Merge pull request #126 from CJackHwang/dev

Merge pull request #125 from CJackHwang/codex/align-documentation-with-configuration-updates

Docs: add `auto_delete.sessions`, rename `claude_model_mapping` to `claude_mapping`, and clarify config token handling

.env.example

@@ -1,93 +1,15 @@
-# DS2API environment template (Go runtime)
-# Copy this file to .env and adjust values.
-# Updated: 2026-02
-
-# ---------------------------------------------------------------
-# Runtime
-# ---------------------------------------------------------------
-# HTTP listen port (default: 5001)
+# DS2API runtime
 PORT=5001
-
-# Log level: DEBUG | INFO | WARN | ERROR
 LOG_LEVEL=INFO
 
-# Max concurrent inflight requests per account in managed-key mode.
-# Default: 2
-# Recommended client concurrency is calculated dynamically as:
-#   account_count * DS2API_ACCOUNT_MAX_INFLIGHT
-# So by default it is account_count * 2.
-# Requests beyond inflight slots enter a waiting queue first.
-# Default queue size equals recommended concurrency, so 429 starts after:
-#   account_count * DS2API_ACCOUNT_MAX_INFLIGHT * 2
-# Alias: DS2API_ACCOUNT_CONCURRENCY
-# DS2API_ACCOUNT_MAX_INFLIGHT=2
+# Admin authentication
+DS2API_ADMIN_KEY=change-me
 
-# Optional waiting queue size override for managed-key mode.
-# Default: recommended_concurrency (same as account_count * inflight_limit)
-# Alias: DS2API_ACCOUNT_QUEUE_SIZE
-# DS2API_ACCOUNT_MAX_QUEUE=10
+# Config loading (choose one)
+# 1) file-based config
+DS2API_CONFIG_PATH=/app/config.json
+# 2) inline JSON or Base64 JSON
+# DS2API_CONFIG_JSON=
 
-# ---------------------------------------------------------------
-# Admin auth
-# ---------------------------------------------------------------
-# Admin key for /admin login and protected admin APIs.
-# Default is "admin" when unset, but setting it explicitly is recommended.
-DS2API_ADMIN_KEY=admin
-
-# Optional JWT signing secret for admin token.
-# Defaults to DS2API_ADMIN_KEY when unset.
-# DS2API_JWT_SECRET=change-me
-
-# Optional admin JWT validity in hours (default: 24)
-# DS2API_JWT_EXPIRE_HOURS=24
-
-# ---------------------------------------------------------------
-# Config source (choose one)
-# ---------------------------------------------------------------
-# Option A: config file path (local/dev recommended)
-# DS2API_CONFIG_PATH=config.json
-
-# Option B: JSON string
-# DS2API_CONFIG_JSON={"keys":["your-api-key"],"accounts":[{"email":"user@example.com","password":"xxx","token":""}]}
-
-# Option C: Base64 encoded JSON (recommended for Vercel env var)
-# DS2API_CONFIG_JSON=eyJrZXlzIjpbInlvdXItYXBpLWtleSJdLCJhY2NvdW50cyI6W3siZW1haWwiOiJ1c2VyQGV4YW1wbGUuY29tIiwicGFzc3dvcmQiOiJ4eHgiLCJ0b2tlbiI6IiJ9XX0=
-#
-# Generate from local config.json:
-#   DS2API_CONFIG_JSON="$(base64 < config.json | tr -d '\n')"
-
-# ---------------------------------------------------------------
-# Paths (optional)
-# ---------------------------------------------------------------
-# WASM file used for PoW solving
-# DS2API_WASM_PATH=sha3_wasm_bg.7b9ca65ddd.wasm
-
-# Built admin static assets directory
-# DS2API_STATIC_ADMIN_DIR=static/admin
-
-# Auto-build WebUI on startup when static/admin is missing.
-# Default: enabled on local/Docker, disabled on Vercel.
-# DS2API_AUTO_BUILD_WEBUI=true
-
-# Internal auth secret used by the Vercel hybrid streaming path
-# (Go prepare endpoint <-> Node stream function).
-# Optional: falls back to DS2API_ADMIN_KEY when unset.
-# DS2API_VERCEL_INTERNAL_SECRET=change-me
-
-# Stream lease TTL seconds for Vercel hybrid streaming.
-# During this window, the managed account stays occupied until Node calls release.
-# Default: 900 (15 minutes)
-# DS2API_VERCEL_STREAM_LEASE_TTL_SECONDS=900
-
-# ---------------------------------------------------------------
-# Vercel sync integration (optional)
-# ---------------------------------------------------------------
-# VERCEL_TOKEN=your-vercel-token
-# VERCEL_PROJECT_ID=prj_xxxxxxxxxxxx
-# VERCEL_TEAM_ID=team_xxxxxxxxxxxx
-
-# Optional: Vercel deployment protection bypass secret.
-# If deployment protection is enabled, DS2API will use this value as
-# x-vercel-protection-bypass for internal Node->Go calls on Vercel.
-# You can also use VERCEL_AUTOMATION_BYPASS_SECRET directly.
-# DS2API_VERCEL_PROTECTION_BYPASS=your-bypass-secret
+# Optional: static admin assets path
+# DS2API_STATIC_ADMIN_DIR=/app/static/admin

