` / `parameter`。例如 `<DSML|tool_calls>``<<|DSML|tool_calls>``<|DSML tool_calls>``<DSMLtool_calls>``<<DSML|DSML|tool_calls>` 都会归一化;相似但非固定标签名(如 `tool_calls_extra`)仍按普通文本处理。
- 如果模型在固定工具标签名后多输出一个尾部管道符,例如 `<|DSML|tool_calls|` / `<|DSML|invoke|` / `<|DSML|parameter|`,兼容层会把这个尾部 `|` 当作异常标签终止符并补齐缺失的 `>`;如果后面已经有 `>`,也会消费这个多余 `|` 后再归一化。
- 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 DSML 外壳。
-`<invoke ...>` / `<parameter ...>` 不会被当成“已支持的工具语法”;只有 `tool_calls` wrapper 或可修复的缺失 opening wrapper 才会进入工具调用路径。
@@ -53,7 +54,7 @@
在流式链路中Go / Node 一致):
- DSML `<|DSML|tool_calls>` wrapper、兼容变体(`<dsml|tool_calls>``<tool_calls>``<|tool_calls>``<DSML|tool_calls>`)、窄容错空格分隔形态(如 `<|DSML tool_calls>`)、黏连形态(如 `<DSMLtool_calls>`)和 canonical `<tool_calls>` wrapper 都会进入结构化捕获
- DSML `<|DSML|tool_calls>` wrapper、基于固定本地标签名的 DSML 噪声容错形态、尾部管道符形态(如 `<|DSML|tool_calls|`)和 canonical `<tool_calls>` wrapper 都会进入结构化捕获
- 如果流里直接从 invoke 开始,但后面补上了 closing wrapperGo 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
@@ -61,6 +62,7 @@
- 支持嵌套围栏(如 4 反引号嵌套 3 反引号)和 CDATA 内围栏保护
- 如果模型把 `<![CDATA[` 打开后却没有闭合,流式扫描阶段仍会保守地继续缓冲,不会误把 CDATA 里的示例 XML 当成真实工具调用;在最终 parse / flush 恢复阶段,会对这类 loose CDATA 做窄修复,尽量保住外层已完整包裹的真实工具调用
- 当文本中 mention 了某种标签名(如 `<dsml|tool_calls>` 或 Markdown inline code 里的 `<|DSML|tool_calls>`而后面紧跟真正工具调用时sieve 会跳过不可解析的 mention 候选并继续匹配后续真实工具块,不会因 mention 导致工具调用丢失,也不会截断 mention 后的正文
- Go 侧 SSE 读取不再使用 `bufio.Scanner` 的固定 token 上限;单个 `data:` 行中包含很长的写文件参数时,非流式收集、流式解析与 auto-continue 透传都应保留完整行,再交给 tool parser 处理
另外,`<parameter>` 的值如果本身是合法 JSON 字面量,也会按结构化值解析,而不是一律保留为字符串。例如 `123``true``null``[1,2]``{"a":1}` 都会还原成对应的 number / boolean / null / array / object。
结构化 XML 参数也会还原为 JSON 结构:如果参数体只包含一个或多个 `<item>...</item>` 子节点,会输出数组;嵌套对象里的 item-only 字段也同样按数组处理。例如 `<parameter name="questions"><item><question>...</question></item></parameter>` 会输出 `{"questions":[{"question":"..."}]}`,而不是 `{"questions":{"item":...}}`
@@ -94,7 +96,7 @@ node --test tests/node/stream-tool-sieve.test.js
- DSML `<|DSML|tool_calls>` wrapper 正常解析
- legacy canonical `<tool_calls>` wrapper 正常解析
- 别名变体(`<dsml|tool_calls>``<tool_calls>``<|tool_calls>`、DSML 空格分隔 typo`<|DSML tool_calls>`)和黏连 typo`<DSMLtool_calls>`)正常解析
- 固定本地标签名的 DSML 噪声容错形态(如 `<DSML|tool_calls>``<<|DSML|tool_calls>``<|DSML tool_calls>``<DSMLtool_calls>``<<DSML|DSML|tool_calls>`)正常解析
- 混搭标签DSML wrapper + canonical inner归一化后正常解析
- 波浪线围栏 `~~~` 内的示例不执行
- 嵌套围栏4 反引号嵌套 3 反引号)内的示例不执行

View File

@@ -133,33 +133,51 @@ func pumpAutoContinue(ctx context.Context, pw *io.PipeWriter, initial io.ReadClo
// sentinels are consumed (not forwarded) so that the downstream only sees
// one final [DONE] at the very end.
func streamBodyWithContinueState(ctx context.Context, pw *io.PipeWriter, body io.Reader, state *continueState) (bool, error) {
scanner := bufio.NewScanner(body)
scanner.Buffer(make([]byte, 0, 64*1024), 2*1024*1024)
reader := bufio.NewReaderSize(body, 64*1024)
hadDone := false
for scanner.Scan() {
for {
select {
case <-ctx.Done():
return hadDone, ctx.Err()
default:
}
line := append([]byte{}, scanner.Bytes()...)
trimmed := strings.TrimSpace(string(line))
if trimmed == "" {
continue
}
if strings.HasPrefix(trimmed, "data:") {
data := strings.TrimSpace(strings.TrimPrefix(trimmed, "data:"))
if data == "[DONE]" {
hadDone = true
continue
line, err := reader.ReadBytes('\n')
if len(line) == 0 && err != nil {
if err == io.EOF {
return hadDone, nil
}
state.observe(data)
return hadDone, err
}
if _, err := io.Copy(pw, bytes.NewReader(append(line, '\n'))); err != nil {
trimmed := strings.TrimSpace(string(line))
if trimmed != "" {
if strings.HasPrefix(trimmed, "data:") {
data := strings.TrimSpace(strings.TrimPrefix(trimmed, "data:"))
if data == "[DONE]" {
hadDone = true
if err != nil && err != io.EOF {
return hadDone, err
}
if err == io.EOF {
return hadDone, nil
}
continue
}
state.observe(data)
}
if !strings.HasSuffix(string(line), "\n") {
line = append(line, '\n')
}
if _, copyErr := io.Copy(pw, bytes.NewReader(line)); copyErr != nil {
return hadDone, copyErr
}
}
if err != nil {
if err == io.EOF {
return hadDone, nil
}
return hadDone, err
}
}
return hadDone, scanner.Err()
}
// observe extracts continue-relevant signals from an SSE JSON chunk.
@@ -175,34 +193,48 @@ func (s *continueState) observe(data string) {
if id := intFrom(chunk["response_message_id"]); id > 0 {
s.responseMessageID = id
}
// Path-based status: {"p": "response/status", "v": "FINISHED"}
if p, _ := chunk["p"].(string); p == "response/status" {
s.setStatus(asString(chunk["v"]))
}
s.observeDirectPatch(asString(chunk["p"]), chunk["v"])
if p, _ := chunk["p"].(string); p == "response" {
s.observeBatchPatches("response", chunk["v"])
} else {
s.observeBatchPatches("", chunk["v"])
}
// Nested v.response
v, _ := chunk["v"].(map[string]any)
if response, _ := v["response"].(map[string]any); response != nil {
if id := intFrom(response["message_id"]); id > 0 {
s.responseMessageID = id
}
s.setStatus(asString(response["status"]))
if autoContinue, ok := response["auto_continue"].(bool); ok && autoContinue {
if v, _ := chunk["v"].(map[string]any); v != nil {
s.observeResponseObject(v["response"])
}
if message, _ := chunk["message"].(map[string]any); message != nil {
s.observeResponseObject(message["response"])
}
}
func (s *continueState) observeDirectPatch(path string, value any) {
if s == nil {
return
}
switch strings.Trim(strings.TrimSpace(path), "/") {
case "response/status", "status", "response/quasi_status", "quasi_status":
s.setStatus(asString(value))
case "response/auto_continue", "auto_continue":
if v, ok := value.(bool); ok && v {
s.lastStatus = "AUTO_CONTINUE"
}
}
// Nested message.response
if message, _ := chunk["message"].(map[string]any); message != nil {
if response, _ := message["response"].(map[string]any); response != nil {
if id := intFrom(response["message_id"]); id > 0 {
s.responseMessageID = id
}
s.setStatus(asString(response["status"]))
}
}
func (s *continueState) observeResponseObject(raw any) {
if s == nil {
return
}
response, _ := raw.(map[string]any)
if response == nil {
return
}
if id := intFrom(response["message_id"]); id > 0 {
s.responseMessageID = id
}
s.setStatus(asString(response["status"]))
if autoContinue, ok := response["auto_continue"].(bool); ok && autoContinue {
s.lastStatus = "AUTO_CONTINUE"
}
}
@@ -230,6 +262,10 @@ func (s *continueState) observeBatchPatches(parentPath string, raw any) {
switch strings.Trim(strings.TrimSpace(fullPath), "/") {
case "response/status", "status", "response/quasi_status", "quasi_status":
s.setStatus(asString(m["v"]))
case "response/auto_continue", "auto_continue":
if v, ok := m["v"].(bool); ok && v {
s.lastStatus = "AUTO_CONTINUE"
}
}
}
}

View File

@@ -150,6 +150,62 @@ func TestAutoContinueDoesNotTriggerOnPlainWIPWithoutExplicitContinuationSignal(t
}
}
func TestAutoContinuePassesThroughLongSingleSSELine(t *testing.T) {
payload := strings.Repeat("x", 2*1024*1024+4096)
initialBody := `data: {"p":"response/content","v":"` + payload + `"}` + "\n" +
`data: [DONE]` + "\n"
body := newAutoContinueBody(context.Background(), io.NopCloser(strings.NewReader(initialBody)), "session-123", 8, func(context.Context, string, int) (*http.Response, error) {
return nil, errors.New("continue should not have been called")
})
defer func() { _ = body.Close() }()
out, err := io.ReadAll(body)
if err != nil {
t.Fatalf("read body failed: %v", err)
}
if !bytes.Contains(out, []byte(payload)) {
t.Fatalf("expected long SSE payload to pass through, got len=%d want payload len=%d", len(out), len(payload))
}
if !bytes.Contains(out, []byte(`data: [DONE]`)) {
t.Fatalf("expected final DONE sentinel in body, got len=%d", len(out))
}
}
func TestAutoContinueTriggersOnDirectQuasiStatusIncomplete(t *testing.T) {
initialBody := strings.Join([]string{
`data: {"response_message_id":321,"p":"response/content","v":"<tool_calls><invoke name=\"write_file\"><parameter name=\"content\"><![CDATA[part-one"}`,
`data: {"p":"response/quasi_status","v":"INCOMPLETE"}`,
`data: [DONE]`,
}, "\n") + "\n"
var continueCalls atomic.Int32
body := newAutoContinueBody(context.Background(), io.NopCloser(strings.NewReader(initialBody)), "session-123", 8, func(context.Context, string, int) (*http.Response, error) {
continueCalls.Add(1)
return &http.Response{
StatusCode: http.StatusOK,
Header: make(http.Header),
Body: io.NopCloser(strings.NewReader(
`data: {"response_message_id":322,"p":"response/content","v":"-part-two]]></parameter></invoke></tool_calls>"}` + "\n" +
`data: {"p":"response/status","v":"FINISHED"}` + "\n" +
`data: [DONE]` + "\n",
)),
}, nil
})
defer func() { _ = body.Close() }()
out, err := io.ReadAll(body)
if err != nil {
t.Fatalf("read body failed: %v", err)
}
if continueCalls.Load() != 1 {
t.Fatalf("expected exactly one continue call, got %d", continueCalls.Load())
}
if !bytes.Contains(out, []byte("part-one")) || !bytes.Contains(out, []byte("-part-two")) {
t.Fatalf("expected continued tool content in body, got=%s", string(out))
}
}
func TestAutoContinueTriggersOnResponseBatchQuasiStatusIncomplete(t *testing.T) {
initialBody := strings.Join([]string{
`data: {"response_message_id":321,"v":{"response":{"message_id":321,"status":"WIP","auto_continue":false}}}`,

View File

@@ -23,6 +23,7 @@ type UploadFileRequest struct {
Filename string
ContentType string
Purpose string
ModelType string
Data []byte
}
@@ -54,6 +55,7 @@ func (c *Client) UploadFile(ctx context.Context, a *auth.RequestAuth, req Upload
contentType = "application/octet-stream"
}
purpose := strings.TrimSpace(req.Purpose)
modelType := strings.ToLower(strings.TrimSpace(req.ModelType))
body, contentTypeHeader, err := buildUploadMultipartBody(filename, contentType, req.Data)
if err != nil {
return nil, err
@@ -64,6 +66,9 @@ func (c *Client) UploadFile(ctx context.Context, a *auth.RequestAuth, req Upload
"purpose": purpose,
"bytes": len(req.Data),
}
if modelType != "" {
capturePayload["model_type"] = modelType
}
captureSession := c.capture.Start("deepseek_upload_file", dsprotocol.DeepSeekUploadFileURL, a.AccountID, capturePayload)
attempts := 0
refreshed := false
@@ -81,6 +86,9 @@ func (c *Client) UploadFile(ctx context.Context, a *auth.RequestAuth, req Upload
}
headers := c.authHeaders(a.DeepSeekToken)
headers["Content-Type"] = contentTypeHeader
if modelType != "" {
headers["x-model-type"] = modelType
}
headers["x-ds-pow-response"] = powHeader
headers["x-file-size"] = strconv.Itoa(len(req.Data))
headers["x-thinking-enabled"] = "1"

View File

@@ -82,6 +82,7 @@ func TestUploadFileUsesUploadTargetPowAndMultipartHeaders(t *testing.T) {
var seenTargetPath string
var seenContentType string
var seenFileSize string
var seenModelType string
var seenBody string
call := 0
client := &Client{
@@ -96,6 +97,7 @@ func TestUploadFileUsesUploadTargetPowAndMultipartHeaders(t *testing.T) {
seenPow = req.Header.Get("x-ds-pow-response")
seenContentType = req.Header.Get("Content-Type")
seenFileSize = req.Header.Get("x-file-size")
seenModelType = req.Header.Get("x-model-type")
seenBody = string(bodyBytes)
return &http.Response{StatusCode: http.StatusOK, Header: make(http.Header), Body: io.NopCloser(strings.NewReader(uploadResponse)), Request: req}, nil
default:
@@ -112,6 +114,7 @@ func TestUploadFileUsesUploadTargetPowAndMultipartHeaders(t *testing.T) {
Filename: "demo.txt",
ContentType: "text/plain",
Purpose: "assistants",
ModelType: "vision",
Data: []byte("hello"),
}, 1)
if err != nil {
@@ -140,6 +143,9 @@ func TestUploadFileUsesUploadTargetPowAndMultipartHeaders(t *testing.T) {
if seenFileSize != "5" {
t.Fatalf("expected x-file-size=5, got %q", seenFileSize)
}
if seenModelType != "vision" {
t.Fatalf("expected x-model-type=vision, got %q", seenModelType)
}
if !strings.HasPrefix(seenContentType, "multipart/form-data; boundary=") {
t.Fatalf("expected multipart content type, got %q", seenContentType)
}

View File

@@ -159,6 +159,6 @@ func toStringSet(in []string) map[string]struct{} {
const (
KeepAliveTimeout = 5
StreamIdleTimeout = 90
MaxKeepaliveCount = 10
StreamIdleTimeout = 300
MaxKeepaliveCount = 40
)

View File

@@ -2,7 +2,7 @@
"client": {
"name": "DeepSeek",
"platform": "android",
"version": "2.0.3",
"version": "2.0.4",
"android_api_level": "35",
"locale": "zh_CN"
},
@@ -24,4 +24,4 @@
"skip_exact_paths": [
"response/search_status"
]
}
}

View File

@@ -2,20 +2,24 @@ package protocol
import (
"bufio"
"io"
"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
reader := bufio.NewReaderSize(resp.Body, 64*1024)
for {
line, err := reader.ReadBytes('\n')
if len(line) > 0 {
if !onLine(line) {
return nil
}
}
if err != nil {
if err == io.EOF {
return nil
}
return err
}
}
if err := scanner.Err(); err != nil {
return err
}
return nil
}

View File

@@ -0,0 +1,26 @@
package protocol
import (
"io"
"net/http"
"strings"
"testing"
)
func TestScanSSELinesHandlesLongSingleLine(t *testing.T) {
payload := strings.Repeat("x", 2*1024*1024+4096)
body := "data: {\"p\":\"response/content\",\"v\":\"" + payload + "\"}\n"
resp := &http.Response{Body: io.NopCloser(strings.NewReader(body))}
var got string
err := ScanSSELines(resp, func(line []byte) bool {
got = string(line)
return true
})
if err != nil {
t.Fatalf("ScanSSELines returned error: %v", err)
}
if !strings.Contains(got, payload) {
t.Fatalf("long SSE line was not preserved: got len=%d want payload len=%d", len(got), len(payload))
}
}

View File

@@ -3,12 +3,14 @@ package claude
import (
"bytes"
"encoding/json"
"errors"
"io"
"net/http"
"net/http/httptest"
"strings"
"ds2api/internal/config"
"ds2api/internal/httpapi/requestbody"
streamengine "ds2api/internal/stream"
"ds2api/internal/translatorcliproxy"
"ds2api/internal/util"
@@ -33,7 +35,11 @@ func (h *Handler) Messages(w http.ResponseWriter, r *http.Request) {
func (h *Handler) proxyViaOpenAI(w http.ResponseWriter, r *http.Request, store ConfigReader) bool {
raw, err := io.ReadAll(r.Body)
if err != nil {
writeClaudeError(w, http.StatusBadRequest, "invalid body")
if errors.Is(err, requestbody.ErrInvalidUTF8Body) {
writeClaudeError(w, http.StatusBadRequest, "invalid json")
} else {
writeClaudeError(w, http.StatusBadRequest, "invalid body")
}
return true
}
var req map[string]any

View File

@@ -2,8 +2,8 @@ package gemini
import (
"bytes"
"ds2api/internal/toolcall"
"encoding/json"
"errors"
"io"
"net/http"
"net/http/httptest"
@@ -11,7 +11,9 @@ import (
"github.com/go-chi/chi/v5"
"ds2api/internal/httpapi/requestbody"
"ds2api/internal/sse"
"ds2api/internal/toolcall"
"ds2api/internal/translatorcliproxy"
"ds2api/internal/util"
@@ -32,7 +34,11 @@ func (h *Handler) handleGenerateContent(w http.ResponseWriter, r *http.Request,
func (h *Handler) proxyViaOpenAI(w http.ResponseWriter, r *http.Request, stream bool) bool {
raw, err := io.ReadAll(r.Body)
if err != nil {
writeGeminiError(w, http.StatusBadRequest, "invalid body")
if errors.Is(err, requestbody.ErrInvalidUTF8Body) {
writeGeminiError(w, http.StatusBadRequest, "invalid json")
} else {
writeGeminiError(w, http.StatusBadRequest, "invalid body")
}
return true
}
routeModel := strings.TrimSpace(chi.URLParam(r, "model"))

View File

@@ -311,16 +311,16 @@ func TestChatCompletionsCurrentInputFilePersistsNeutralPrompt(t *testing.T) {
if len(ds.uploadCalls) != 1 {
t.Fatalf("expected current input upload to happen, got %d", len(ds.uploadCalls))
}
if ds.uploadCalls[0].Filename != "history.txt" {
t.Fatalf("expected history.txt upload, got %q", ds.uploadCalls[0].Filename)
if ds.uploadCalls[0].Filename != "DS2API_HISTORY.txt" {
t.Fatalf("expected DS2API_HISTORY.txt upload, got %q", ds.uploadCalls[0].Filename)
}
if full.HistoryText != string(ds.uploadCalls[0].Data) {
t.Fatalf("expected uploaded current input file to be persisted in history text")
}
if len(full.Messages) != 1 {
t.Fatalf("expected neutral prompt to be the only persisted message, got %#v", full.Messages)
t.Fatalf("expected continuation prompt to be the only persisted message, got %#v", full.Messages)
}
if !strings.Contains(full.Messages[0].Content, "Answer the latest user request directly.") {
t.Fatalf("expected neutral prompt to be persisted, got %#v", full.Messages[0])
if !strings.Contains(full.Messages[0].Content, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") {
t.Fatalf("expected continuation prompt to be persisted, got %#v", full.Messages[0])
}
}