.github/workflows/release-artifacts.yml

@@ -51,6 +51,10 @@ jobs:
         run: |
           set -euo pipefail
           TAG="${RELEASE_TAG}"
+          BUILD_VERSION="${TAG}"
+          if [ -z "${BUILD_VERSION}" ] && [ -f VERSION ]; then
+            BUILD_VERSION="$(cat VERSION | tr -d '[:space:]')"
+          fi
           mkdir -p dist
 
           targets=(
@@ -73,7 +77,7 @@ jobs:
 
             mkdir -p "${STAGE}/static"
             CGO_ENABLED=0 GOOS="${GOOS}" GOARCH="${GOARCH}" \
-              go build -trimpath -ldflags="-s -w" -o "${STAGE}/${BIN}" ./cmd/ds2api
+              go build -trimpath -ldflags="-s -w -X ds2api/internal/version.BuildVersion=${BUILD_VERSION}" -o "${STAGE}/${BIN}" ./cmd/ds2api
 
             cp config.example.json .env.example sha3_wasm_bg.7b9ca65ddd.wasm LICENSE README.MD README.en.md "${STAGE}/"
             cp -R static/admin "${STAGE}/static/admin"

.github/workflows/release-dockerhub.yml

@@ -123,5 +123,7 @@ jobs:
           labels: |
             org.opencontainers.image.version=${{ steps.next_version.outputs.new_version }}
             org.opencontainers.image.revision=${{ github.sha }}
+          build-args: |
+            BUILD_VERSION=${{ steps.next_version.outputs.new_tag }}
           cache-from: type=gha
           cache-to: type=gha,mode=max