View File

@@ -173,6 +173,15 @@ func (s *chatStreamRuntime) sendFailedChunk(status int, message, code string) {
s.sendDone()
}
func (s *chatStreamRuntime) markContextCancelled() {
s.finalErrorStatus = 499
s.finalErrorMessage = "request context cancelled"
s.finalErrorCode = string(streamengine.StopReasonContextCancelled)
s.finalThinking = s.thinking.String()
s.finalText = cleanVisibleOutput(s.text.String(), s.stripReferenceMarkers)
s.finalFinishReason = string(streamengine.StopReasonContextCancelled)
}
func (s *chatStreamRuntime) resetStreamToolCallState() {
s.streamToolCallIDs = map[int]string{}
s.streamToolNames = map[int]string{}

View File

@@ -247,11 +247,15 @@ func (h *Handler) consumeChatStreamAttempt(r *http.Request, resp *http.Response,
}
},
OnContextDone: func() {
streamRuntime.markContextCancelled()
if historySession != nil {
historySession.stopped(streamRuntime.thinking.String(), streamRuntime.text.String(), string(streamengine.StopReasonContextCancelled))
}
},
})
if streamRuntime.finalErrorCode == string(streamengine.StopReasonContextCancelled) {
return true, false
}
terminalWritten := streamRuntime.finalize(finalReason, allowDeferEmpty && finalReason != "content_filter")
if terminalWritten {
recordChatStreamHistory(streamRuntime, historySession)
@@ -283,6 +287,10 @@ func logChatStreamTerminal(streamRuntime *chatStreamRuntime, attempts int) {
if attempts > 0 {
source = "synthetic_retry"
}
if streamRuntime.finalErrorCode == string(streamengine.StopReasonContextCancelled) {
config.Logger.Info("[openai_empty_retry] terminal cancelled", "surface", "chat.completions", "stream", true, "retry_attempts", attempts, "error_code", streamRuntime.finalErrorCode)
return
}
if streamRuntime.finalErrorMessage != "" {
config.Logger.Info("[openai_empty_retry] terminal empty output", "surface", "chat.completions", "stream", true, "retry_attempts", attempts, "success_source", "none", "error_code", streamRuntime.finalErrorCode)
return

View File

@@ -0,0 +1,85 @@
package chat
import (
"context"
"net/http"
"net/http/httptest"
"testing"
"time"
"ds2api/internal/chathistory"
"ds2api/internal/stream"
)
func TestConsumeChatStreamAttemptMarksContextCancelledState(t *testing.T) {
historyStore := newTestChatHistoryStore(t)
entry, err := historyStore.Start(chathistory.StartParams{
CallerID: "caller:test",
Model: "deepseek-v4-flash",
Stream: true,
UserInput: "hello",
})
if err != nil {
t.Fatalf("start history failed: %v", err)
}
session := &chatHistorySession{
store: historyStore,
entryID: entry.ID,
startedAt: time.Now(),
lastPersist: time.Now(),
finalPrompt: "prompt",
}
ctx, cancel := context.WithCancel(context.Background())
cancel()
req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", nil).WithContext(ctx)
rec := httptest.NewRecorder()
streamRuntime := newChatStreamRuntime(
rec,
http.NewResponseController(rec),
true,
"cid-cancelled",
time.Now().Unix(),
"deepseek-v4-flash",
"prompt",
false,
false,
true,
nil,
nil,
false,
false,
)
resp := makeOpenAISSEHTTPResponse(
`data: {"p":"response/content","v":"hello"}`,
`data: [DONE]`,
)
h := &Handler{}
terminalWritten, retryable := h.consumeChatStreamAttempt(req, resp, streamRuntime, "text", false, session, true)
if !terminalWritten || retryable {
t.Fatalf("expected cancelled attempt to terminate without retry, got terminalWritten=%v retryable=%v", terminalWritten, retryable)
}
if got, want := streamRuntime.finalErrorCode, string(stream.StopReasonContextCancelled); got != want {
t.Fatalf("expected cancelled final error code %q, got %q", want, got)
}
if streamRuntime.finalErrorMessage == "" {
t.Fatalf("expected cancelled final error message to be preserved")
}
snapshot, err := historyStore.Snapshot()
if err != nil {
t.Fatalf("snapshot failed: %v", err)
}
if len(snapshot.Items) != 1 {
t.Fatalf("expected one history item, got %d", len(snapshot.Items))
}
full, err := historyStore.Get(snapshot.Items[0].ID)
if err != nil {
t.Fatalf("get detail failed: %v", err)
}
if full.Status != "stopped" {
t.Fatalf("expected stopped status, got %#v", full)
}
}

View File

@@ -130,8 +130,8 @@ func TestHandleVercelStreamPrepareAppliesCurrentInputFile(t *testing.T) {
t.Fatalf("expected payload object, got %#v", body["payload"])
}
promptText, _ := payload["prompt"].(string)
if !strings.Contains(promptText, "Answer the latest user request directly.") {
t.Fatalf("expected neutral prompt, got %s", promptText)
if !strings.Contains(promptText, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") {
t.Fatalf("expected continuation prompt, got %s", promptText)
}
if strings.Contains(promptText, "first user turn") || strings.Contains(promptText, "latest user turn") {
t.Fatalf("expected original turns hidden from prompt, got %s", promptText)

View File

@@ -94,6 +94,9 @@ func TestPreprocessInlineFileInputsReplacesDataURLAndCollectsRefFileIDs(t *testi
if len(ds.uploadCalls) != 1 {
t.Fatalf("expected 1 upload, got %d", len(ds.uploadCalls))
}
if ds.uploadCalls[0].ModelType != "default" {
t.Fatalf("expected default model type when request omits model, got %q", ds.uploadCalls[0].ModelType)
}
if ds.lastCtx != ctx {
t.Fatalf("expected upload to use request context")
}
@@ -149,7 +152,7 @@ func TestPreprocessInlineFileInputsDeduplicatesIdenticalPayloads(t *testing.T) {
func TestChatCompletionsUploadsInlineFilesBeforeCompletion(t *testing.T) {
ds := &inlineUploadDSStub{}
h := &openAITestSurface{Store: mockOpenAIConfig{wideInput: true}, Auth: streamStatusAuthStub{}, DS: ds}
reqBody := `{"model":"deepseek-v4-flash","messages":[{"role":"user","content":[{"type":"input_text","text":"hi"},{"type":"image_url","image_url":{"url":"data:image/png;base64,QUJDRA=="}}]}],"stream":false}`
reqBody := `{"model":"deepseek-v4-vision","messages":[{"role":"user","content":[{"type":"input_text","text":"hi"},{"type":"image_url","image_url":{"url":"data:image/png;base64,QUJDRA=="}}]}],"stream":false}`
req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", strings.NewReader(reqBody))
req.Header.Set("Authorization", "Bearer direct-token")
req.Header.Set("Content-Type", "application/json")
@@ -163,6 +166,9 @@ func TestChatCompletionsUploadsInlineFilesBeforeCompletion(t *testing.T) {
if len(ds.uploadCalls) != 1 {
t.Fatalf("expected 1 upload call, got %d", len(ds.uploadCalls))
}
if ds.uploadCalls[0].ModelType != "vision" {
t.Fatalf("expected vision model type for vision request, got %q", ds.uploadCalls[0].ModelType)
}
if ds.completionReq == nil {
t.Fatal("expected completion payload to be captured")
}
@@ -177,7 +183,7 @@ func TestResponsesUploadsInlineFilesBeforeCompletion(t *testing.T) {
h := &openAITestSurface{Store: mockOpenAIConfig{wideInput: true}, Auth: streamStatusAuthStub{}, DS: ds}
r := chi.NewRouter()
registerOpenAITestRoutes(r, h)
reqBody := `{"model":"deepseek-v4-flash","input":[{"role":"user","content":[{"type":"input_text","text":"hi"},{"type":"input_image","image_url":{"url":"data:image/png;base64,QUJDRA=="}}]}],"stream":false}`
reqBody := `{"model":"deepseek-v4-pro","input":[{"role":"user","content":[{"type":"input_text","text":"hi"},{"type":"input_image","image_url":{"url":"data:image/png;base64,QUJDRA=="}}]}],"stream":false}`
req := httptest.NewRequest(http.MethodPost, "/v1/responses", strings.NewReader(reqBody))
req.Header.Set("Authorization", "Bearer direct-token")
req.Header.Set("Content-Type", "application/json")
@@ -191,6 +197,9 @@ func TestResponsesUploadsInlineFilesBeforeCompletion(t *testing.T) {
if len(ds.uploadCalls) != 1 {
t.Fatalf("expected 1 upload call, got %d", len(ds.uploadCalls))
}
if ds.uploadCalls[0].ModelType != "expert" {
t.Fatalf("expected expert model type for pro request, got %q", ds.uploadCalls[0].ModelType)
}
refIDs, _ := ds.completionReq["ref_file_ids"].([]any)
if len(refIDs) != 1 || refIDs[0] != "file-inline-1" {
t.Fatalf("unexpected completion ref_file_ids: %#v", ds.completionReq["ref_file_ids"])

View File

@@ -12,6 +12,7 @@ import (
"strings"
"ds2api/internal/auth"
"ds2api/internal/config"
dsclient "ds2api/internal/deepseek/client"
"ds2api/internal/httpapi/openai/shared"
"ds2api/internal/promptcompat"
@@ -42,6 +43,7 @@ type inlineUploadState struct {
ctx context.Context
handler *Handler
auth *auth.RequestAuth
modelType string
uploadedByID map[string]string
uploadCount int
inlineFileBytes int
@@ -58,10 +60,19 @@ func (h *Handler) PreprocessInlineFileInputs(ctx context.Context, a *auth.Reques
if h == nil || h.DS == nil || len(req) == 0 {
return nil
}
modelType := "default"
if requestedModel, ok := req["model"].(string); ok {
if resolvedModel, ok := config.ResolveModel(h.Store, requestedModel); ok {
if resolvedType, ok := config.GetModelType(resolvedModel); ok {
modelType = resolvedType
}
}
}
state := &inlineUploadState{
ctx: ctx,
handler: h,
auth: a,
modelType: modelType,
uploadedByID: map[string]string{},
}
for _, key := range []string{"messages", "input", "attachments"} {
@@ -174,6 +185,7 @@ func (s *inlineUploadState) uploadInlineFile(file inlineDecodedFile) (string, er
result, err := s.handler.DS.UploadFile(s.ctx, s.auth, dsclient.UploadFileRequest{
Filename: file.Filename,
ContentType: contentType,
ModelType: s.modelType,
Data: file.Data,
}, 3)
if err != nil {

View File

@@ -8,6 +8,7 @@ import (
"ds2api/internal/auth"
"ds2api/internal/chathistory"
"ds2api/internal/config"
dsclient "ds2api/internal/deepseek/client"
"ds2api/internal/httpapi/openai/shared"
)
@@ -66,10 +67,12 @@ func (h *Handler) UploadFile(w http.ResponseWriter, r *http.Request) {
if contentType == "" && len(data) > 0 {
contentType = http.DetectContentType(data)
}
modelType := resolveUploadModelType(h.Store, r)
result, err := h.DS.UploadFile(r.Context(), a, dsclient.UploadFileRequest{
Filename: header.Filename,
ContentType: contentType,
Purpose: strings.TrimSpace(r.FormValue("purpose")),
ModelType: modelType,
Data: data,
}, 3)
if err != nil {
@@ -82,6 +85,32 @@ func (h *Handler) UploadFile(w http.ResponseWriter, r *http.Request) {
shared.WriteJSON(w, http.StatusOK, buildOpenAIFileObject(result))
}
func resolveUploadModelType(store shared.ConfigReader, r *http.Request) string {
for _, candidate := range []string{r.FormValue("model_type"), r.Header.Get("X-Model-Type")} {
if modelType := normalizeUploadModelType(candidate); modelType != "" {
return modelType
}
}
requestedModel := strings.TrimSpace(r.FormValue("model"))
if requestedModel != "" {
if resolvedModel, ok := config.ResolveModel(store, requestedModel); ok {
if modelType, ok := config.GetModelType(resolvedModel); ok {
return modelType
}
}
}
return "default"
}
func normalizeUploadModelType(raw string) string {
switch strings.ToLower(strings.TrimSpace(raw)) {
case "default", "expert", "vision":
return strings.ToLower(strings.TrimSpace(raw))
default:
return ""
}
}
func buildOpenAIFileObject(result *dsclient.UploadFileResult) map[string]any {
if result == nil {
obj := map[string]any{

View File

@@ -77,7 +77,7 @@ func (m *filesRouteDSStub) DeleteAllSessionsForToken(_ context.Context, _ string
return nil
}
func newMultipartUploadRequest(t *testing.T, purpose string, filename string, data []byte) *http.Request {
func newMultipartUploadRequest(t *testing.T, purpose string, filename string, data []byte, model string) *http.Request {
t.Helper()
var body bytes.Buffer
writer := multipart.NewWriter(&body)
@@ -86,6 +86,11 @@ func newMultipartUploadRequest(t *testing.T, purpose string, filename string, da
t.Fatalf("write purpose failed: %v", err)
}
}
if model != "" {
if err := writer.WriteField("model", model); err != nil {
t.Fatalf("write model failed: %v", err)
}
}
part, err := writer.CreateFormFile("file", filename)
if err != nil {
t.Fatalf("create form file failed: %v", err)
@@ -108,7 +113,7 @@ func TestFilesRouteUploadSuccess(t *testing.T) {
r := chi.NewRouter()
registerOpenAITestRoutes(r, h)
req := newMultipartUploadRequest(t, "assistants", "notes.txt", []byte("hello world"))
req := newMultipartUploadRequest(t, "assistants", "notes.txt", []byte("hello world"), "deepseek-v4-vision")
rec := httptest.NewRecorder()
r.ServeHTTP(rec, req)
@@ -121,6 +126,9 @@ func TestFilesRouteUploadSuccess(t *testing.T) {
if ds.lastReq.Purpose != "assistants" {
t.Fatalf("expected purpose assistants, got %q", ds.lastReq.Purpose)
}
if ds.lastReq.ModelType != "vision" {
t.Fatalf("expected vision model type, got %q", ds.lastReq.ModelType)
}
if string(ds.lastReq.Data) != "hello world" {
t.Fatalf("unexpected uploaded data: %q", string(ds.lastReq.Data))
}
@@ -145,7 +153,7 @@ func TestFilesRouteUploadIncludesAccountIDForManagedAccount(t *testing.T) {
r := chi.NewRouter()
registerOpenAITestRoutes(r, h)
req := newMultipartUploadRequest(t, "assistants", "notes.txt", []byte("hello world"))
req := newMultipartUploadRequest(t, "assistants", "notes.txt", []byte("hello world"), "deepseek-v4-vision")
rec := httptest.NewRecorder()
r.ServeHTTP(rec, req)

View File

@@ -7,6 +7,7 @@ import (
"strings"
"ds2api/internal/auth"
"ds2api/internal/config"
dsclient "ds2api/internal/deepseek/client"
"ds2api/internal/httpapi/openai/shared"
"ds2api/internal/promptcompat"
@@ -35,10 +36,15 @@ func (s Service) ApplyCurrentInputFile(ctx context.Context, a *auth.RequestAuth,
if strings.TrimSpace(fileText) == "" {
return stdReq, errors.New("current user input file produced empty transcript")
}
modelType := "default"
if resolvedType, ok := config.GetModelType(stdReq.ResolvedModel); ok {
modelType = resolvedType
}
result, err := s.DS.UploadFile(ctx, a, dsclient.UploadFileRequest{
Filename: currentInputFilename,
ContentType: currentInputContentType,
Purpose: currentInputPurpose,
ModelType: modelType,
Data: []byte(fileText),
}, 3)
if err != nil {
@@ -62,7 +68,7 @@ func (s Service) ApplyCurrentInputFile(ctx context.Context, a *auth.RequestAuth,
stdReq.RefFileIDs = prependUniqueRefFileID(stdReq.RefFileIDs, fileID)
stdReq.FinalPrompt, stdReq.ToolNames = promptcompat.BuildOpenAIPrompt(messages, stdReq.ToolsRaw, "", stdReq.ToolChoice, stdReq.Thinking)
// Token accounting must reflect the actual downstream context:
// the uploaded history.txt file content + the neutral live prompt.
// the uploaded DS2API_HISTORY.txt file content + the continuation live prompt.
stdReq.PromptTokenText = fileText + "\n" + stdReq.FinalPrompt
return stdReq, nil
}
@@ -87,5 +93,5 @@ func latestUserInputForFile(messages []any) (int, string) {
}
func currentInputFilePrompt() string {
return "The current request and prior conversation context have already been provided. Answer the latest user request directly."
return "Continue from the latest state in the attached DS2API_HISTORY.txt context. Treat it as the current working state and answer the latest user request directly."
}