.github/workflows/release.yml

@@ -1,128 +1,130 @@
-name: Release to Aliyun CR
-
-on:
-  workflow_dispatch:
-    inputs:
-      version_type:
-        description: '版本类型'
-        required: true
-        default: 'patch'
-        type: choice
-        options:
-          - patch
-          - minor
-          - major
-
-permissions:
-  contents: write
-
-jobs:
-  release:
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@v5
-        with:
-          fetch-depth: 0
-          token: ${{ secrets.GITHUB_TOKEN }}
-
-      - name: Get current version
-        id: get_version
-        run: |
-          LATEST_TAG=$(git describe --tags --abbrev=0 2>/dev/null || echo "v0.0.0")
-          TAG_VERSION=${LATEST_TAG#v}
-
-          if [ -f VERSION ]; then
-            FILE_VERSION=$(cat VERSION | tr -d '[:space:]')
-          else
-            FILE_VERSION="0.0.0"
-          fi
-
-          function version_gt() { test "$(printf '%s\n' "$@" | sort -V | head -n 1)" != "$1"; }
-
-          if version_gt "$FILE_VERSION" "$TAG_VERSION"; then
-            VERSION="$FILE_VERSION"
-          else
-            VERSION="$TAG_VERSION"
-          fi
-
-          echo "Current version: $VERSION"
-          echo "current_version=$VERSION" >> $GITHUB_OUTPUT
-
-      - name: Calculate next version
-        id: next_version
-        env:
-          VERSION_TYPE: ${{ github.event.inputs.version_type }}
-        run: |
-          VERSION="${{ steps.get_version.outputs.current_version }}"
-          BASE_VERSION=$(echo "$VERSION" | sed 's/-.*$//')
-
-          IFS='.' read -r -a version_parts <<< "$BASE_VERSION"
-          MAJOR="${version_parts[0]:-0}"
-          MINOR="${version_parts[1]:-0}"
-          PATCH="${version_parts[2]:-0}"
-
-          case "$VERSION_TYPE" in
-            major)
-              NEW_VERSION="$((MAJOR + 1)).0.0"
-              ;;
-            minor)
-              NEW_VERSION="${MAJOR}.$((MINOR + 1)).0"
-              ;;
-            *)
-              NEW_VERSION="${MAJOR}.${MINOR}.$((PATCH + 1))"
-              ;;
-          esac
-
-          echo "New version: $NEW_VERSION"
-          echo "new_version=$NEW_VERSION" >> $GITHUB_OUTPUT
-          echo "new_tag=v$NEW_VERSION" >> $GITHUB_OUTPUT
-
-      - name: Update VERSION file
-        run: |
-          echo "${{ steps.next_version.outputs.new_version }}" > VERSION
-
-      - name: Commit VERSION and create tag
-        run: |
-          git config user.name "github-actions[bot]"
-          git config user.email "github-actions[bot]@users.noreply.github.com"
-
-          git add VERSION
-          if ! git diff --cached --quiet; then
-            git commit -m "chore: bump version to ${{ steps.next_version.outputs.new_tag }} [skip ci]"
-          fi
-
-          NEW_TAG="${{ steps.next_version.outputs.new_tag }}"
-          git tag -a "$NEW_TAG" -m "Release $NEW_TAG"
-          git push origin HEAD:main "$NEW_TAG"
-
-      # Docker 构建并推送到阿里云
-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@v3
-
-      - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
-
-      - name: Log in to Aliyun Container Registry
-        uses: docker/login-action@v3
-        with:
-          registry: ${{ secrets.ALIYUN_REGISTRY }}
-          username: ${{ secrets.ALIYUN_REGISTRY_USER }}
-          password: ${{ secrets.ALIYUN_REGISTRY_PASSWORD }}
-
-      - name: Build and push Docker image
-        uses: docker/build-push-action@v6
-        with:
-          context: .
-          file: ./Dockerfile
-          platforms: linux/amd64,linux/arm64
-          push: true
-          tags: |
-            ${{ secrets.ALIYUN_REGISTRY }}/${{ secrets.ALIYUN_REGISTRY_NAMESPACE }}/ds2api:${{ steps.next_version.outputs.new_tag }}
-            ${{ secrets.ALIYUN_REGISTRY }}/${{ secrets.ALIYUN_REGISTRY_NAMESPACE }}/ds2api:${{ steps.next_version.outputs.new_version }}
-            ${{ secrets.ALIYUN_REGISTRY }}/${{ secrets.ALIYUN_REGISTRY_NAMESPACE }}/ds2api:latest
-          labels: |
-            org.opencontainers.image.version=${{ steps.next_version.outputs.new_version }}
-            org.opencontainers.image.revision=${{ github.sha }}
-          cache-from: type=gha