View File

@@ -61,26 +61,33 @@ func (streamStatusManagedAuthStub) DetermineCaller(_ *http.Request) (*auth.Reque
func (streamStatusManagedAuthStub) Release(_ *auth.RequestAuth) {}
func TestBuildOpenAICurrentInputContextTranscriptUsesInjectedFileWrapper(t *testing.T) {
func TestBuildOpenAICurrentInputContextTranscriptUsesNumberedHistorySections(t *testing.T) {
_, historyMessages := splitOpenAIHistoryMessages(historySplitTestMessages(), 1)
transcript := buildOpenAICurrentInputContextTranscript(historyMessages)
if strings.Contains(transcript, "[file content end]") || strings.Contains(transcript, "[file content begin]") || strings.Contains(transcript, "[file name]:") {
t.Fatalf("expected plain transcript without file wrapper tags, got %q", transcript)
t.Fatalf("expected transcript without file wrapper tags, got %q", transcript)
}
if !strings.Contains(transcript, "<begin▁of▁sentence>") {
t.Fatalf("expected serialized conversation markers, got %q", transcript)
if !strings.Contains(transcript, "# DS2API_HISTORY.txt") {
t.Fatalf("expected history transcript header, got %q", transcript)
}
if !strings.Contains(transcript, "first user turn") || !strings.Contains(transcript, "tool result") {
t.Fatalf("expected historical turns preserved, got %q", transcript)
if !strings.Contains(transcript, "Prior conversation history and tool progress.") {
t.Fatalf("expected history transcript description, got %q", transcript)
}
if !strings.Contains(transcript, "[reasoning_content]") || !strings.Contains(transcript, "hidden reasoning") {
t.Fatalf("expected reasoning block preserved, got %q", transcript)
for _, want := range []string{
"=== 1. USER ===",
"=== 2. ASSISTANT ===",
"=== 3. TOOL ===",
"first user turn",
"tool result",
"[reasoning_content]",
"hidden reasoning",
"<|DSML|tool_calls>",
} {
if !strings.Contains(transcript, want) {
t.Fatalf("expected transcript to contain %q, got %q", want, transcript)
}
}
if !strings.Contains(transcript, "<|DSML|tool_calls>") {
t.Fatalf("expected tool calls preserved, got %q", transcript)
}
}
func TestSplitOpenAIHistoryMessagesUsesLatestUserTurn(t *testing.T) {
@@ -220,7 +227,7 @@ func TestApplyCurrentInputFileDisabledPassThrough(t *testing.T) {
DS: ds,
}
req := map[string]any{
"model": "deepseek-v4-flash",
"model": "deepseek-v4-vision",
"messages": historySplitTestMessages(),
}
stdReq, err := promptcompat.NormalizeOpenAIChatRequest(h.Store, req, "")
@@ -243,7 +250,7 @@ func TestApplyCurrentInputFileDisabledPassThrough(t *testing.T) {
}
}
func TestApplyCurrentInputFileUploadsFirstTurnWithInjectedWrapper(t *testing.T) {
func TestApplyCurrentInputFileUploadsFirstTurnWithNumberedHistoryTranscript(t *testing.T) {
ds := &inlineUploadDSStub{}
h := &openAITestSurface{
Store: mockOpenAIConfig{
@@ -273,15 +280,21 @@ func TestApplyCurrentInputFileUploadsFirstTurnWithInjectedWrapper(t *testing.T)
t.Fatalf("expected 1 current input upload, got %d", len(ds.uploadCalls))
}
upload := ds.uploadCalls[0]
if upload.Filename != "history.txt" {
if upload.Filename != "DS2API_HISTORY.txt" {
t.Fatalf("unexpected upload filename: %q", upload.Filename)
}
uploadedText := string(upload.Data)
if strings.Contains(uploadedText, "[file content end]") || strings.Contains(uploadedText, "[file content begin]") || strings.Contains(uploadedText, "[file name]:") {
t.Fatalf("expected uploaded transcript without file wrapper tags, got %q", uploadedText)
}
if !strings.Contains(uploadedText, "<begin▁of▁sentence><User>first turn content that is long enough") {
t.Fatalf("expected serialized current user turn markers, got %q", uploadedText)
for _, want := range []string{
"# DS2API_HISTORY.txt",
"=== 1. USER ===",
"first turn content that is long enough",
} {
if !strings.Contains(uploadedText, want) {
t.Fatalf("expected uploaded transcript to contain %q, got %q", want, uploadedText)
}
}
if !strings.Contains(uploadedText, promptcompat.ThinkingInjectionMarker) {
t.Fatalf("expected thinking injection in current input file, got %q", uploadedText)
@@ -290,11 +303,11 @@ func TestApplyCurrentInputFileUploadsFirstTurnWithInjectedWrapper(t *testing.T)
if strings.Contains(out.FinalPrompt, "first turn content that is long enough") {
t.Fatalf("expected current input text to be replaced in live prompt, got %s", out.FinalPrompt)
}
if strings.Contains(out.FinalPrompt, "CURRENT_USER_INPUT.txt") || strings.Contains(out.FinalPrompt, "history.txt") || strings.Contains(out.FinalPrompt, "Read that file") {
if strings.Contains(out.FinalPrompt, "CURRENT_USER_INPUT.txt") || strings.Contains(out.FinalPrompt, "Read that file") {
t.Fatalf("expected live prompt not to instruct file reads, got %s", out.FinalPrompt)
}
if !strings.Contains(out.FinalPrompt, "Answer the latest user request directly.") {
t.Fatalf("expected neutral continuation instruction in live prompt, got %s", out.FinalPrompt)
if !strings.Contains(out.FinalPrompt, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") {
t.Fatalf("expected continuation-oriented prompt in live prompt, got %s", out.FinalPrompt)
}
if len(out.RefFileIDs) != 1 || out.RefFileIDs[0] != "file-inline-1" {
t.Fatalf("expected current input file id in ref_file_ids, got %#v", out.RefFileIDs)
@@ -302,6 +315,9 @@ func TestApplyCurrentInputFileUploadsFirstTurnWithInjectedWrapper(t *testing.T)
if !strings.Contains(out.PromptTokenText, "first turn content that is long enough") {
t.Fatalf("expected prompt token text to preserve original full context, got %q", out.PromptTokenText)
}
if !strings.Contains(out.PromptTokenText, "# DS2API_HISTORY.txt") || !strings.Contains(out.PromptTokenText, "=== 1. USER ===") {
t.Fatalf("expected prompt token text to include numbered history transcript, got %q", out.PromptTokenText)
}
}
func TestApplyCurrentInputFilePreservesFullContextPromptForTokenCounting(t *testing.T) {
@@ -316,7 +332,7 @@ func TestApplyCurrentInputFilePreservesFullContextPromptForTokenCounting(t *test
DS: ds,
}
req := map[string]any{
"model": "deepseek-v4-flash",
"model": "deepseek-v4-vision",
"messages": historySplitTestMessages(),
}
stdReq, err := promptcompat.NormalizeOpenAIChatRequest(h.Store, req, "")
@@ -337,10 +353,13 @@ func TestApplyCurrentInputFilePreservesFullContextPromptForTokenCounting(t *test
t.Fatalf("expected prompt token text to contain file context with full conversation, got %q", out.PromptTokenText)
}
if strings.Contains(out.PromptTokenText, "[file content end]") || strings.Contains(out.PromptTokenText, "[file name]:") {
t.Fatalf("expected prompt token text to use raw transcript without wrapper tags, got %q", out.PromptTokenText)
t.Fatalf("expected prompt token text to omit file wrapper tags, got %q", out.PromptTokenText)
}
if !strings.Contains(out.PromptTokenText, "Answer the latest user request directly.") {
t.Fatalf("expected prompt token text to also include neutral live prompt, got %q", out.PromptTokenText)
if !strings.Contains(out.PromptTokenText, "# DS2API_HISTORY.txt") || !strings.Contains(out.PromptTokenText, "=== 1. SYSTEM ===") {
t.Fatalf("expected prompt token text to include numbered history transcript, got %q", out.PromptTokenText)
}
if !strings.Contains(out.PromptTokenText, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") {
t.Fatalf("expected prompt token text to also include continuation prompt, got %q", out.PromptTokenText)
}
if strings.Contains(out.FinalPrompt, "first user turn") || strings.Contains(out.FinalPrompt, "latest user turn") {
t.Fatalf("expected live prompt to hide original turns, got %q", out.FinalPrompt)
@@ -359,7 +378,7 @@ func TestApplyCurrentInputFileUploadsFullContextFile(t *testing.T) {
DS: ds,
}
req := map[string]any{
"model": "deepseek-v4-flash",
"model": "deepseek-v4-vision",
"messages": historySplitTestMessages(),
}
stdReq, err := promptcompat.NormalizeOpenAIChatRequest(h.Store, req, "")
@@ -378,20 +397,23 @@ func TestApplyCurrentInputFileUploadsFullContextFile(t *testing.T) {
t.Fatalf("expected one current input upload, got %d", len(ds.uploadCalls))
}
upload := ds.uploadCalls[0]
if upload.Filename != "history.txt" {
t.Fatalf("expected history.txt upload, got %q", upload.Filename)
if upload.Filename != "DS2API_HISTORY.txt" {
t.Fatalf("expected DS2API_HISTORY.txt upload, got %q", upload.Filename)
}
if upload.ModelType != "vision" {
t.Fatalf("expected vision model type for vision request, got %q", upload.ModelType)
}
uploadedText := string(upload.Data)
for _, want := range []string{"system instructions", "first user turn", "hidden reasoning", "tool result", "latest user turn", promptcompat.ThinkingInjectionMarker} {
for _, want := range []string{"# DS2API_HISTORY.txt", "=== 1. SYSTEM ===", "=== 2. USER ===", "=== 3. ASSISTANT ===", "=== 4. TOOL ===", "=== 5. USER ===", "system instructions", "first user turn", "hidden reasoning", "tool result", "latest user turn", promptcompat.ThinkingInjectionMarker} {
if !strings.Contains(uploadedText, want) {
t.Fatalf("expected full context file to contain %q, got %q", want, uploadedText)
}
}
if strings.Contains(out.FinalPrompt, "first user turn") || strings.Contains(out.FinalPrompt, "latest user turn") || strings.Contains(out.FinalPrompt, "CURRENT_USER_INPUT.txt") || strings.Contains(out.FinalPrompt, "history.txt") || strings.Contains(out.FinalPrompt, "Read that file") {
t.Fatalf("expected live prompt to use only a neutral continuation instruction, got %s", out.FinalPrompt)
if strings.Contains(out.FinalPrompt, "first user turn") || strings.Contains(out.FinalPrompt, "latest user turn") || strings.Contains(out.FinalPrompt, "CURRENT_USER_INPUT.txt") || strings.Contains(out.FinalPrompt, "Read that file") {
t.Fatalf("expected live prompt to use only a continuation instruction, got %s", out.FinalPrompt)
}
if !strings.Contains(out.FinalPrompt, "Answer the latest user request directly.") {
t.Fatalf("expected neutral continuation instruction in live prompt, got %s", out.FinalPrompt)
if !strings.Contains(out.FinalPrompt, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") {
t.Fatalf("expected continuation-oriented prompt in live prompt, got %s", out.FinalPrompt)
}
}
@@ -423,6 +445,9 @@ func TestApplyCurrentInputFileCarriesHistoryText(t *testing.T) {
if out.HistoryText != string(ds.uploadCalls[0].Data) {
t.Fatalf("expected current input file flow to preserve uploaded text in history, got %q", out.HistoryText)
}
if !strings.Contains(out.HistoryText, "# DS2API_HISTORY.txt") || !strings.Contains(out.HistoryText, "=== 1. SYSTEM ===") {
t.Fatalf("expected history text to use numbered transcript format, got %q", out.HistoryText)
}
}
func TestChatCompletionsCurrentInputFileUploadsContextAndKeepsNeutralPrompt(t *testing.T) {
@@ -454,7 +479,7 @@ func TestChatCompletionsCurrentInputFileUploadsContextAndKeepsNeutralPrompt(t *t
t.Fatalf("expected 1 upload call, got %d", len(ds.uploadCalls))
}
upload := ds.uploadCalls[0]
if upload.Filename != "history.txt" {
if upload.Filename != "DS2API_HISTORY.txt" {
t.Fatalf("unexpected upload filename: %q", upload.Filename)
}
if upload.Purpose != "assistants" {
@@ -462,7 +487,10 @@ func TestChatCompletionsCurrentInputFileUploadsContextAndKeepsNeutralPrompt(t *t
}
historyText := string(upload.Data)
if strings.Contains(historyText, "[file content end]") || strings.Contains(historyText, "[file content begin]") || strings.Contains(historyText, "[file name]:") {
t.Fatalf("expected plain history transcript without wrapper tags, got %s", historyText)
t.Fatalf("expected history transcript without file wrapper tags, got %s", historyText)
}
if !strings.Contains(historyText, "# DS2API_HISTORY.txt") || !strings.Contains(historyText, "=== 1. SYSTEM ===") {
t.Fatalf("expected history transcript to use numbered sections, got %s", historyText)
}
if !strings.Contains(historyText, "latest user turn") {
t.Fatalf("expected full context to include latest turn, got %s", historyText)
@@ -471,8 +499,8 @@ func TestChatCompletionsCurrentInputFileUploadsContextAndKeepsNeutralPrompt(t *t
t.Fatal("expected completion payload to be captured")
}
promptText, _ := ds.completionReq["prompt"].(string)
if !strings.Contains(promptText, "Answer the latest user request directly.") {
t.Fatalf("expected neutral completion prompt, got %s", promptText)
if !strings.Contains(promptText, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") {
t.Fatalf("expected continuation-oriented prompt, got %s", promptText)
}
if strings.Contains(promptText, "first user turn") || strings.Contains(promptText, "latest user turn") {
t.Fatalf("expected prompt to hide original turns, got %s", promptText)
@@ -523,12 +551,16 @@ func TestResponsesCurrentInputFileUploadsContextAndKeepsNeutralPrompt(t *testing
if len(ds.uploadCalls) != 1 {
t.Fatalf("expected 1 upload call, got %d", len(ds.uploadCalls))
}
historyText := string(ds.uploadCalls[0].Data)
if !strings.Contains(historyText, "# DS2API_HISTORY.txt") || !strings.Contains(historyText, "=== 1. SYSTEM ===") {
t.Fatalf("expected uploaded history text to use numbered transcript format, got %s", historyText)
}
if ds.completionReq == nil {
t.Fatal("expected completion payload to be captured")
}
promptText, _ := ds.completionReq["prompt"].(string)
if !strings.Contains(promptText, "Answer the latest user request directly.") {
t.Fatalf("expected neutral completion prompt, got %s", promptText)
if !strings.Contains(promptText, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") {
t.Fatalf("expected continuation-oriented prompt, got %s", promptText)
}
if strings.Contains(promptText, "first user turn") || strings.Contains(promptText, "latest user turn") {
t.Fatalf("expected prompt to hide original turns, got %s", promptText)
@@ -669,11 +701,15 @@ func TestCurrentInputFileWorksAcrossAutoDeleteModes(t *testing.T) {
if len(ds.uploadCalls) != 1 {
t.Fatalf("expected current input upload for mode=%s, got %d", mode, len(ds.uploadCalls))
}
historyText := string(ds.uploadCalls[0].Data)
if !strings.Contains(historyText, "# DS2API_HISTORY.txt") || !strings.Contains(historyText, "=== 1. SYSTEM ===") {
t.Fatalf("expected uploaded history text to use numbered transcript format, got %s", historyText)
}
if ds.completionReq == nil {
t.Fatalf("expected completion payload for mode=%s", mode)
}
promptText, _ := ds.completionReq["prompt"].(string)
if !strings.Contains(promptText, "Answer the latest user request directly.") || strings.Contains(promptText, "first user turn") || strings.Contains(promptText, "latest user turn") {
if !strings.Contains(promptText, "Continue from the latest state in the attached DS2API_HISTORY.txt context.") || strings.Contains(promptText, "first user turn") || strings.Contains(promptText, "latest user turn") {
t.Fatalf("unexpected prompt for mode=%s: %s", mode, promptText)
}
})

View File

@@ -222,7 +222,13 @@ func (h *Handler) consumeResponsesStreamAttempt(r *http.Request, resp *http.Resp
finalReason = "content_filter"
}
},
OnContextDone: func() {
streamRuntime.markContextCancelled()
},
})
if streamRuntime.finalErrorCode == string(streamengine.StopReasonContextCancelled) {
return true, false
}
terminalWritten := streamRuntime.finalize(finalReason, allowDeferEmpty && finalReason != "content_filter")
if terminalWritten {
return true, false
@@ -235,6 +241,10 @@ func logResponsesStreamTerminal(streamRuntime *responsesStreamRuntime, attempts
if attempts > 0 {
source = "synthetic_retry"
}
if streamRuntime.finalErrorCode == string(streamengine.StopReasonContextCancelled) {
config.Logger.Info("[openai_empty_retry] terminal cancelled", "surface", "responses", "stream", true, "retry_attempts", attempts, "error_code", streamRuntime.finalErrorCode)
return
}
if streamRuntime.failed {
config.Logger.Info("[openai_empty_retry] terminal empty output", "surface", "responses", "stream", true, "retry_attempts", attempts, "success_source", "none", "error_code", streamRuntime.finalErrorCode)
return

View File

@@ -0,0 +1,70 @@
package responses
import (
"context"
"io"
"net/http"
"net/http/httptest"
"strings"
"testing"
"ds2api/internal/promptcompat"
"ds2api/internal/stream"
)
func makeResponsesOpenAISSEHTTPResponse(lines ...string) *http.Response {
body := strings.Join(lines, "\n")
if !strings.HasSuffix(body, "\n") {
body += "\n"
}
return &http.Response{
StatusCode: http.StatusOK,
Header: make(http.Header),
Body: io.NopCloser(strings.NewReader(body)),
}
}
func TestConsumeResponsesStreamAttemptMarksContextCancelledState(t *testing.T) {
ctx, cancel := context.WithCancel(context.Background())
cancel()
req := httptest.NewRequest(http.MethodPost, "/v1/responses", nil).WithContext(ctx)
rec := httptest.NewRecorder()
streamRuntime := newResponsesStreamRuntime(
rec,
http.NewResponseController(rec),
true,
"resp-cancelled",
"deepseek-v4-flash",
"prompt",
false,
false,
true,
nil,
nil,
false,
false,
promptcompat.DefaultToolChoicePolicy(),
"",
nil,
)
resp := makeResponsesOpenAISSEHTTPResponse(
`data: {"p":"response/content","v":"hello"}`,
`data: [DONE]`,
)
h := &Handler{}
terminalWritten, retryable := h.consumeResponsesStreamAttempt(req, resp, streamRuntime, "text", false, true)
if !terminalWritten || retryable {
t.Fatalf("expected cancelled attempt to terminate without retry, got terminalWritten=%v retryable=%v", terminalWritten, retryable)
}
if !streamRuntime.failed {
t.Fatalf("expected cancelled response stream to be marked failed")
}
if got, want := streamRuntime.finalErrorCode, string(stream.StopReasonContextCancelled); got != want {
t.Fatalf("expected cancelled final error code %q, got %q", want, got)
}
if streamRuntime.finalErrorMessage == "" {
t.Fatalf("expected cancelled final error message to be preserved")
}
}

View File

@@ -139,6 +139,13 @@ func (s *responsesStreamRuntime) failResponse(status int, message, code string)
s.sendDone()
}
func (s *responsesStreamRuntime) markContextCancelled() {
s.failed = true
s.finalErrorStatus = 499
s.finalErrorMessage = "request context cancelled"
s.finalErrorCode = string(streamengine.StopReasonContextCancelled)
}
func (s *responsesStreamRuntime) finalize(finishReason string, deferEmptyOutput bool) bool {
s.failed = false
s.finalErrorStatus = 0

View File

@@ -0,0 +1,134 @@
package requestbody
import (
"bytes"
"errors"
"io"
"mime"
"net/http"
"strings"
"unicode/utf8"
)
var (
ErrInvalidUTF8Body = errors.New("invalid utf-8 request body")
errRequestBodyTooLarge = errors.New("request body too large")
)
const maxJSONUTF8ValidationSize = 100 << 20
// ValidateJSONUTF8 validates complete JSON request bodies before downstream
// decoders can silently replace malformed UTF-8 or stop before trailing bytes.
func ValidateJSONUTF8(next http.Handler) http.Handler {
if next == nil {
return http.HandlerFunc(func(http.ResponseWriter, *http.Request) {})
}
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if shouldValidateJSONBody(r) {
r.Body = validateAndReplayBody(r.Body)
}
next.ServeHTTP(w, r)
})
}
func shouldValidateJSONBody(r *http.Request) bool {
if r == nil || r.Body == nil {
return false
}
path := ""
if r.URL != nil {
path = r.URL.Path
}
return isJSONContentType(r.Header.Get("Content-Type")) || isKnownJSONRequestPath(r.Method, path)
}
func isJSONContentType(raw string) bool {
raw = strings.TrimSpace(raw)
if raw == "" {
return false
}
mediaType, _, err := mime.ParseMediaType(raw)
if err != nil {
mediaType = raw
}
mediaType = strings.ToLower(strings.TrimSpace(mediaType))
return strings.Contains(mediaType, "json")
}
func isKnownJSONRequestPath(method, path string) bool {
switch strings.ToUpper(strings.TrimSpace(method)) {
case http.MethodPost, http.MethodPut, http.MethodPatch, http.MethodDelete:
default:
return false
}
path = strings.TrimSpace(path)
if path == "" {
return false
}
switch {
case path == "/v1/chat/completions" || path == "/chat/completions":
return true
case path == "/v1/responses" || path == "/responses":
return true
case path == "/v1/embeddings" || path == "/embeddings":
return true
case path == "/anthropic/v1/messages" || path == "/v1/messages" || path == "/messages":
return true
case path == "/anthropic/v1/messages/count_tokens" || path == "/v1/messages/count_tokens" || path == "/messages/count_tokens":
return true
case strings.HasPrefix(path, "/v1beta/models/") || strings.HasPrefix(path, "/v1/models/"):
return strings.Contains(path, ":generateContent") || strings.Contains(path, ":streamGenerateContent")
case strings.HasPrefix(path, "/admin/"):
return true
default:
return false
}
}
func validateAndReplayBody(body io.ReadCloser) io.ReadCloser {
if body == nil {
return body
}
raw, err := io.ReadAll(io.LimitReader(body, maxJSONUTF8ValidationSize+1))
if err != nil {
return &errorReadCloser{err: err, closer: body}
}
if len(raw) > maxJSONUTF8ValidationSize {
return &errorReadCloser{err: errRequestBodyTooLarge, closer: body}
}
if !utf8.Valid(raw) {
return &errorReadCloser{err: ErrInvalidUTF8Body, closer: body}
}
return &replayReadCloser{Reader: bytes.NewReader(raw), closer: body}
}
type replayReadCloser struct {
*bytes.Reader
closer io.Closer
}
func (r *replayReadCloser) Close() error {
if r == nil || r.closer == nil {
return nil
}
return r.closer.Close()
}
type errorReadCloser struct {
err error
closer io.Closer
}
func (r *errorReadCloser) Read([]byte) (int, error) {
if r == nil || r.err == nil {
return 0, io.EOF
}
return 0, r.err
}
func (r *errorReadCloser) Close() error {
if r == nil || r.closer == nil {
return nil
}
return r.closer.Close()
}

View File

@@ -0,0 +1,158 @@
package requestbody
import (
"bytes"
"encoding/json"
"io"
"net/http"
"net/http/httptest"
"strings"
"testing"
)
type singleByteReadCloser struct {
data []byte
pos int
}
func (r *singleByteReadCloser) Read(p []byte) (int, error) {
if r.pos >= len(r.data) {
return 0, io.EOF
}
p[0] = r.data[r.pos]
r.pos++
return 1, nil
}
func (r *singleByteReadCloser) Close() error {
return nil
}
func TestValidateJSONUTF8AllowsSplitMultibyteRunes(t *testing.T) {
body := []byte(`{"text":"你好"}`)
handler := ValidateJSONUTF8(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
var req map[string]any
if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
t.Fatalf("unexpected decode error: %v", err)
}
w.WriteHeader(http.StatusNoContent)
}))
req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", &singleByteReadCloser{data: body})
req.Header.Set("Content-Type", "application/json")
rec := httptest.NewRecorder()
handler.ServeHTTP(rec, req)
if rec.Code != http.StatusNoContent {
t.Fatalf("expected 204 for valid utf-8 json, got %d body=%q", rec.Code, rec.Body.String())
}
}
func TestValidateJSONUTF8RejectsInvalidBytesBeforeJSONDecode(t *testing.T) {
body := append([]byte(`{"text":"`), 0xff)
body = append(body, []byte(`"}`)...)
handler := ValidateJSONUTF8(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
var req map[string]any
if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
w.WriteHeader(http.StatusBadRequest)
_, _ = w.Write([]byte(err.Error()))
return
}
w.WriteHeader(http.StatusOK)
}))
req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", bytes.NewReader(body))
req.Header.Set("Content-Type", "application/json; charset=utf-8")
rec := httptest.NewRecorder()
handler.ServeHTTP(rec, req)
if rec.Code != http.StatusBadRequest {
t.Fatalf("expected 400 for invalid utf-8 json, got %d body=%q", rec.Code, rec.Body.String())
}
if !strings.Contains(strings.ToLower(rec.Body.String()), "invalid utf-8") {
t.Fatalf("expected utf-8 validation error, got %q", rec.Body.String())
}
}
func TestValidateJSONUTF8RejectsInvalidBytesWithoutJSONContentTypeOnKnownPath(t *testing.T) {
body := append([]byte(`{"text":"`), 0xff)
body = append(body, []byte(`"}`)...)
handler := ValidateJSONUTF8(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
var req map[string]any
if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
w.WriteHeader(http.StatusBadRequest)
_, _ = w.Write([]byte(err.Error()))
return
}
w.WriteHeader(http.StatusOK)
}))
req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", bytes.NewReader(body))
req.Header.Set("Content-Type", "text/plain")
rec := httptest.NewRecorder()
handler.ServeHTTP(rec, req)
if rec.Code != http.StatusBadRequest {
t.Fatalf("expected 400 for invalid utf-8 json, got %d body=%q", rec.Code, rec.Body.String())
}
if !strings.Contains(strings.ToLower(rec.Body.String()), "invalid utf-8") {
t.Fatalf("expected utf-8 validation error, got %q", rec.Body.String())
}
}
func TestValidateJSONUTF8RejectsTrailingInvalidBytesAfterJSONValue(t *testing.T) {
body := append([]byte(`{"text":"ok"}`), 0xff)
handler := ValidateJSONUTF8(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
var req map[string]any
if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
w.WriteHeader(http.StatusBadRequest)
_, _ = w.Write([]byte(err.Error()))
return
}
w.WriteHeader(http.StatusOK)
}))
req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", bytes.NewReader(body))
req.Header.Set("Content-Type", "application/json")
rec := httptest.NewRecorder()
handler.ServeHTTP(rec, req)
if rec.Code != http.StatusBadRequest {
t.Fatalf("expected 400 for trailing invalid utf-8, got %d body=%q", rec.Code, rec.Body.String())
}
if !strings.Contains(strings.ToLower(rec.Body.String()), "invalid utf-8") {
t.Fatalf("expected utf-8 validation error, got %q", rec.Body.String())
}
}
func TestIsJSONContentType(t *testing.T) {
for _, raw := range []string{
"application/json",
"application/json; charset=utf-8",
"application/problem+json",
"application/vnd.api+json",
} {
if !isJSONContentType(raw) {
t.Fatalf("expected %q to be recognized as json", raw)
}
}
for _, raw := range []string{
"multipart/form-data; boundary=abc",
"text/plain",
"application/octet-stream",
} {
if isJSONContentType(raw) {
t.Fatalf("expected %q not to be recognized as json", raw)
}
}
}
func TestIsKnownJSONRequestPathIncludesGeminiStream(t *testing.T) {
if !isKnownJSONRequestPath(http.MethodPost, "/v1beta/models/gemini-pro:streamGenerateContent") {
t.Fatal("expected Gemini stream generate path to be recognized as json")
}
}

View File

@@ -516,32 +516,51 @@ function observeContinueState(state, chunk) {
if (topID > 0) {
state.responseMessageID = topID;
}
if (chunk.p === 'response/status') {
setContinueStatus(state, asString(chunk.v));
}
observeContinueDirectPatch(state, chunk.p, chunk.v);
if (chunk.p === 'response') {
observeContinueBatchPatches(state, 'response', chunk.v);
} else {
observeContinueBatchPatches(state, '', chunk.v);
}
const response = chunk.v && typeof chunk.v === 'object' ? chunk.v.response : null;
if (response && typeof response === 'object') {
const id = numberValue(response.message_id);
if (id > 0) {
state.responseMessageID = id;
}
setContinueStatus(state, asString(response.status));
if (response.auto_continue === true) {
state.lastStatus = 'AUTO_CONTINUE';
}
}
observeContinueResponseObject(state, response);
const messageResponse = chunk.message && typeof chunk.message === 'object' && chunk.message.response;
if (messageResponse && typeof messageResponse === 'object') {
const id = numberValue(messageResponse.message_id);
if (id > 0) {
state.responseMessageID = id;
}
setContinueStatus(state, asString(messageResponse.status));
observeContinueResponseObject(state, messageResponse);
}
function observeContinueDirectPatch(state, path, value) {
if (!state) {
return;
}
switch (asString(path).trim().replace(/^\/+|\/+$/g, '')) {
case 'response/status':
case 'status':
case 'response/quasi_status':
case 'quasi_status':
setContinueStatus(state, asString(value));
break;
case 'response/auto_continue':
case 'auto_continue':
if (value === true) {
state.lastStatus = 'AUTO_CONTINUE';
}
break;
default:
break;
}
}
function observeContinueResponseObject(state, response) {
if (!state || !response || typeof response !== 'object') {
return;
}
const id = numberValue(response.message_id);
if (id > 0) {
state.responseMessageID = id;
}
setContinueStatus(state, asString(response.status));
if (response.auto_continue === true) {
state.lastStatus = 'AUTO_CONTINUE';
}
}
@@ -569,6 +588,12 @@ function observeContinueBatchPatches(state, parentPath, raw) {
case 'quasi_status':
setContinueStatus(state, asString(patch.v));
break;
case 'response/auto_continue':
case 'auto_continue':
if (patch.v === true) {
state.lastStatus = 'AUTO_CONTINUE';
}
break;
default:
break;
}

View File

@@ -248,6 +248,9 @@ function replaceDSMLToolMarkupOutsideIgnored(text) {
if (tag) {
if (tag.dsmlLike) {
out += `<${tag.closing ? '/' : ''}${tag.name}${raw.slice(tag.nameEnd, tag.end + 1)}`;
if (raw[tag.end] !== '>') {
out += '>';
}
} else {
out += raw.slice(tag.start, tag.end + 1);
}
@@ -424,31 +427,42 @@ function scanToolMarkupTagAt(text, start) {
}
const lower = raw.toLowerCase();
let i = start + 1;
while (i < raw.length && raw[i] === '<') {
i += 1;
}
const closing = raw[i] === '/';
if (closing) {
i += 1;
}
let dsmlLike = false;
if (i < raw.length && isToolMarkupPipe(raw[i])) {
dsmlLike = true;
i += 1;
}
if (lower.startsWith('dsml', i)) {
dsmlLike = true;
i += 'dsml'.length;
while (i < raw.length && isToolMarkupSeparator(raw[i])) {
i += 1;
}
}
const prefix = consumeToolMarkupNamePrefix(raw, lower, i);
i = prefix.next;
const dsmlLike = prefix.dsmlLike;
const { name, len } = matchToolMarkupName(lower, i);
if (!name) {
return null;
}
const nameEnd = i + len;
const originalNameEnd = i + len;
let nameEnd = originalNameEnd;
while (nameEnd < raw.length && isToolMarkupPipe(raw[nameEnd])) {
nameEnd += 1;
}
const hasTrailingPipe = nameEnd > originalNameEnd;
if (!hasXmlTagBoundary(raw, nameEnd)) {
return null;
}
const end = findXmlTagEnd(raw, nameEnd);
let end = findXmlTagEnd(raw, nameEnd);
if (end < 0) {
if (!hasTrailingPipe) {
return null;
}
end = nameEnd - 1;
}
if (hasTrailingPipe) {
const nextLT = raw.indexOf('<', nameEnd);
if (nextLT >= 0 && end >= nextLT) {
end = nameEnd - 1;
}
}
if (end < 0) {
return null;
}
@@ -520,37 +534,94 @@ function findPartialToolMarkupStart(text) {
if (lastLT < 0) {
return -1;
}
const tail = raw.slice(lastLT);
const start = includeDuplicateLeadingLessThan(raw, lastLT);
const tail = raw.slice(start);
if (tail.includes('>')) {
return -1;
}
const lowerTail = tail.toLowerCase();
const candidates = [
'<tool_calls', '<invoke', '<parameter',
'<|tool_calls', '<|invoke', '<|parameter',
'<tool_calls', '<invoke', '<parameter',
'<|dsml|tool_calls', '<|dsml|invoke', '<|dsml|parameter',
'<dsml|tool_calls', '<dsml|invoke', '<dsml|parameter',
'<dsmltool_calls', '<dsmlinvoke', '<dsmlparameter',
'<dsml tool_calls', '<dsml invoke', '<dsml parameter',
'<dsml|tool_calls', '<dsml|invoke', '<dsml|parameter',
'<|dsmltool_calls', '<|dsmlinvoke', '<|dsmlparameter',
'<|dsml tool_calls', '<|dsml invoke', '<|dsml parameter',
];
for (const candidate of candidates) {
if (candidate.startsWith(lowerTail)) {
return lastLT;
}
return isPartialToolMarkupTagPrefix(tail) ? start : -1;
}
function includeDuplicateLeadingLessThan(text, idx) {
let out = idx;
while (out > 0 && text[out - 1] === '<') {
out -= 1;
}
return -1;
return out;
}
function isToolMarkupPipe(ch) {
return ch === '|' || ch === '';
}
function isToolMarkupSeparator(ch) {
return ch === ' ' || ch === '\t' || ch === '\r' || ch === '\n' || isToolMarkupPipe(ch);
function isPartialToolMarkupTagPrefix(text) {
const raw = toStringSafe(text);
if (!raw || raw[0] !== '<' || raw.includes('>')) {
return false;
}
const lower = raw.toLowerCase();
let i = 1;
while (i < raw.length && raw[i] === '<') {
i += 1;
}
if (i >= raw.length) {
return true;
}
if (raw[i] === '/') {
i += 1;
}
while (i <= raw.length) {
if (i === raw.length) {
return true;
}
if (hasToolMarkupNamePrefix(lower.slice(i))) {
return true;
}
if ('dsml'.startsWith(lower.slice(i))) {
return true;
}
const next = consumeToolMarkupNamePrefixOnce(raw, lower, i);
if (!next.ok) {
return false;
}
i = next.next;
}
return false;
}
function consumeToolMarkupNamePrefix(raw, lower, idx) {
let next = idx;
let dsmlLike = false;
while (true) {
const consumed = consumeToolMarkupNamePrefixOnce(raw, lower, next);
if (!consumed.ok) {
return { next, dsmlLike };
}
next = consumed.next;
dsmlLike = true;
}
}
function consumeToolMarkupNamePrefixOnce(raw, lower, idx) {
if (idx < raw.length && isToolMarkupPipe(raw[idx])) {
return { next: idx + 1, ok: true };
}
if (idx < raw.length && [' ', '\t', '\r', '\n'].includes(raw[idx])) {
return { next: idx + 1, ok: true };
}
if (lower.startsWith('dsml', idx)) {
return { next: idx + 'dsml'.length, ok: true };
}
return { next: idx, ok: false };
}
function hasToolMarkupNamePrefix(lowerTail) {
for (const name of TOOL_MARKUP_NAMES) {
if (lowerTail.startsWith(name) || name.startsWith(lowerTail)) {
return true;
}
}
return false;
}
function matchToolMarkupName(lower, start) {
@@ -775,10 +846,18 @@ function parseMarkupValue(raw, paramName = '') {
if (cdata.ok) {
const literal = parseJSONLiteralValue(cdata.value);
if (literal.ok) {
const literalArray = coerceArrayValue(literal.value, paramName);
if (literalArray.ok) {
return literalArray.value;
}
return literal.value;
}
const structured = parseStructuredCDATAParameterValue(paramName, cdata.value);
return structured.ok ? structured.value : cdata.value;
if (structured.ok) {
return structured.value;
}
const looseArray = parseLooseJSONArrayValue(cdata.value, paramName);
return looseArray.ok ? looseArray.value : cdata.value;
}
const s = toStringSafe(extractRawTagValue(raw)).trim();
if (!s) {
@@ -791,8 +870,14 @@ function parseMarkupValue(raw, paramName = '') {
return nested;
}
if (nested && typeof nested === 'object') {
const nestedArray = coerceArrayValue(nested, paramName);
if (nestedArray.ok) {
return nestedArray.value;
}
if (isOnlyRawValue(nested)) {
return toStringSafe(nested._raw);
const rawValue = toStringSafe(nested._raw);
const looseArray = parseLooseJSONArrayValue(rawValue, paramName);
return looseArray.ok ? looseArray.value : rawValue;
}
return nested;
}
@@ -800,8 +885,16 @@ function parseMarkupValue(raw, paramName = '') {
const literal = parseJSONLiteralValue(s);
if (literal.ok) {
const literalArray = coerceArrayValue(literal.value, paramName);
if (literalArray.ok) {
return literalArray.value;
}
return literal.value;
}
const looseArray = parseLooseJSONArrayValue(s, paramName);
if (looseArray.ok) {
return looseArray.value;
}
return s;
}
@@ -937,6 +1030,226 @@ function parseJSONLiteralValue(raw) {
}
}
function parseLooseJSONArrayValue(raw, paramName = '') {
if (preservesCDATAStringParameter(paramName)) {
return { ok: false, value: null };
}
const s = toStringSafe(raw).trim();
if (!s) {
return { ok: false, value: null };
}
const candidate = parseLooseJSONArrayCandidate(s, paramName);
if (candidate.ok) {
return candidate;
}
const segments = splitTopLevelJSONValues(s);
if (segments.length < 2) {
return { ok: false, value: null };
}
const out = [];
for (const segment of segments) {
const parsed = parseLooseArrayElementValue(segment);
if (!parsed.ok) {
return { ok: false, value: null };
}
out.push(parsed.value);
}
return { ok: true, value: out };
}
function parseLooseJSONArrayCandidate(raw, paramName = '') {
const parsed = parseLooseArrayElementValue(raw);
if (!parsed.ok) {
return { ok: false, value: null };
}
return coerceArrayValue(parsed.value, paramName);
}
function parseLooseArrayElementValue(raw) {
const s = toStringSafe(raw).trim();
if (!s) {
return { ok: false, value: null };
}
const literal = parseJSONLiteralValue(s);
if (literal.ok) {
return literal;
}
const repairedBackslashes = repairInvalidJSONBackslashes(s);
if (repairedBackslashes !== s) {
try {
const parsed = JSON.parse(repairedBackslashes);
return { ok: true, value: parsed };
} catch (_err) {
// Fall through.
}
}
const repairedLoose = repairLooseJSON(s);
if (repairedLoose !== s) {
try {
const parsed = JSON.parse(repairedLoose);
return { ok: true, value: parsed };
} catch (_err) {
// Fall through.
}
}
if (s.includes('<') && s.includes('>')) {
const parsed = parseMarkupInput(s);
if (Array.isArray(parsed)) {
return { ok: true, value: parsed };
}
if (parsed && typeof parsed === 'object') {
return { ok: true, value: parsed };
}
}
return { ok: false, value: null };
}
function coerceArrayValue(value, paramName = '') {
if (Array.isArray(value)) {
return { ok: true, value };
}
if (!value || typeof value !== 'object') {
return { ok: false, value: null };
}
const keys = Object.keys(value);
if (keys.length !== 1) {
return { ok: false, value: null };
}
if (Object.prototype.hasOwnProperty.call(value, 'item')) {
const items = value.item;
const nested = coerceArrayValue(items, '');
return nested.ok ? nested : { ok: true, value: [items] };
}
if (paramName && Object.prototype.hasOwnProperty.call(value, paramName)) {
const nested = coerceArrayValue(value[paramName], '');
if (nested.ok) {
return nested;
}
}
return { ok: false, value: null };
}
function splitTopLevelJSONValues(raw) {
const s = toStringSafe(raw).trim();
if (!s) {
return [];
}
const values = [];
let start = 0;
let depth = 0;
let inString = false;
let escaped = false;
for (let i = 0; i < s.length; i += 1) {
const ch = s[i];
if (inString) {
if (escaped) {
escaped = false;
continue;
}
if (ch === '\\') {
escaped = true;
continue;
}
if (ch === '"') {
inString = false;
}
continue;
}
if (ch === '"') {
inString = true;
continue;
}
if (ch === '{' || ch === '[') {
depth += 1;
continue;
}
if (ch === '}' || ch === ']') {
if (depth > 0) {
depth -= 1;
}
continue;
}
if (ch === ',' && depth === 0) {
const segment = s.slice(start, i).trim();
if (!segment) {
return [];
}
values.push(segment);
start = i + 1;
}
}
const last = s.slice(start).trim();
if (!last) {
return [];
}
values.push(last);
return values.length > 1 ? values : [];
}
function repairInvalidJSONBackslashes(s) {
if (!s || !s.includes('\\')) {
return s;
}
let out = '';
for (let i = 0; i < s.length; i += 1) {
const ch = s[i];
if (ch !== '\\') {
out += ch;
continue;
}
if (i + 1 < s.length) {
const next = s[i + 1];
if ('"\\/bfnrt'.includes(next)) {
out += `\\${next}`;
i += 1;
continue;
}
if (next === 'u' && i + 5 < s.length) {
let isHex = true;
for (let j = 1; j <= 4; j += 1) {
const r = s[i + 1 + j];
if (!/[0-9a-fA-F]/.test(r)) {
isHex = false;
break;
}
}
if (isHex) {
out += `\\u${s.slice(i + 2, i + 6)}`;
i += 5;
continue;
}
}
}
out += '\\\\';
}
return out;
}
function repairLooseJSON(s) {
const raw = toStringSafe(s).trim();
if (!raw) {
return raw;
}
let out = raw.replace(/([{,]\s*)([a-zA-Z_][a-zA-Z0-9_]*)\s*:/g, '$1"$2":');
out = out.replace(/(:\s*)(\{(?:[^{}]|\{[^{}]*\})*\}(?:\s*,\s*\{(?:[^{}]|\{[^{}]*\})*\})+)/g, '$1[$2]');
return out;
}
function sanitizeLooseCDATA(text) {
const raw = toStringSafe(text);
if (!raw) {

View File

@@ -1,55 +0,0 @@
'use strict';
const XML_TOOL_SEGMENT_TAGS = [
'<|dsml|tool_calls>', '<|dsml|tool_calls\n', '<|dsml|tool_calls ',
'<dsml|tool_calls>', '<dsml|tool_calls\n', '<dsml|tool_calls ',
'<|dsml|invoke ', '<|dsml|invoke\n', '<|dsml|invoke\t', '<|dsml|invoke\r',
'<|dsmltool_calls>', '<|dsmltool_calls\n', '<|dsmltool_calls ',
'<|dsmlinvoke ', '<|dsmlinvoke\n', '<|dsmlinvoke\t', '<|dsmlinvoke\r',
'<|dsml tool_calls>', '<|dsml tool_calls\n', '<|dsml tool_calls ',
'<|dsml invoke ', '<|dsml invoke\n', '<|dsml invoke\t', '<|dsml invoke\r',
'<dsml|tool_calls>', '<dsml|tool_calls\n', '<dsml|tool_calls ',
'<dsml|invoke ', '<dsml|invoke\n', '<dsml|invoke\t', '<dsml|invoke\r',
'<dsmltool_calls>', '<dsmltool_calls\n', '<dsmltool_calls ',
'<dsmlinvoke ', '<dsmlinvoke\n', '<dsmlinvoke\t', '<dsmlinvoke\r',
'<dsml tool_calls>', '<dsml tool_calls\n', '<dsml tool_calls ',
'<dsml invoke ', '<dsml invoke\n', '<dsml invoke\t', '<dsml invoke\r',
'<tool_calls>', '<tool_calls\n', '<tool_calls ',
'<invoke ', '<invoke\n', '<invoke\t', '<invoke\r',
'<|tool_calls>', '<|tool_calls\n', '<|tool_calls ',
'<|invoke ', '<|invoke\n', '<|invoke\t', '<|invoke\r',
'<tool_calls>', '<tool_calls\n', '<tool_calls ',
'<invoke ', '<invoke\n', '<invoke\t', '<invoke\r',
];
const XML_TOOL_OPENING_TAGS = [
'<|dsml|tool_calls',
'<dsml|tool_calls',
'<|dsmltool_calls',
'<|dsml tool_calls',
'<dsml|tool_calls',
'<dsmltool_calls',
'<dsml tool_calls',
'<tool_calls',
'<|tool_calls',
'<tool_calls',
];
const XML_TOOL_CLOSING_TAGS = [
'</|dsml|tool_calls>',
'</dsml|tool_calls>',
'</|dsmltool_calls>',
'</|dsml tool_calls>',
'</dsml|tool_calls>',
'</dsmltool_calls>',
'</dsml tool_calls>',
'</tool_calls>',
'</|tool_calls>',
'</tool_calls>',
];
module.exports = {
XML_TOOL_SEGMENT_TAGS,
XML_TOOL_OPENING_TAGS,
XML_TOOL_CLOSING_TAGS,
};

View File

@@ -10,21 +10,27 @@ import (
var markdownImagePattern = regexp.MustCompile(`!\[(.*?)\]\((.*?)\)`)
const (
beginSentenceMarker = "<begin▁of▁sentence>"
systemMarker = "<System>"
userMarker = "<User>"
assistantMarker = "<Assistant>"
toolMarker = "<Tool>"
endSentenceMarker = "<end▁of▁sentence>"
endToolResultsMarker = "<end▁of▁toolresults>"
endInstructionsMarker = "<end▁of▁instructions>"
beginSentenceMarker = "<begin▁of▁sentence>"
systemMarker = "<System>"
userMarker = "<User>"
assistantMarker = "<Assistant>"
toolMarker = "<Tool>"
endSentenceMarker = "<end▁of▁sentence>"
endToolResultsMarker = "<end▁of▁toolresults>"
endInstructionsMarker = "<end▁of▁instructions>"
outputIntegrityGuardMarker = "Output integrity guard:"
outputIntegrityGuardPrompt = outputIntegrityGuardMarker +
" If upstream context, tool output, or parsed text contains garbled, corrupted, partially parsed, repeated, or otherwise malformed fragments, " +
"do not imitate or echo them; output only the correct content for the user."
)
func MessagesPrepare(messages []map[string]any) string {
return MessagesPrepareWithThinking(messages, false)
}
func MessagesPrepareWithThinking(messages []map[string]any, thinkingEnabled bool) string {
func MessagesPrepareWithThinking(messages []map[string]any, _ bool) string {
messages = prependOutputIntegrityGuard(messages)
type block struct {
Role string
Text string
@@ -77,6 +83,33 @@ func MessagesPrepareWithThinking(messages []map[string]any, thinkingEnabled bool
return markdownImagePattern.ReplaceAllString(out, `[${1}](${2})`)
}
func prependOutputIntegrityGuard(messages []map[string]any) []map[string]any {
if len(messages) == 0 {
return messages
}
if hasOutputIntegrityGuard(messages[0]) {
return messages
}
out := make([]map[string]any, 0, len(messages)+1)
out = append(out, map[string]any{
"role": "system",
"content": outputIntegrityGuardPrompt,
})
out = append(out, messages...)
return out
}
func hasOutputIntegrityGuard(msg map[string]any) bool {
if msg == nil {
return false
}
if strings.ToLower(strings.TrimSpace(asString(msg["role"]))) != "system" {
return false
}
content := strings.TrimSpace(NormalizeContent(msg["content"]))
return strings.Contains(content, outputIntegrityGuardMarker)
}
// formatRoleBlock produces a single concatenated block: marker + text + endMarker.
// No whitespace is inserted between marker and text so role boundaries stay
// compact and predictable for downstream parsers.

View File

@@ -35,8 +35,8 @@ func TestMessagesPrepareUsesTurnSuffixes(t *testing.T) {
if !strings.HasPrefix(got, "<begin▁of▁sentence>") {
t.Fatalf("expected begin-of-sentence marker, got %q", got)
}
if !strings.Contains(got, "<System>System rule<end▁of▁instructions>") {
t.Fatalf("expected system instructions suffix, got %q", got)
if !strings.Contains(got, "<System>") || !strings.Contains(got, "<end▁of▁instructions>") || !strings.Contains(got, "System rule") {
t.Fatalf("expected system instructions to remain present, got %q", got)
}
if !strings.Contains(got, "<User>Question") {
t.Fatalf("expected user question, got %q", got)
@@ -49,6 +49,23 @@ func TestMessagesPrepareUsesTurnSuffixes(t *testing.T) {
}
}
func TestMessagesPreparePrependsOutputIntegrityGuard(t *testing.T) {
messages := []map[string]any{
{"role": "system", "content": "System rule"},
{"role": "user", "content": "Question"},
}
got := MessagesPrepare(messages)
if !strings.HasPrefix(got, beginSentenceMarker+systemMarker+outputIntegrityGuardPrompt) {
t.Fatalf("expected output integrity guard to be prepended, got %q", got)
}
if !strings.Contains(got, outputIntegrityGuardPrompt+"\n\nSystem rule") {
t.Fatalf("expected output integrity guard to precede system prompt content, got %q", got)
}
if !strings.Contains(got, "<User>Question") {
t.Fatalf("expected user question after guard, got %q", got)
}
}
func TestNormalizeContentArrayFallsBackToContentWhenTextEmpty(t *testing.T) {
got := NormalizeContent([]any{
map[string]any{"type": "text", "text": "", "content": "from-content"},

View File

@@ -1,35 +1,108 @@
package promptcompat
import (
"fmt"
"strings"
"ds2api/internal/prompt"
)
const CurrentInputContextFilename = "history.txt"
const CurrentInputContextFilename = "DS2API_HISTORY.txt"
const historyTranscriptTitle = "# DS2API_HISTORY.txt"
const historyTranscriptSummary = "Prior conversation history and tool progress."
func BuildOpenAIHistoryTranscript(messages []any) string {
return buildOpenAIInjectedFileTranscript(messages)
return buildOpenAIHistoryTranscript(messages)
}
func BuildOpenAICurrentUserInputTranscript(text string) string {
if strings.TrimSpace(text) == "" {
return ""
}
return BuildOpenAICurrentInputContextTranscript([]any{
return buildOpenAIHistoryTranscript([]any{
map[string]any{"role": "user", "content": text},
})
}
func BuildOpenAICurrentInputContextTranscript(messages []any) string {
return buildOpenAIInjectedFileTranscript(messages)
return buildOpenAIHistoryTranscript(messages)
}
func buildOpenAIInjectedFileTranscript(messages []any) string {
normalized := NormalizeOpenAIMessagesForPrompt(messages, "")
transcript := strings.TrimSpace(prompt.MessagesPrepare(normalized))
func buildOpenAIHistoryTranscript(messages []any) string {
if len(messages) == 0 {
return ""
}
var b strings.Builder
b.WriteString(historyTranscriptTitle)
b.WriteString("\n")
b.WriteString(historyTranscriptSummary)
b.WriteString("\n\n")
entry := 0
for _, raw := range messages {
msg, ok := raw.(map[string]any)
if !ok {
continue
}
role := normalizeOpenAIRoleForPrompt(strings.ToLower(strings.TrimSpace(asString(msg["role"]))))
content := strings.TrimSpace(buildOpenAIHistoryEntry(role, msg))
if content == "" {
continue
}
entry++
fmt.Fprintf(&b, "=== %d. %s ===\n%s\n\n", entry, strings.ToUpper(roleLabelForHistory(role)), content)
}
transcript := strings.TrimSpace(b.String())
if transcript == "" {
return ""
}
return transcript
return transcript + "\n"
}
func buildOpenAIHistoryEntry(role string, msg map[string]any) string {
switch role {
case "assistant":
return strings.TrimSpace(buildAssistantContentForPrompt(msg))
case "tool", "function":
return strings.TrimSpace(buildToolHistoryContent(msg))
case "system", "user":
return strings.TrimSpace(NormalizeOpenAIContentForPrompt(msg["content"]))
default:
return strings.TrimSpace(NormalizeOpenAIContentForPrompt(msg["content"]))
}
}
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 roleLabelForHistory(role string) string {
role = strings.ToLower(strings.TrimSpace(role))
switch role {
case "function":
return "tool"
case "":
return "unknown"
default:
return role
}
}

View File

@@ -88,6 +88,38 @@ func TestBuildOpenAIFinalPrompt_VercelPreparePathKeepsFinalAnswerInstruction(t *
}
}
func TestBuildOpenAIFinalPromptPrependsOutputIntegrityGuard(t *testing.T) {
messages := []any{
map[string]any{"role": "system", "content": "You are helpful"},
map[string]any{"role": "user", "content": "请调用工具"},
}
tools := []any{
map[string]any{
"type": "function",
"function": map[string]any{
"name": "search",
"description": "search docs",
"parameters": map[string]any{
"type": "object",
},
},
},
}
finalPrompt, _ := buildOpenAIFinalPrompt(messages, tools, "", false)
guardIdx := strings.Index(finalPrompt, "Output integrity guard")
toolIdx := strings.Index(finalPrompt, "TOOL CALL FORMAT")
if guardIdx < 0 {
t.Fatalf("expected output integrity guard in final prompt, got: %q", finalPrompt)
}
if toolIdx < 0 {
t.Fatalf("expected tool instructions in final prompt, got: %q", finalPrompt)
}
if guardIdx > toolIdx {
t.Fatalf("expected output integrity guard to precede tool instructions, got: %q", finalPrompt)
}
}
func TestBuildOpenAIFinalPromptReadLikeToolIncludesCacheGuard(t *testing.T) {
messages := []any{
map[string]any{"role": "user", "content": "请读取文件"},

View File

@@ -27,6 +27,7 @@ import (
"ds2api/internal/httpapi/openai/files"
"ds2api/internal/httpapi/openai/responses"
"ds2api/internal/httpapi/openai/shared"
"ds2api/internal/httpapi/requestbody"
"ds2api/internal/webui"
)
@@ -75,6 +76,7 @@ func NewApp() (*App, error) {
r.Use(filteredLogger())
r.Use(middleware.Recoverer)
r.Use(cors)
r.Use(requestbody.ValidateJSONUTF8)
r.Use(timeout(0))
healthzHandler := func(w http.ResponseWriter, _ *http.Request) {

View File

@@ -0,0 +1,89 @@
package server
import (
"bytes"
"net/http"
"net/http/httptest"
"strings"
"testing"
)
func TestJSONRequestsRejectInvalidUTF8BeforeDecode(t *testing.T) {
t.Setenv("DS2API_CONFIG_JSON", `{"keys":["managed-key"],"accounts":[{"email":"u@example.com","password":"p"}]}`)
t.Setenv("DS2API_ENV_WRITEBACK", "0")
app, err := NewApp()
if err != nil {
t.Fatalf("NewApp() error: %v", err)
}
body := append([]byte(`{"model":"deepseek-v4-flash","messages":[{"role":"user","content":"`), 0xff)
body = append(body, []byte(`"}]}`)...)
req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", bytes.NewReader(body))
req.Header.Set("Content-Type", "application/json; charset=utf-8")
req.Header.Set("x-api-key", "direct-token")
rec := httptest.NewRecorder()
app.Router.ServeHTTP(rec, req)
if rec.Code != http.StatusBadRequest {
t.Fatalf("expected 400 for invalid utf-8 request body, got %d body=%q", rec.Code, rec.Body.String())
}
if !strings.Contains(strings.ToLower(rec.Body.String()), "invalid json") {
t.Fatalf("expected invalid json error, got %q", rec.Body.String())
}
}
func TestKnownJSONRequestsRejectInvalidUTF8WithoutJSONContentType(t *testing.T) {
t.Setenv("DS2API_CONFIG_JSON", `{"keys":["managed-key"],"accounts":[{"email":"u@example.com","password":"p"}]}`)
t.Setenv("DS2API_ENV_WRITEBACK", "0")
app, err := NewApp()
if err != nil {
t.Fatalf("NewApp() error: %v", err)
}
body := append([]byte(`{"model":"deepseek-v4-flash","messages":[{"role":"user","content":"`), 0xff)
body = append(body, []byte(`"}]}`)...)
req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", bytes.NewReader(body))
req.Header.Set("Content-Type", "text/plain")
req.Header.Set("x-api-key", "direct-token")
rec := httptest.NewRecorder()
app.Router.ServeHTTP(rec, req)
if rec.Code != http.StatusBadRequest {
t.Fatalf("expected 400 for invalid utf-8 request body, got %d body=%q", rec.Code, rec.Body.String())
}
if !strings.Contains(strings.ToLower(rec.Body.String()), "invalid json") {
t.Fatalf("expected invalid json error, got %q", rec.Body.String())
}
}
func TestJSONRequestsRejectTrailingInvalidUTF8AfterCompleteJSON(t *testing.T) {
t.Setenv("DS2API_CONFIG_JSON", `{"keys":["managed-key"],"accounts":[{"email":"u@example.com","password":"p"}]}`)
t.Setenv("DS2API_ENV_WRITEBACK", "0")
app, err := NewApp()
if err != nil {
t.Fatalf("NewApp() error: %v", err)
}
body := append([]byte(`{"model":"deepseek-v4-flash","messages":[{"role":"user","content":"ok"}]}`), 0xff)
req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", bytes.NewReader(body))
req.Header.Set("Content-Type", "application/json")
req.Header.Set("x-api-key", "direct-token")
rec := httptest.NewRecorder()
app.Router.ServeHTTP(rec, req)
if rec.Code != http.StatusBadRequest {
t.Fatalf("expected 400 for trailing invalid utf-8, got %d body=%q", rec.Code, rec.Body.String())
}
if !strings.Contains(strings.ToLower(rec.Body.String()), "invalid json") {
t.Fatalf("expected invalid json error, got %q", rec.Body.String())
}
}

View File

@@ -41,6 +41,15 @@ func TestCollectStreamTextOnly(t *testing.T) {
}
}
func TestCollectStreamHandlesLongSingleSSELine(t *testing.T) {
payload := strings.Repeat("x", 2*1024*1024+4096)
resp := makeHTTPResponse(makeLargeContentSSEBody(t, payload))
result := CollectStream(resp, false, true)
if result.Text != payload {
t.Fatalf("long SSE line payload mismatch: got len=%d want len=%d", len(result.Text), len(payload))
}
}
func TestCollectStreamThinkingAndText(t *testing.T) {
resp := makeHTTPResponse(
"data: {\"p\":\"response/thinking_content\",\"v\":\"Thinking...\"}\n" +

View File

@@ -9,8 +9,7 @@ import (
const (
parsedLineBufferSize = 128
scannerBufferSize = 64 * 1024
maxScannerLineSize = 2 * 1024 * 1024
lineReaderBufferSize = 64 * 1024
minFlushChars = 160
maxFlushWait = 80 * time.Millisecond
)
@@ -29,8 +28,8 @@ func StartParsedLinePump(ctx context.Context, body io.Reader, thinkingEnabled bo
eof bool
}
lineCh := make(chan scanItem, 1)
stopScanner := make(chan struct{})
defer close(stopScanner)
stopReader := make(chan struct{})
defer close(stopReader)
go func() {
sendScanItem := func(item scanItem) bool {
select {
@@ -38,20 +37,28 @@ func StartParsedLinePump(ctx context.Context, body io.Reader, thinkingEnabled bo
return true
case <-ctx.Done():
return false
case <-stopScanner:
case <-stopReader:
return false
}
}
defer close(lineCh)
scanner := bufio.NewScanner(body)
scanner.Buffer(make([]byte, 0, scannerBufferSize), maxScannerLineSize)
for scanner.Scan() {
line := append([]byte{}, scanner.Bytes()...)
if !sendScanItem(scanItem{line: line}) {
reader := bufio.NewReaderSize(body, lineReaderBufferSize)
for {
line, err := reader.ReadBytes('\n')
if len(line) > 0 {
line = append([]byte{}, line...)
if !sendScanItem(scanItem{line: line}) {
return
}
}
if err != nil {
if err == io.EOF {
err = nil
}
_ = sendScanItem(scanItem{err: err, eof: true})
return
}
}
_ = sendScanItem(scanItem{err: scanner.Err(), eof: true})
}()
ticker := time.NewTicker(maxFlushWait)

View File

@@ -2,10 +2,23 @@ package sse
import (
"context"
"encoding/json"
"strings"
"testing"
)
func makeLargeContentSSEBody(t *testing.T, payload string) string {
t.Helper()
line, err := json.Marshal(map[string]any{
"p": "response/content",
"v": payload,
})
if err != nil {
t.Fatalf("marshal SSE line failed: %v", err)
}
return "data: " + string(line) + "\n" + "data: [DONE]\n"
}
func TestStartParsedLinePumpParsesAndStops(t *testing.T) {
body := strings.NewReader("data: {\"p\":\"response/content\",\"v\":\"hi\"}\n\ndata: [DONE]\n")
results, done := StartParsedLinePump(context.Background(), body, false, "text")
@@ -28,3 +41,28 @@ func TestStartParsedLinePumpParsesAndStops(t *testing.T) {
t.Fatalf("expected last line to stop stream, got parsed=%v stop=%v", last.Parsed, last.Stop)
}
}
func TestStartParsedLinePumpHandlesLongSingleSSELine(t *testing.T) {
payload := strings.Repeat("x", 2*1024*1024+4096)
results, done := StartParsedLinePump(context.Background(), strings.NewReader(makeLargeContentSSEBody(t, payload)), false, "text")
var got strings.Builder
var sawDone bool
for r := range results {
for _, p := range r.Parts {
got.WriteString(p.Text)
}
if r.Stop {
sawDone = true
}
}
if err := <-done; err != nil {
t.Fatalf("unexpected long-line read error: %v", err)
}
if got.String() != payload {
t.Fatalf("long SSE line payload mismatch: got len=%d want len=%d", got.Len(), len(payload))
}
if !sawDone {
t.Fatal("expected DONE after long SSE line")
}
}

View File

@@ -0,0 +1,164 @@
package toolcall
import (
"encoding/json"
"html"
"strings"
)
func parseLooseJSONArrayValue(raw, paramName string) ([]any, bool) {
if preservesCDATAStringParameter(paramName) {
return nil, false
}
trimmed := strings.TrimSpace(html.UnescapeString(raw))
if trimmed == "" {
return nil, false
}
if parsed, ok := parseLooseJSONArrayCandidate(trimmed, paramName); ok {
return parsed, true
}
segments, ok := splitTopLevelJSONValues(trimmed)
if !ok {
return nil, false
}
out := make([]any, 0, len(segments))
for _, segment := range segments {
parsed, ok := parseLooseArrayElementValue(segment)
if !ok {
return nil, false
}
out = append(out, parsed)
}
return out, true
}
func parseLooseJSONArrayCandidate(raw, paramName string) ([]any, bool) {
parsed, ok := parseLooseArrayElementValue(raw)
if !ok {
return nil, false
}
return coerceArrayValue(parsed, paramName)
}
func parseLooseArrayElementValue(raw string) (any, bool) {
trimmed := strings.TrimSpace(html.UnescapeString(raw))
if trimmed == "" {
return nil, false
}
var parsed any
if err := json.Unmarshal([]byte(trimmed), &parsed); err == nil {
return parsed, true
}
repairedBackslashes := repairInvalidJSONBackslashes(trimmed)
if repairedBackslashes != trimmed {
if err := json.Unmarshal([]byte(repairedBackslashes), &parsed); err == nil {
return parsed, true
}
}
repairedLoose := RepairLooseJSON(trimmed)
if repairedLoose != trimmed {
if err := json.Unmarshal([]byte(repairedLoose), &parsed); err == nil {
return parsed, true
}
}
if strings.Contains(trimmed, "<") && strings.Contains(trimmed, ">") {
if parsedXML, ok := parseXMLFragmentValue(trimmed); ok {
return parsedXML, true
}
}
return nil, false
}
func coerceArrayValue(value any, paramName string) ([]any, bool) {
switch x := value.(type) {
case []any:
return x, true
case map[string]any:
if len(x) != 1 {
return nil, false
}
if items, ok := x["item"]; ok {
if arr, ok := coerceArrayValue(items, ""); ok {
return arr, true
}
return []any{items}, true
}
if paramName != "" {
if wrapped, ok := x[paramName]; ok {
if arr, ok := coerceArrayValue(wrapped, ""); ok {
return arr, true
}
}
}
}
return nil, false
}
func splitTopLevelJSONValues(raw string) ([]string, bool) {
trimmed := strings.TrimSpace(raw)
if trimmed == "" {
return nil, false
}
values := make([]string, 0, 2)
start := 0
depth := 0
inString := false
escaped := false
for i, r := range trimmed {
if inString {
if escaped {
escaped = false
continue
}
switch r {
case '\\':
escaped = true
case '"':
inString = false
}
continue
}
switch r {
case '"':
inString = true
case '{', '[':
depth++
case '}', ']':
if depth > 0 {
depth--
}
case ',':
if depth == 0 {
segment := strings.TrimSpace(trimmed[start:i])
if segment == "" {
return nil, false
}
values = append(values, segment)
start = i + 1
}
}
}
last := strings.TrimSpace(trimmed[start:])
if last == "" {
return nil, false
}
values = append(values, last)
if len(values) < 2 {
return nil, false
}
return values, true
}

View File

@@ -44,6 +44,9 @@ func rewriteDSMLToolMarkupOutsideIgnored(text string) string {
}
b.WriteString(tag.Name)
b.WriteString(text[tag.NameEnd : tag.End+1])
if text[tag.End] != '>' {
b.WriteByte('>')
}
i = tag.End + 1
continue
}

View File

@@ -298,11 +298,17 @@ func parseInvokeParameterValue(paramName, raw string) any {
}
if value, ok := extractStandaloneCDATA(trimmed); ok {
if parsed, ok := parseJSONLiteralValue(value); ok {
if parsedArray, ok := coerceArrayValue(parsed, paramName); ok {
return parsedArray
}
return parsed
}
if parsed, ok := parseStructuredCDATAParameterValue(paramName, value); ok {
return parsed
}
if parsed, ok := parseLooseJSONArrayValue(value, paramName); ok {
return parsed
}
return value
}
decoded := html.UnescapeString(extractRawTagValue(trimmed))
@@ -311,6 +317,9 @@ func parseInvokeParameterValue(paramName, raw string) any {
switch v := parsedValue.(type) {
case map[string]any:
if len(v) > 0 {
if parsedArray, ok := coerceArrayValue(v, paramName); ok {
return parsedArray
}
return v
}
case []any:
@@ -321,6 +330,12 @@ func parseInvokeParameterValue(paramName, raw string) any {
return ""
}
if parsedText, ok := parseJSONLiteralValue(text); ok {
if parsedArray, ok := coerceArrayValue(parsedText, paramName); ok {
return parsedArray
}
return parsedText
}
if parsedText, ok := parseLooseJSONArrayValue(text, paramName); ok {
return parsedText
}
return v
@@ -331,13 +346,25 @@ func parseInvokeParameterValue(paramName, raw string) any {
if parsed := parseStructuredToolCallInput(decoded); len(parsed) > 0 {
if len(parsed) == 1 {
if rawValue, ok := parsed["_raw"].(string); ok {
if parsedText, ok := parseLooseJSONArrayValue(rawValue, paramName); ok {
return parsedText
}
return rawValue
}
}
if parsedArray, ok := coerceArrayValue(parsed, paramName); ok {
return parsedArray
}
return parsed
}
}
if parsed, ok := parseJSONLiteralValue(decoded); ok {
if parsedArray, ok := coerceArrayValue(parsed, paramName); ok {
return parsedArray
}
return parsed
}
if parsed, ok := parseLooseJSONArrayValue(decoded, paramName); ok {
return parsed
}
return decoded

View File

@@ -128,34 +128,39 @@ func scanToolMarkupTagAt(text string, start int) (ToolMarkupTag, bool) {
}
lower := strings.ToLower(text)
i := start + 1
for i < len(text) && text[i] == '<' {
i++
}
closing := false
if i < len(text) && text[i] == '/' {
closing = true
i++
}
dsmlLike := false
if next, ok := consumeToolMarkupPipe(text, i); ok {
dsmlLike = true
i = next
}
if strings.HasPrefix(lower[i:], "dsml") {
dsmlLike = true
i += len("dsml")
for next, ok := consumeToolMarkupSeparator(text, i); ok; next, ok = consumeToolMarkupSeparator(text, i) {
i = next
}
}
i, dsmlLike := consumeToolMarkupNamePrefix(lower, text, i)
name, nameLen := matchToolMarkupName(lower, i)
if nameLen == 0 {
return ToolMarkupTag{}, false
}
nameEnd := i + nameLen
nameEndBeforePipes := nameEnd
for next, ok := consumeToolMarkupPipe(text, nameEnd); ok; next, ok = consumeToolMarkupPipe(text, nameEnd) {
nameEnd = next
}
hasTrailingPipe := nameEnd > nameEndBeforePipes
if !hasToolMarkupBoundary(text, nameEnd) {
return ToolMarkupTag{}, false
}
end := findXMLTagEnd(text, nameEnd)
if end < 0 {
return ToolMarkupTag{}, false
if !hasTrailingPipe {
return ToolMarkupTag{}, false
}
end = nameEnd - 1
}
if hasTrailingPipe {
if nextLT := strings.IndexByte(text[nameEnd:], '<'); nextLT >= 0 && end >= nameEnd+nextLT {
end = nameEnd - 1
}
}
trimmed := strings.TrimSpace(text[start : end+1])
return ToolMarkupTag{
@@ -171,6 +176,74 @@ func scanToolMarkupTagAt(text string, start int) (ToolMarkupTag, bool) {
}, true
}
func IsPartialToolMarkupTagPrefix(text string) bool {
if text == "" || text[0] != '<' || strings.Contains(text, ">") {
return false
}
lower := strings.ToLower(text)
i := 1
for i < len(text) && text[i] == '<' {
i++
}
if i >= len(text) {
return true
}
if text[i] == '/' {
i++
}
for i <= len(text) {
if i == len(text) {
return true
}
if hasToolMarkupNamePrefix(lower[i:]) {
return true
}
if strings.HasPrefix("dsml", lower[i:]) {
return true
}
next, ok := consumeToolMarkupNamePrefixOnce(lower, text, i)
if !ok {
return false
}
i = next
}
return false
}
func consumeToolMarkupNamePrefix(lower, text string, idx int) (int, bool) {
dsmlLike := false
for {
next, ok := consumeToolMarkupNamePrefixOnce(lower, text, idx)
if !ok {
return idx, dsmlLike
}
idx = next
dsmlLike = true
}
}
func consumeToolMarkupNamePrefixOnce(lower, text string, idx int) (int, bool) {
if next, ok := consumeToolMarkupPipe(text, idx); ok {
return next, true
}
if idx < len(text) && (text[idx] == ' ' || text[idx] == '\t' || text[idx] == '\r' || text[idx] == '\n') {
return idx + 1, true
}
if strings.HasPrefix(lower[idx:], "dsml") {
return idx + len("dsml"), true
}
return idx, false
}
func hasToolMarkupNamePrefix(lowerTail string) bool {
for _, name := range toolMarkupNames {
if strings.HasPrefix(lowerTail, name) || strings.HasPrefix(name, lowerTail) {
return true
}
}
return false
}
func matchToolMarkupName(lower string, start int) (string, int) {
for _, name := range toolMarkupNames {
if strings.HasPrefix(lower[start:], name) {
@@ -193,19 +266,6 @@ func consumeToolMarkupPipe(text string, idx int) (int, bool) {
return idx, false
}
func consumeToolMarkupSeparator(text string, idx int) (int, bool) {
if idx >= len(text) {
return idx, false
}
if text[idx] == ' ' || text[idx] == '\t' || text[idx] == '\r' || text[idx] == '\n' {
return idx + 1, true
}
if next, ok := consumeToolMarkupPipe(text, idx); ok {
return next, true
}
return idx, false
}
func hasToolMarkupBoundary(text string, idx int) bool {
if idx >= len(text) {
return true

View File

@@ -41,6 +41,52 @@ func TestParseToolCallsSupportsDSMLShell(t *testing.T) {
}
}
func TestParseToolCallsToleratesDSMLTrailingPipeTagTerminator(t *testing.T) {
text := strings.Join([]string{
`<|DSML|tool_calls| `,
` <|DSML|invoke name="terminal">`,
` <|DSML|parameter name="command"><![CDATA[find "/home" -type d]]></|DSML|parameter>`,
` <|DSML|parameter name="timeout"><![CDATA[10]]></|DSML|parameter>`,
` </|DSML|invoke>`,
`</|DSML|tool_calls>`,
}, "\n")
calls := ParseToolCalls(text, []string{"terminal"})
if len(calls) != 1 {
t.Fatalf("expected one trailing-pipe DSML call, got %#v", calls)
}
if calls[0].Name != "terminal" {
t.Fatalf("expected terminal tool, got %#v", calls[0])
}
if calls[0].Input["command"] != `find "/home" -type d` {
t.Fatalf("expected command argument, got %#v", calls[0].Input)
}
if calls[0].Input["timeout"] != float64(10) {
t.Fatalf("expected numeric timeout, got %#v", calls[0].Input)
}
}
func TestParseToolCallsToleratesExtraLeadingLessThanBeforeDSML(t *testing.T) {
text := `<<|DSML|tool_calls><<|DSML|invoke name="Bash"><<|DSML|parameter name="command"><![CDATA[pwd]]></|DSML|parameter></|DSML|invoke></|DSML|tool_calls>`
calls := ParseToolCalls(text, []string{"Bash"})
if len(calls) != 1 {
t.Fatalf("expected one extra-leading-less-than DSML call, got %#v", calls)
}
if calls[0].Name != "Bash" || calls[0].Input["command"] != "pwd" {
t.Fatalf("unexpected extra-leading-less-than DSML parse result: %#v", calls[0])
}
}
func TestParseToolCallsToleratesRepeatedDSMLPrefixNoise(t *testing.T) {
text := `<<DSML|DSML|tool_calls><<DSML|DSML|invoke name="Bash"><<DSML|DSML|parameter name="command"><![CDATA[git status]]></DSML|DSML|parameter></DSML|DSML|invoke></DSML|DSML|tool_calls>`
calls := ParseToolCalls(text, []string{"Bash"})
if len(calls) != 1 {
t.Fatalf("expected one repeated-prefix DSML call, got %#v", calls)
}
if calls[0].Name != "Bash" || calls[0].Input["command"] != "git status" {
t.Fatalf("unexpected repeated-prefix DSML parse result: %#v", calls[0])
}
}
func TestParseToolCallsSupportsDSMLShellWithCanonicalExampleInCDATA(t *testing.T) {
content := `<tool_calls><invoke name="demo"><parameter name="value">x</parameter></invoke></tool_calls>`
text := `<|DSML|tool_calls><|DSML|invoke name="Write"><|DSML|parameter name="file_path">notes.md</|DSML|parameter><|DSML|parameter name="content"><![CDATA[` + content + `]]></|DSML|parameter></|DSML|invoke></|DSML|tool_calls>`
@@ -248,6 +294,59 @@ func TestParseToolCallsTreatsSingleItemCDATAAsArray(t *testing.T) {
}
}
func TestParseToolCallsTreatsLooseJSONListAsArray(t *testing.T) {
tests := []struct {
name string
body string
}{
{
name: "plain text",
body: `{"content":"Test TodoWrite tool","status":"completed"}, {"content":"Another task","status":"pending"}`,
},
{
name: "cdata",
body: `<![CDATA[{"content":"Test TodoWrite tool","status":"completed"}, {"content":"Another task","status":"pending"}]]>`,
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
text := `<tool_calls><invoke name="TodoWrite"><parameter name="todos">` + tt.body + `</parameter></invoke></tool_calls>`
calls := ParseToolCalls(text, []string{"TodoWrite"})
if len(calls) != 1 {
t.Fatalf("expected one TodoWrite call, got %#v", calls)
}
items, ok := calls[0].Input["todos"].([]any)
if !ok || len(items) != 2 {
t.Fatalf("expected loose JSON list to parse as array, got %#v", calls[0].Input["todos"])
}
first, ok := items[0].(map[string]any)
if !ok {
t.Fatalf("expected first todo object, got %#v", items[0])
}
if first["content"] != "Test TodoWrite tool" || first["status"] != "completed" {
t.Fatalf("unexpected first todo: %#v", first)
}
})
}
}
func TestParseToolCallsKeepsPreservedTextParametersAsText(t *testing.T) {
text := `<tool_calls><invoke name="Write"><parameter name="content"><![CDATA[{"content":"Test TodoWrite tool","status":"completed"}, {"content":"Another task","status":"pending"}]]></parameter></invoke></tool_calls>`
calls := ParseToolCalls(text, []string{"Write"})
if len(calls) != 1 {
t.Fatalf("expected one Write call, got %#v", calls)
}
got, ok := calls[0].Input["content"].(string)
if !ok {
t.Fatalf("expected content to stay a string, got %#v", calls[0].Input["content"])
}
want := `{"content":"Test TodoWrite tool","status":"completed"}, {"content":"Another task","status":"pending"}`
if got != want {
t.Fatalf("expected content to stay raw, got %q", got)
}
}
func TestParseToolCallsTreatsCDATAObjectFragmentAsObject(t *testing.T) {
payload := `<question><![CDATA[Pick one]]></question><options><item><label><![CDATA[A]]></label></item><item><label><![CDATA[B]]></label></item></options>`
text := `<tool_calls><invoke name="AskUserQuestion"><parameter name="questions"><![CDATA[` + payload + `]]></parameter></invoke></tool_calls>`

View File

@@ -1,10 +1,6 @@
package toolstream
import (
"strings"
"ds2api/internal/toolcall"
)
import "ds2api/internal/toolcall"
func ProcessChunk(state *State, chunk string, toolNames []string) []Event {
if state == nil {
@@ -174,31 +170,27 @@ func findToolSegmentStart(state *State, s string) int {
if s == "" {
return -1
}
lower := strings.ToLower(s)
offset := 0
for {
bestKeyIdx := -1
matchedTag := ""
for _, tag := range xmlToolTagsToDetect {
idx := strings.Index(lower[offset:], tag)
if idx >= 0 {
idx += offset
if bestKeyIdx < 0 || idx < bestKeyIdx {
bestKeyIdx = idx
matchedTag = tag
}
}
}
if bestKeyIdx < 0 {
tag, ok := toolcall.FindToolMarkupTagOutsideIgnored(s, offset)
if !ok {
return -1
}
if !insideCodeFenceWithState(state, s[:bestKeyIdx]) {
return bestKeyIdx
start := includeDuplicateLeadingLessThan(s, tag.Start)
if !insideCodeFenceWithState(state, s[:start]) {
return start
}
offset = bestKeyIdx + len(matchedTag)
offset = tag.End + 1
}
}
func includeDuplicateLeadingLessThan(s string, idx int) int {
for idx > 0 && s[idx-1] == '<' {
idx--
}
return idx
}
func consumeToolCapture(state *State, toolNames []string) (prefix string, calls []toolcall.ParsedToolCall, suffix string, ready bool) {
captured := state.capture.String()
if captured == "" {

View File

@@ -153,27 +153,14 @@ func findPartialXMLToolTagStart(s string) int {
if lastLT < 0 {
return -1
}
tail := s[lastLT:]
start := includeDuplicateLeadingLessThan(s, lastLT)
tail := s[start:]
// If there's a '>' in the tail, the tag is closed — not partial.
if strings.Contains(tail, ">") {
return -1
}
lowerTail := strings.ToLower(tail)
for _, tag := range []string{
"<tool_calls", "<invoke", "<parameter",
"<|tool_calls", "<|invoke", "<|parameter",
"<tool_calls", "<invoke", "<parameter",
"<|dsml|tool_calls", "<|dsml|invoke", "<|dsml|parameter",
"<dsml|tool_calls", "<dsml|invoke", "<dsml|parameter",
"<dsmltool_calls", "<dsmlinvoke", "<dsmlparameter",
"<dsml tool_calls", "<dsml invoke", "<dsml parameter",
"<dsml|tool_calls", "<dsml|invoke", "<dsml|parameter",
"<|dsmltool_calls", "<|dsmlinvoke", "<|dsmlparameter",
"<|dsml tool_calls", "<|dsml invoke", "<|dsml parameter",
} {
if strings.HasPrefix(tag, lowerTail) {
return lastLT
}
if toolcall.IsPartialToolMarkupTagPrefix(tail) {
return start
}
return -1
}

View File

@@ -1,35 +0,0 @@
package toolstream
import "regexp"
// --- XML tool call support for the streaming sieve ---
//nolint:unused // kept as explicit tag inventory for future XML sieve refinements.
var xmlToolCallClosingTags = []string{"</tool_calls>", "</|dsml|tool_calls>", "</|dsmltool_calls>", "</|dsml tool_calls>", "</dsml|tool_calls>", "</dsmltool_calls>", "</dsml tool_calls>", "</tool_calls>", "</|tool_calls>"}
// xmlToolCallBlockPattern matches a complete canonical XML tool call block.
//
//nolint:unused // reserved for future fast-path XML block detection.
var xmlToolCallBlockPattern = regexp.MustCompile(`(?is)((?:<tool_calls\b|<\|dsml\|tool_calls\b)[^>]*>\s*(?:.*?)\s*(?:</tool_calls>|</\|dsml\|tool_calls>))`)
// xmlToolTagsToDetect is the set of XML tag prefixes used by findToolSegmentStart.
var xmlToolTagsToDetect = []string{
"<|dsml|tool_calls>", "<|dsml|tool_calls\n", "<|dsml|tool_calls ",
"<dsml|tool_calls>", "<dsml|tool_calls\n", "<dsml|tool_calls ",
"<|dsml|invoke ", "<|dsml|invoke\n", "<|dsml|invoke\t", "<|dsml|invoke\r",
"<|dsmltool_calls>", "<|dsmltool_calls\n", "<|dsmltool_calls ",
"<|dsmlinvoke ", "<|dsmlinvoke\n", "<|dsmlinvoke\t", "<|dsmlinvoke\r",
"<|dsml tool_calls>", "<|dsml tool_calls\n", "<|dsml tool_calls ",
"<|dsml invoke ", "<|dsml invoke\n", "<|dsml invoke\t", "<|dsml invoke\r",
"<dsml|tool_calls>", "<dsml|tool_calls\n", "<dsml|tool_calls ",
"<dsml|invoke ", "<dsml|invoke\n", "<dsml|invoke\t", "<dsml|invoke\r",
"<dsmltool_calls>", "<dsmltool_calls\n", "<dsmltool_calls ",
"<dsmlinvoke ", "<dsmlinvoke\n", "<dsmlinvoke\t", "<dsmlinvoke\r",
"<dsml tool_calls>", "<dsml tool_calls\n", "<dsml tool_calls ",
"<dsml invoke ", "<dsml invoke\n", "<dsml invoke\t", "<dsml invoke\r",
"<tool_calls>", "<tool_calls\n", "<tool_calls ",
"<invoke ", "<invoke\n", "<invoke\t", "<invoke\r",
"<|tool_calls>", "<|tool_calls\n", "<|tool_calls ",
"<|invoke ", "<|invoke\n", "<|invoke\t", "<|invoke\r",
"<tool_calls>", "<tool_calls\n", "<tool_calls ", "<invoke ", "<invoke\n", "<invoke\t", "<invoke\r",
}

View File

@@ -72,6 +72,97 @@ func TestProcessToolSieveInterceptsDSMLToolCallWithoutLeak(t *testing.T) {
}
}
func TestProcessToolSieveInterceptsDSMLTrailingPipeToolCallWithoutLeak(t *testing.T) {
var state State
chunks := []string{
"<|DSML|tool_calls| \n",
` <|DSML|invoke name="terminal">` + "\n",
` <|DSML|parameter name="command"><![CDATA[find "/home" -type d]]></|DSML|parameter>` + "\n",
` <|DSML|parameter name="timeout"><![CDATA[10]]></|DSML|parameter>` + "\n",
" </|DSML|invoke>\n",
"</|DSML|tool_calls>",
}
var events []Event
for _, c := range chunks {
events = append(events, ProcessChunk(&state, c, []string{"terminal"})...)
}
events = append(events, Flush(&state, []string{"terminal"})...)
var textContent strings.Builder
var calls []any
for _, evt := range events {
textContent.WriteString(evt.Content)
for _, call := range evt.ToolCalls {
calls = append(calls, call)
}
}
if text := textContent.String(); strings.Contains(strings.ToLower(text), "dsml") || strings.Contains(text, "terminal") {
t.Fatalf("trailing-pipe DSML tool call leaked to text: %q events=%#v", text, events)
}
if len(calls) != 1 {
t.Fatalf("expected one trailing-pipe DSML tool call, got %d events=%#v", len(calls), events)
}
}
func TestProcessToolSieveInterceptsExtraLeadingLessThanDSMLToolCallWithoutLeak(t *testing.T) {
var state State
chunks := []string{
"<<|DSML|tool_calls>\n",
` <<|DSML|invoke name="Bash">` + "\n",
` <<|DSML|parameter name="command"><![CDATA[pwd]]></|DSML|parameter>` + "\n",
" </|DSML|invoke>\n",
"</|DSML|tool_calls>",
}
var events []Event
for _, c := range chunks {
events = append(events, ProcessChunk(&state, c, []string{"Bash"})...)
}
events = append(events, Flush(&state, []string{"Bash"})...)
var textContent strings.Builder
toolCalls := 0
for _, evt := range events {
textContent.WriteString(evt.Content)
toolCalls += len(evt.ToolCalls)
}
if text := textContent.String(); strings.Contains(text, "<") || strings.Contains(text, "Bash") {
t.Fatalf("extra-leading-less-than DSML tool call leaked to text: %q events=%#v", text, events)
}
if toolCalls != 1 {
t.Fatalf("expected one extra-leading-less-than DSML tool call, got %d events=%#v", toolCalls, events)
}
}
func TestProcessToolSieveInterceptsRepeatedDSMLPrefixNoiseWithoutLeak(t *testing.T) {
var state State
chunks := []string{
"<<DSML|DSML|tool",
"_calls>\n",
` <<DSML|DSML|invoke name="Bash">` + "\n",
` <<DSML|DSML|parameter name="command"><![CDATA[git status]]></DSML|DSML|parameter>` + "\n",
" </DSML|DSML|invoke>\n",
"</DSML|DSML|tool_calls>",
}
var events []Event
for _, c := range chunks {
events = append(events, ProcessChunk(&state, c, []string{"Bash"})...)
}
events = append(events, Flush(&state, []string{"Bash"})...)
var textContent strings.Builder
toolCalls := 0
for _, evt := range events {
textContent.WriteString(evt.Content)
toolCalls += len(evt.ToolCalls)
}
if text := textContent.String(); strings.Contains(strings.ToLower(text), "dsml") || strings.Contains(text, "Bash") {
t.Fatalf("repeated-prefix DSML tool call leaked to text: %q events=%#v", text, events)
}
if toolCalls != 1 {
t.Fatalf("expected one repeated-prefix DSML tool call, got %d events=%#v", toolCalls, events)
}
}
func TestProcessToolSieveHandlesLongXMLToolCall(t *testing.T) {
var state State
const toolName = "write_to_file"
@@ -442,6 +533,8 @@ func TestFindToolSegmentStartDetectsXMLToolCalls(t *testing.T) {
want int
}{
{"tool_calls_tag", "some text <tool_calls>\n", 10},
{"dsml_trailing_pipe_tag", "some text <|DSML|tool_calls| \n", 10},
{"dsml_extra_leading_less_than", "some text <<|DSML|tool_calls>\n", 10},
{"invoke_tag_missing_wrapper", "some text <invoke name=\"read_file\">\n", 10},
{"bare_tool_call_text", "prefix <tool_call>\n", -1},
{"xml_inside_code_fence", "```xml\n<tool_calls><invoke name=\"read_file\"></invoke></tool_calls>\n```", -1},
@@ -465,6 +558,8 @@ func TestFindPartialXMLToolTagStart(t *testing.T) {
want int
}{
{"partial_tool_calls", "Hello <tool_ca", 6},
{"partial_dsml_trailing_pipe", "Hello <|DSML|tool_calls|", 6},
{"partial_dsml_extra_leading_less_than", "Hello <<|DSML|tool_calls", 6},
{"partial_invoke", "Hello <inv", 6},
{"bare_tool_call_not_held", "Hello <tool_name", -1},
{"partial_lt_only", "Text <", 5},

View File

@@ -1,6 +1,7 @@
package util
import (
"strings"
"testing"
"ds2api/internal/config"
@@ -12,7 +13,10 @@ func TestMessagesPrepareBasic(t *testing.T) {
if got == "" {
t.Fatal("expected non-empty prompt")
}
if got != "<begin▁of▁sentence><User>Hello<Assistant>" {
if !strings.HasPrefix(got, "<begin▁of▁sentence><System>") {
t.Fatalf("expected output integrity guard at the start, got %q", got)
}
if !strings.Contains(got, "Hello") || !strings.HasSuffix(got, "<Assistant>") {
t.Fatalf("unexpected prompt: %q", got)
}
}
@@ -26,8 +30,11 @@ func TestMessagesPrepareRoles(t *testing.T) {
{"role": "user", "content": "How are you"},
}
got := MessagesPrepare(messages)
if !contains(got, "<System>You are helper<end▁of▁instructions><User>Hi") {
t.Fatalf("expected system/user separation in %q", got)
if !contains(got, "Output integrity guard") {
t.Fatalf("expected output integrity guard in %q", got)
}
if !contains(got, "You are helper") || !contains(got, "<User>Hi") {
t.Fatalf("expected system/user content in %q", got)
}
if !contains(got, "<begin▁of▁sentence>") {
t.Fatalf("expected begin marker in %q", got)
@@ -77,9 +84,12 @@ func TestMessagesPrepareArrayTextVariants(t *testing.T) {
},
}
got := MessagesPrepare(messages)
if got != "<begin▁of▁sentence><User>line1\nline2<Assistant>" {
if !contains(got, "line1\nline2") {
t.Fatalf("unexpected content from text variants: %q", got)
}
if !strings.Contains(got, "Output integrity guard") {
t.Fatalf("expected output integrity guard in %q", got)
}
}
func TestConvertClaudeToDeepSeek(t *testing.T) {

View File

@@ -1,11 +1,5 @@
package util
import (
"strings"
tiktoken "github.com/hupe1980/go-tiktoken"
)
const (
defaultTokenizerModel = "gpt-4o"
claudeTokenizerModel = "claude"
@@ -33,41 +27,6 @@ func CountOutputTokens(text, model string) int {
return base
}
func countWithTokenizer(text, model string) int {
text = strings.TrimSpace(text)
if text == "" {
return 0
}
encoding, err := tiktoken.NewEncodingForModel(tokenizerModelForCount(model))
if err != nil {
return 0
}
ids, _, err := encoding.Encode(text, nil, nil)
if err != nil {
return 0
}
return len(ids)
}
func tokenizerModelForCount(model string) string {
model = strings.ToLower(strings.TrimSpace(model))
if model == "" {
return defaultTokenizerModel
}
switch {
case strings.HasPrefix(model, "claude"):
return claudeTokenizerModel
case strings.HasPrefix(model, "gpt-4"), strings.HasPrefix(model, "gpt-5"), strings.HasPrefix(model, "o1"), strings.HasPrefix(model, "o3"), strings.HasPrefix(model, "o4"):
return defaultTokenizerModel
case strings.HasPrefix(model, "deepseek-v4"):
return defaultTokenizerModel
case strings.HasPrefix(model, "deepseek"):
return defaultTokenizerModel
default:
return defaultTokenizerModel
}
}
func conservativePromptPadding(base int) int {
padding := base / 50
if padding < 4 {

View File

@@ -0,0 +1,7 @@
//go:build 386 || arm || mips || mipsle || wasm
package util
func countWithTokenizer(_, _ string) int {
return 0
}

View File

@@ -0,0 +1,98 @@
//go:build !386 && !arm && !mips && !mipsle && !wasm
package util
import (
"strings"
"sync"
tiktoken "github.com/hupe1980/go-tiktoken"
)
var (
tokenEncodingPools sync.Map
tokenEncodingUnsupported sync.Map
)
func countWithTokenizer(text, model string) int {
text = strings.TrimSpace(text)
if text == "" {
return 0
}
encoding, release := tokenizerEncodingForCount(tokenizerModelForCount(model))
if encoding == nil {
return 0
}
defer release()
ids, _, err := encoding.Encode(text, nil, nil)
if err != nil {
return 0
}
return len(ids)
}
func tokenizerEncodingForCount(model string) (*tiktoken.Encoding, func()) {
model = strings.TrimSpace(model)
if model == "" {
model = defaultTokenizerModel
}
if _, ok := tokenEncodingUnsupported.Load(model); ok {
return nil, func() {}
}
if rawPool, ok := tokenEncodingPools.Load(model); ok {
pool, _ := rawPool.(*sync.Pool)
return getEncodingFromPool(pool)
}
encoding, err := tiktoken.NewEncodingForModel(model)
if err != nil {
tokenEncodingUnsupported.Store(model, struct{}{})
return nil, func() {}
}
pool := &sync.Pool{
New: func() any {
encoding, err := tiktoken.NewEncodingForModel(model)
if err != nil {
return nil
}
return encoding
},
}
actualPool, _ := tokenEncodingPools.LoadOrStore(model, pool)
pool, _ = actualPool.(*sync.Pool)
return encoding, func() {
pool.Put(encoding)
}
}
func getEncodingFromPool(pool *sync.Pool) (*tiktoken.Encoding, func()) {
if pool == nil {
return nil, func() {}
}
encoding, _ := pool.Get().(*tiktoken.Encoding)
if encoding == nil {
return nil, func() {}
}
return encoding, func() {
pool.Put(encoding)
}
}
func tokenizerModelForCount(model string) string {
model = strings.ToLower(strings.TrimSpace(model))
if model == "" {
return defaultTokenizerModel
}
switch {
case strings.HasPrefix(model, "claude"):
return claudeTokenizerModel
case strings.HasPrefix(model, "gpt-4"), strings.HasPrefix(model, "gpt-5"), strings.HasPrefix(model, "o1"), strings.HasPrefix(model, "o3"), strings.HasPrefix(model, "o4"):
return defaultTokenizerModel
case strings.HasPrefix(model, "deepseek-v4"):
return defaultTokenizerModel
case strings.HasPrefix(model, "deepseek"):
return defaultTokenizerModel
default:
return defaultTokenizerModel
}
}

View File

@@ -0,0 +1,35 @@
//go:build !386 && !arm && !mips && !mipsle && !wasm
package util
import "testing"
func TestTokenizerEncodingForCountCachesSupportedModel(t *testing.T) {
encoding, release := tokenizerEncodingForCount(defaultTokenizerModel)
if encoding == nil {
t.Fatalf("expected tokenizer encoding for %q", defaultTokenizerModel)
}
release()
if _, ok := tokenEncodingPools.Load(defaultTokenizerModel); !ok {
t.Fatalf("expected tokenizer encoding pool for %q", defaultTokenizerModel)
}
encoding, release = tokenizerEncodingForCount(defaultTokenizerModel)
if encoding == nil {
t.Fatalf("expected cached tokenizer encoding for %q", defaultTokenizerModel)
}
release()
}
func TestTokenizerEncodingForCountCachesUnsupportedModel(t *testing.T) {
const model = "__ds2api_unsupported_tokenizer_model__"
encoding, release := tokenizerEncodingForCount(model)
release()
if encoding != nil {
t.Fatalf("expected nil encoding for unsupported model %q", model)
}
if _, ok := tokenEncodingUnsupported.Load(model); !ok {
t.Fatalf("expected unsupported tokenizer model to be cached")
}
}

View File

@@ -1,23 +0,0 @@
# Node split syntax gate targets
# Keep this list in sync with api/chat-stream and internal/js/helpers/stream-tool-sieve split modules.
api/chat-stream.js
internal/js/chat-stream/index.js
internal/js/chat-stream/error_shape.js
internal/js/chat-stream/http_internal.js
internal/js/chat-stream/proxy_go.js
internal/js/chat-stream/sse_parse.js
internal/js/chat-stream/stream_emitter.js
internal/js/chat-stream/token_usage.js
internal/js/chat-stream/toolcall_policy.js
internal/js/chat-stream/vercel_stream.js
internal/js/helpers/stream-tool-sieve.js
internal/js/helpers/stream-tool-sieve/index.js
internal/js/helpers/stream-tool-sieve/state.js
internal/js/helpers/stream-tool-sieve/sieve.js
internal/js/helpers/stream-tool-sieve/sieve-xml.js
internal/js/helpers/stream-tool-sieve/jsonscan.js
internal/js/helpers/stream-tool-sieve/parse.js
internal/js/helpers/stream-tool-sieve/format.js
internal/js/helpers/stream-tool-sieve/tool-keywords.js

View File

@@ -1,149 +0,0 @@
# Line gate targets for large-file decoupling refactor.
# Backend default limit: 300 lines
# Frontend (webui/) default limit: 500 lines
# Entry/facade limit: 120 lines (enforced in script)
# Test files are ignored by the gate script.
internal/config/config.go
internal/config/logger.go
internal/config/paths.go
internal/config/codec.go
internal/config/store.go
internal/config/store_index.go
internal/config/store_accessors.go
internal/config/account.go
internal/httpapi/admin/configmgmt/handler_config_read.go
internal/httpapi/admin/configmgmt/handler_config_write.go
internal/httpapi/admin/configmgmt/handler_config_import.go
internal/httpapi/admin/settings/handler_settings_read.go
internal/httpapi/admin/settings/handler_settings_write.go
internal/httpapi/admin/settings/handler_settings_parse.go
internal/httpapi/admin/settings/handler_settings_runtime.go
internal/httpapi/admin/accounts/handler_accounts_crud.go
internal/httpapi/admin/accounts/handler_accounts_testing.go
internal/httpapi/admin/accounts/handler_accounts_queue.go
internal/account/pool_core.go
internal/account/pool_acquire.go
internal/account/pool_waiters.go
internal/account/pool_limits.go
internal/deepseek/client/client_core.go
internal/deepseek/client/client_auth.go
internal/deepseek/client/client_completion.go
internal/deepseek/client/client_http_json.go
internal/deepseek/client/client_http_helpers.go
internal/format/openai/render_chat.go
internal/format/openai/render_responses.go
internal/format/openai/render_stream_events.go
internal/format/openai/render_usage.go
internal/httpapi/openai/shared/models.go
internal/httpapi/openai/chat/handler_chat.go
internal/httpapi/openai/shared/handler_errors.go
internal/httpapi/openai/shared/handler_toolcall_policy.go
internal/httpapi/openai/shared/handler_toolcall_format.go
internal/httpapi/openai/responses/responses_handler.go
internal/promptcompat/responses_input_normalize.go
internal/promptcompat/responses_input_items.go
internal/httpapi/openai/responses/responses_stream_runtime_core.go
internal/httpapi/openai/responses/responses_stream_runtime_events.go
internal/httpapi/openai/responses/responses_stream_runtime_toolcalls.go
internal/toolstream/tool_sieve_state.go
internal/toolstream/tool_sieve_core.go
internal/toolstream/tool_sieve_xml.go
internal/toolstream/tool_sieve_jsonscan.go
internal/toolcall/toolcalls_parse.go
internal/toolcall/toolcalls_candidates.go
internal/toolcall/toolcalls_format.go
internal/httpapi/claude/handler_routes.go
internal/httpapi/claude/handler_messages.go
internal/httpapi/claude/handler_tokens.go
internal/httpapi/claude/handler_errors.go
internal/httpapi/claude/handler_utils.go
internal/httpapi/claude/stream_runtime_core.go
internal/httpapi/claude/stream_runtime_emit.go
internal/httpapi/claude/stream_runtime_finalize.go
internal/httpapi/gemini/handler_routes.go
internal/httpapi/gemini/handler_generate.go
internal/httpapi/gemini/handler_stream_runtime.go
internal/httpapi/gemini/handler_errors.go
internal/httpapi/gemini/convert_request.go
internal/httpapi/gemini/convert_messages.go
internal/httpapi/gemini/convert_tools.go
internal/httpapi/gemini/convert_passthrough.go
internal/testsuite/runner_core.go
internal/testsuite/runner_env.go
internal/testsuite/runner_http.go
internal/testsuite/runner_cases_openai.go
internal/testsuite/runner_cases_openai_advanced.go
internal/testsuite/runner_cases_admin.go
internal/testsuite/runner_cases_claude.go
internal/testsuite/runner_summary.go
internal/testsuite/runner_utils.go
internal/testsuite/runner_defaults.go
internal/testsuite/runner_registry.go
internal/testsuite/edge_cases_abort.go
internal/testsuite/edge_cases_error_contract.go
api/chat-stream.js
internal/js/chat-stream/index.js
internal/js/chat-stream/vercel_stream.js
internal/js/chat-stream/proxy_go.js
internal/js/chat-stream/sse_parse.js
internal/js/chat-stream/http_internal.js
internal/js/chat-stream/toolcall_policy.js
internal/js/chat-stream/error_shape.js
internal/js/chat-stream/token_usage.js
internal/js/chat-stream/stream_emitter.js
internal/js/helpers/stream-tool-sieve.js
internal/js/helpers/stream-tool-sieve/index.js
internal/js/helpers/stream-tool-sieve/state.js
internal/js/helpers/stream-tool-sieve/sieve.js
internal/js/helpers/stream-tool-sieve/sieve-xml.js
internal/js/helpers/stream-tool-sieve/jsonscan.js
internal/js/helpers/stream-tool-sieve/parse.js
internal/js/helpers/stream-tool-sieve/format.js
webui/src/App.jsx
webui/src/app/AppRoutes.jsx
webui/src/app/useAdminAuth.js
webui/src/app/useAdminConfig.js
webui/src/layout/DashboardShell.jsx
webui/src/features/account/AccountManagerContainer.jsx
webui/src/features/account/useAccountsData.js
webui/src/features/account/useAccountActions.js
webui/src/features/account/QueueCards.jsx
webui/src/features/account/ApiKeysPanel.jsx
webui/src/features/account/AccountsTable.jsx
webui/src/features/account/AddKeyModal.jsx
webui/src/features/account/AddAccountModal.jsx
webui/src/features/apiTester/ApiTesterContainer.jsx
webui/src/features/apiTester/useApiTesterState.js
webui/src/features/apiTester/useChatStreamClient.js
webui/src/features/apiTester/ConfigPanel.jsx
webui/src/features/apiTester/ChatPanel.jsx
webui/src/features/settings/SettingsContainer.jsx
webui/src/features/settings/useSettingsForm.js
webui/src/features/settings/settingsApi.js
webui/src/features/settings/SecuritySection.jsx
webui/src/features/settings/RuntimeSection.jsx
webui/src/features/settings/BehaviorSection.jsx
webui/src/features/settings/ModelSection.jsx
webui/src/features/settings/BackupSection.jsx
webui/src/features/vercel/VercelSyncContainer.jsx
webui/src/features/vercel/useVercelSyncState.js
webui/src/features/vercel/VercelSyncForm.jsx
webui/src/features/vercel/VercelSyncStatus.jsx
webui/src/features/vercel/VercelGuide.jsx

View File

@@ -1,28 +0,0 @@
# Stage 6 Manual Smoke Checklist
- Date: 2026-02-22
- Tester: release-maintainer
- Environment: local macOS + latest Chrome
## Items
1. Login flow (`/admin/login`) succeeds and failure message shape unchanged.
2. Account manager:
- add/edit/delete account
- queue status cards render and refresh
3. API tester:
- non-stream request succeeds
- stream request receives incremental output and final state
4. Settings:
- read settings
- save settings
- backup/export path works
5. Vercel sync:
- status poll
- manual refresh
- sync action and status feedback text
## Result
- Status: `PASS`
- Notes: login/account/api-tester/settings/vercel-sync smoke passed with no behavior regressions.

View File

@@ -258,6 +258,28 @@ test('vercel stream exhausts DeepSeek continue before synthetic retry', async ()
assert.equal(fetchBodies.some((body) => String(body.prompt || '').includes('Previous reply had no visible output')), false);
});
test('vercel stream continues direct quasi_status incomplete before final tool call', async () => {
const { frames, fetchURLs } = await runMockVercelStreamSequence([
[
'data: {"response_message_id":7,"p":"response/content","v":"<tool_calls><invoke name=\\"write_file\\"><parameter name=\\"content\\"><![CDATA[part-one"}\n\n',
'data: {"p":"response/quasi_status","v":"INCOMPLETE"}\n\n',
'data: [DONE]\n\n',
],
[
'data: {"response_message_id":8,"p":"response/content","v":"-part-two]]></parameter></invoke></tool_calls>"}\n\n',
'data: {"p":"response/status","v":"FINISHED"}\n\n',
'data: [DONE]\n\n',
],
], { tool_names: ['write_file'] });
const parsed = frames.filter((frame) => frame !== '[DONE]').map((frame) => JSON.parse(frame));
const toolDelta = parsed.find((item) => item.choices?.[0]?.delta?.tool_calls);
assert.equal(fetchURLs.filter((url) => url === 'https://chat.deepseek.com/api/v0/chat/continue').length, 1);
assert.ok(toolDelta);
const args = JSON.parse(toolDelta.choices[0].delta.tool_calls[0].function.arguments);
assert.equal(args.content, 'part-one-part-two');
assert.equal(parsed.at(-1).choices[0].finish_reason, 'tool_calls');
});
test('vercel stream usage completion_tokens does not double-count visible output', async () => {

View File

@@ -57,6 +57,49 @@ test('parseToolCalls parses DSML shell as XML-compatible tool call', () => {
assert.deepEqual(calls[0].input, { path: 'README.MD' });
});
test('parseToolCalls tolerates DSML trailing pipe tag terminator', () => {
const payload = [
'<|DSML|tool_calls| ',
' <|DSML|invoke name="terminal">',
' <|DSML|parameter name="command"><![CDATA[find "/home" -type d]]></|DSML|parameter>',
' <|DSML|parameter name="timeout"><![CDATA[10]]></|DSML|parameter>',
' </|DSML|invoke>',
'</|DSML|tool_calls>',
].join('\n');
const calls = parseToolCalls(payload, ['terminal']);
assert.equal(calls.length, 1);
assert.equal(calls[0].name, 'terminal');
assert.deepEqual(calls[0].input, { command: 'find "/home" -type d', timeout: 10 });
});
test('parseToolCalls tolerates extra leading less-than before DSML tags', () => {
const payload = [
'<<|DSML|tool_calls>',
' <<|DSML|invoke name="Bash">',
' <<|DSML|parameter name="command"><![CDATA[pwd]]></|DSML|parameter>',
' </|DSML|invoke>',
'</|DSML|tool_calls>',
].join('\n');
const calls = parseToolCalls(payload, ['Bash']);
assert.equal(calls.length, 1);
assert.equal(calls[0].name, 'Bash');
assert.deepEqual(calls[0].input, { command: 'pwd' });
});
test('parseToolCalls tolerates repeated DSML prefix noise', () => {
const payload = [
'<<DSML|DSML|tool_calls>',
' <<DSML|DSML|invoke name="Bash">',
' <<DSML|DSML|parameter name="command"><![CDATA[git status]]></DSML|DSML|parameter>',
' </DSML|DSML|invoke>',
'</DSML|DSML|tool_calls>',
].join('\n');
const calls = parseToolCalls(payload, ['Bash']);
assert.equal(calls.length, 1);
assert.equal(calls[0].name, 'Bash');
assert.deepEqual(calls[0].input, { command: 'git status' });
});
test('parseToolCalls tolerates DSML space-separator typo', () => {
const payload = '<|DSML tool_calls><|DSML invoke name="Read"><|DSML parameter name="file_path"><![CDATA[/tmp/input.txt]]></|DSML parameter></|DSML invoke></|DSML tool_calls>';
const calls = parseToolCalls(payload, ['Read']);
@@ -188,6 +231,28 @@ test('parseToolCalls treats single-item CDATA body as array', () => {
assert.deepEqual(calls[0].input.todos, ['one']);
});
test('parseToolCalls treats loose JSON list as array', () => {
for (const [label, body] of [
['plain text', '{"content":"Test TodoWrite tool","status":"completed"}, {"content":"Another task","status":"pending"}'],
['cdata', '<![CDATA[{"content":"Test TodoWrite tool","status":"completed"}, {"content":"Another task","status":"pending"}]]>'],
]) {
const payload = `<tool_calls><invoke name="TodoWrite"><parameter name="todos">${body}</parameter></invoke></tool_calls>`;
const calls = parseToolCalls(payload, ['TodoWrite']);
assert.equal(calls.length, 1, label);
assert.deepEqual(calls[0].input.todos, [
{ content: 'Test TodoWrite tool', status: 'completed' },
{ content: 'Another task', status: 'pending' },
]);
}
});
test('parseToolCalls keeps preserved text parameters as text', () => {
const payload = '<tool_calls><invoke name="Write"><parameter name="content"><![CDATA[{"content":"Test TodoWrite tool","status":"completed"}, {"content":"Another task","status":"pending"}]]></parameter></invoke></tool_calls>';
const calls = parseToolCalls(payload, ['Write']);
assert.equal(calls.length, 1);
assert.equal(calls[0].input.content, '{"content":"Test TodoWrite tool","status":"completed"}, {"content":"Another task","status":"pending"}');
});
test('formatOpenAIStreamToolCalls normalizes camelCase inputSchema string fields', () => {
const formatted = formatOpenAIStreamToolCalls([
{ name: 'Write', input: { content: { message: 'hi' }, taskId: 1 } },
@@ -285,6 +350,39 @@ test('sieve emits tool_calls for DSML space-separator typo', () => {
assert.equal(text.includes('<|DSML invoke'), false);
});
test('sieve emits tool_calls for DSML trailing pipe tag terminator', () => {
const events = runSieve([
'<|DSML|tool_calls| \n',
'<|DSML|invoke name="terminal">\n',
'<|DSML|parameter name="command"><![CDATA[find "/home" -type d]]></|DSML|parameter>\n',
'<|DSML|parameter name="timeout"><![CDATA[10]]></|DSML|parameter>\n',
'</|DSML|invoke>\n',
'</|DSML|tool_calls>',
], ['terminal']);
const finalCalls = events.filter((evt) => evt.type === 'tool_calls').flatMap((evt) => evt.calls || []);
const text = collectText(events);
assert.equal(finalCalls.length, 1);
assert.equal(finalCalls[0].name, 'terminal');
assert.deepEqual(finalCalls[0].input, { command: 'find "/home" -type d', timeout: 10 });
assert.equal(text.toLowerCase().includes('dsml'), false);
});
test('sieve emits tool_calls for extra leading less-than DSML tags without leaking prefix', () => {
const events = runSieve([
'<<|DSML|tool_calls>\n',
'<<|DSML|invoke name="Bash">\n',
'<<|DSML|parameter name="command"><![CDATA[pwd]]></|DSML|parameter>\n',
'</|DSML|invoke>\n',
'</|DSML|tool_calls>',
], ['Bash']);
const finalCalls = events.filter((evt) => evt.type === 'tool_calls').flatMap((evt) => evt.calls || []);
const text = collectText(events);
assert.equal(finalCalls.length, 1);
assert.equal(finalCalls[0].name, 'Bash');
assert.deepEqual(finalCalls[0].input, { command: 'pwd' });
assert.equal(text.includes('<'), false);
});
test('sieve keeps DSML space lookalike tag names as text', () => {
const input = '<|DSML tool_calls_extra><|DSML invoke name="Read"><|DSML parameter name="file_path">/tmp/input.txt</|DSML parameter></|DSML invoke></|DSML tool_calls_extra>';
const events = runSieve([input], ['Read']);

View File

@@ -1,4 +1,4 @@
{"code":0,"msg":"","data":{"biz_code":0,"biz_msg":"","biz_data":{"id":"file-b10a2aca-39e9-4a38-be9d-9f22e398cb62","status":"PENDING","file_name":"history.txt","from_share":false,"file_size":732,"model_kind":"NORMAL","token_usage":null,"error_code":null,"inserted_at":1777485015.255,"updated_at":1777485015.255,"is_image":false,"audit_result":null}}}
{"code":0,"msg":"","data":{"biz_code":0,"biz_msg":"","biz_data":{"id":"file-b10a2aca-39e9-4a38-be9d-9f22e398cb62","status":"PENDING","file_name":"DS2API_HISTORY.txt","from_share":false,"file_size":732,"model_kind":"NORMAL","token_usage":null,"error_code":null,"inserted_at":1777485015.255,"updated_at":1777485015.255,"is_image":false,"audit_result":null}}}
event: ready
data: {"request_message_id":1,"response_message_id":2,"model_type":"default"}

View File

@@ -1,4 +1,4 @@
{"code":0,"msg":"","data":{"biz_code":0,"biz_msg":"","biz_data":{"id":"file-9c8ae986-75f7-4611-9956-5e1b502f3ec2","status":"SUCCESS","file_name":"history.txt","from_share":false,"file_size":732,"model_kind":"NORMAL","token_usage":145,"error_code":null,"inserted_at":1777485076.42,"updated_at":1777485076.42,"signed_path":"/file?file_id=9c8ae986-75f7-4611-9956-5e1b502f3ec2&state=a1REa2AdO8JmDuxMFiUTPJfpiyY4ie2weyUpYxfvEOrk5lxUCZifpRw9toZAEzn3DAjkgbR6blgZf41KLkHBKwwrcYTIjfxTRKijDqjEfguis03yddpuVrii6keG4%2BXIlcLAsyZG3qcGhfTGVZhsr%2BRl17J%2BcnT9roslhxBcEy4rthFJVMWUI%2BSHjuo2gLEUDfvMfULQ1gSLVGtr%2Fpq%2FcNPCPSxZapIQv04ZVmJLcdbzRkz%2Bb%2BxM5RWUIPujp%2B3ke1WDa3%2B6S4pP0Pv%2BAJ0MFUjQsloUwO4AsJ8YhGBFWg8Ehe1b2yt1N%2Fi%2BIjLRPt5xiNmALcJJXIY%3D","is_image":false,"audit_result":null}}}
{"code":0,"msg":"","data":{"biz_code":0,"biz_msg":"","biz_data":{"id":"file-9c8ae986-75f7-4611-9956-5e1b502f3ec2","status":"SUCCESS","file_name":"DS2API_HISTORY.txt","from_share":false,"file_size":732,"model_kind":"NORMAL","token_usage":145,"error_code":null,"inserted_at":1777485076.42,"updated_at":1777485076.42,"signed_path":"/file?file_id=9c8ae986-75f7-4611-9956-5e1b502f3ec2&state=a1REa2AdO8JmDuxMFiUTPJfpiyY4ie2weyUpYxfvEOrk5lxUCZifpRw9toZAEzn3DAjkgbR6blgZf41KLkHBKwwrcYTIjfxTRKijDqjEfguis03yddpuVrii6keG4%2BXIlcLAsyZG3qcGhfTGVZhsr%2BRl17J%2BcnT9roslhxBcEy4rthFJVMWUI%2BSHjuo2gLEUDfvMfULQ1gSLVGtr%2Fpq%2FcNPCPSxZapIQv04ZVmJLcdbzRkz%2Bb%2BxM5RWUIPujp%2B3ke1WDa3%2B6S4pP0Pv%2BAJ0MFUjQsloUwO4AsJ8YhGBFWg8Ehe1b2yt1N%2Fi%2BIjLRPt5xiNmALcJJXIY%3D","is_image":false,"audit_result":null}}}
event: ready
data: {"request_message_id":1,"response_message_id":2,"model_type":"expert"}

View File

@@ -5,8 +5,8 @@ ROOT_DIR="$(cd "$(dirname "$0")/../.." && pwd)"
TARGETS_FILE="${1:-$ROOT_DIR/plans/node-syntax-gate-targets.txt}"
if [[ ! -f "$TARGETS_FILE" ]]; then
echo "missing targets file: $TARGETS_FILE" >&2
exit 1
echo "checked=0 missing=0 invalid=0"
exit 0
fi
checked=0

View File

@@ -41,8 +41,8 @@ is_test_file() {
}
if [[ ! -f "$TARGETS_FILE" ]]; then
echo "missing targets file: $TARGETS_FILE" >&2
exit 1
echo "checked=0 missing=0 over_limit=0"
exit 0
fi
missing=0

View File

@@ -81,6 +81,22 @@
"source": "/admin/version",
"destination": "/api/index"
},
{
"source": "/admin/chat-history(.*)",
"destination": "/api/index"
},
{
"source": "/admin/proxies(.*)",
"destination": "/api/index"
},
{
"source": "/admin/dev/raw-samples/(.*)",
"destination": "/api/index"
},
{
"source": "/admin/dev/captures(.*)",
"destination": "/api/index"
},
{
"source": "/admin",
"destination": "/admin/index.html"

View File

@@ -217,6 +217,7 @@ export default function ApiTesterContainer({ config, onMessage, authFetch }) {
setSelectedAccount={setSelectedAccount}
effectiveKey={effectiveKey}
selectedAccount={selectedAccount}
model={model}
onMessage={onMessage}
response={response}
isStreaming={isStreaming}

View File

@@ -13,6 +13,7 @@ export default function ChatPanel({
setSelectedAccount,
effectiveKey,
selectedAccount,
model,
onMessage,
response,
isStreaming,
@@ -37,11 +38,15 @@ export default function ChatPanel({
setUploadingFiles(true)
const initialSelectedAccount = String(selectedAccount || '').trim()
const selectedModel = String(model || '').trim()
let boundAccount = initialSelectedAccount
for (const file of files) {
const formData = new FormData()
formData.append('file', file)
formData.append('purpose', 'assistants')
if (selectedModel) {
formData.append('model', selectedModel)
}
const headers = {
'Authorization': `Bearer ${effectiveKey}`,
@@ -181,8 +186,9 @@ export default function ChatPanel({
/>
<div className="absolute left-2 bottom-2 z-10">
<button
type="button"
onClick={() => fileInputRef.current?.click()}
disabled={uploadingFiles || isStreaming || !hasAvailableModel}
disabled={uploadingFiles || isStreaming}
className="p-2 text-muted-foreground hover:text-primary transition-colors disabled:opacity-50 disabled:cursor-not-allowed"
title="Attach files"
>

View File

@@ -16,7 +16,15 @@ const TOOL_MARKER = '<Tool>'
const END_INSTRUCTIONS_MARKER = '<end▁of▁instructions>'
const END_SENTENCE_MARKER = '<end▁of▁sentence>'
const END_TOOL_RESULTS_MARKER = '<end▁of▁toolresults>'
const CURRENT_INPUT_FILE_PROMPT = 'The current request and prior conversation context have already been provided. Answer the latest user request directly.'
const CURRENT_INPUT_FILE_PROMPT = 'Continue from the latest state in the attached DS2API_HISTORY.txt context. Treat it as the current working state and answer the latest user request directly.'
const LEGACY_CURRENT_INPUT_FILE_PROMPTS = new Set([
'The current request and prior conversation context have already been provided. Answer the latest user request directly.',
])
function isCurrentInputFilePrompt(value) {
const text = String(value || '').trim()
return text === CURRENT_INPUT_FILE_PROMPT || LEGACY_CURRENT_INPUT_FILE_PROMPTS.has(text)
}
function formatDateTime(value, lang) {
if (!value) return '-'
@@ -312,7 +320,7 @@ function buildListModeMessages(item, t) {
const placeholderOnly = liveMessages.length === 1
&& String(liveMessages[0]?.role || '').trim().toLowerCase() === 'user'
&& String(liveMessages[0]?.content || '').trim() === CURRENT_INPUT_FILE_PROMPT
&& isCurrentInputFilePrompt(liveMessages[0]?.content)
if (placeholderOnly) {
return { messages: historyMessages, historyMerged: true }

View File

@@ -394,7 +394,7 @@
"thinkingInjectionPromptHelp": "Leave empty to use the built-in default prompt shown as the input placeholder.",
"currentInputFileTitle": "Independent Split",
"currentInputFileEnabled": "Independent split (by size)",
"currentInputFileDesc": "Enabled by default. Once the character threshold is reached, upload the full context as a history.txt context file.",
"currentInputFileDesc": "Enabled by default. Once the character threshold is reached, upload the full context as a DS2API_HISTORY.txt context file.",
"currentInputFileMinChars": "Current input threshold (characters)",
"currentInputFileHelp": "Default is 0, which uses independent split for any non-empty input.",
"compatibilityTitle": "Compatibility",
@@ -485,4 +485,4 @@
"four": "Trigger a redeploy to apply the updated environment variables."
}
}
}
}

View File

@@ -394,7 +394,7 @@
"thinkingInjectionPromptHelp": "留空时使用内置默认提示词;默认内容会显示在输入框占位文本中。",
"currentInputFileTitle": "独立拆分",
"currentInputFileEnabled": "独立拆分(按量)",
"currentInputFileDesc": "默认开启。达到字符阈值后,将完整上下文上传为 history.txt 上下文文件。",
"currentInputFileDesc": "默认开启。达到字符阈值后,将完整上下文上传为 DS2API_HISTORY.txt 上下文文件。",
"currentInputFileMinChars": "当前输入阈值(字符数)",
"currentInputFileHelp": "默认 0表示只要有输入就会使用独立拆分。",
"compatibilityTitle": "兼容性设置",
@@ -485,4 +485,4 @@
"four": "触发重新部署以应用新的环境变量。"
}
}
}
}