-          cache-to: type=gha,mode=max
+name: Release to Aliyun CR
+
+on:
+  workflow_dispatch:
+    inputs:
+      version_type:
+        description: '版本类型'
+        required: true
+        default: 'patch'
+        type: choice
+        options:
+          - patch
+          - minor
+          - major
+
+permissions:
+  contents: write
+
+jobs:
+  release:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v5
+        with:
+          fetch-depth: 0
+          token: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Get current version
+        id: get_version
+        run: |
+          LATEST_TAG=$(git describe --tags --abbrev=0 2>/dev/null || echo "v0.0.0")
+          TAG_VERSION=${LATEST_TAG#v}
+
+          if [ -f VERSION ]; then
+            FILE_VERSION=$(cat VERSION | tr -d '[:space:]')
+          else
+            FILE_VERSION="0.0.0"
+          fi
+
+          function version_gt() { test "$(printf '%s\n' "$@" | sort -V | head -n 1)" != "$1"; }
+
+          if version_gt "$FILE_VERSION" "$TAG_VERSION"; then
+            VERSION="$FILE_VERSION"
+          else
+            VERSION="$TAG_VERSION"
+          fi
+
+          echo "Current version: $VERSION"
+          echo "current_version=$VERSION" >> $GITHUB_OUTPUT
+
+      - name: Calculate next version
+        id: next_version
+        env:
+          VERSION_TYPE: ${{ github.event.inputs.version_type }}
+        run: |
+          VERSION="${{ steps.get_version.outputs.current_version }}"
+          BASE_VERSION=$(echo "$VERSION" | sed 's/-.*$//')
+
+          IFS='.' read -r -a version_parts <<< "$BASE_VERSION"
+          MAJOR="${version_parts[0]:-0}"
+          MINOR="${version_parts[1]:-0}"
+          PATCH="${version_parts[2]:-0}"
+
+          case "$VERSION_TYPE" in
+            major)
+              NEW_VERSION="$((MAJOR + 1)).0.0"
+              ;;
+            minor)
+              NEW_VERSION="${MAJOR}.$((MINOR + 1)).0"
+              ;;
+            *)
+              NEW_VERSION="${MAJOR}.${MINOR}.$((PATCH + 1))"
+              ;;
+          esac
+
+          echo "New version: $NEW_VERSION"
+          echo "new_version=$NEW_VERSION" >> $GITHUB_OUTPUT
+          echo "new_tag=v$NEW_VERSION" >> $GITHUB_OUTPUT
+
+      - name: Update VERSION file
+        run: |
+          echo "${{ steps.next_version.outputs.new_version }}" > VERSION
+
+      - name: Commit VERSION and create tag
+        run: |
+          git config user.name "github-actions[bot]"
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+
+          git add VERSION
+          if ! git diff --cached --quiet; then
+            git commit -m "chore: bump version to ${{ steps.next_version.outputs.new_tag }} [skip ci]"
+          fi
+
+          NEW_TAG="${{ steps.next_version.outputs.new_tag }}"
+          git tag -a "$NEW_TAG" -m "Release $NEW_TAG"
+          git push origin HEAD:main "$NEW_TAG"
+
+      # Docker 构建并推送到阿里云
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@v3
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Log in to Aliyun Container Registry
+        uses: docker/login-action@v3
+        with:
+          registry: ${{ secrets.ALIYUN_REGISTRY }}
+          username: ${{ secrets.ALIYUN_REGISTRY_USER }}
+          password: ${{ secrets.ALIYUN_REGISTRY_PASSWORD }}
+
+      - name: Build and push Docker image
+        uses: docker/build-push-action@v6
+        with:
+          context: .
+          file: ./Dockerfile
+          platforms: linux/amd64,linux/arm64
+          push: true
+          tags: |
+            ${{ secrets.ALIYUN_REGISTRY }}/${{ secrets.ALIYUN_REGISTRY_NAMESPACE }}/ds2api:${{ steps.next_version.outputs.new_tag }}
+            ${{ secrets.ALIYUN_REGISTRY }}/${{ secrets.ALIYUN_REGISTRY_NAMESPACE }}/ds2api:${{ steps.next_version.outputs.new_version }}
+            ${{ secrets.ALIYUN_REGISTRY }}/${{ secrets.ALIYUN_REGISTRY_NAMESPACE }}/ds2api:latest
+          labels: |
+            org.opencontainers.image.version=${{ steps.next_version.outputs.new_version }}
+            org.opencontainers.image.revision=${{ github.sha }}
+          build-args: |
+            BUILD_VERSION=${{ steps.next_version.outputs.new_tag }}
+          cache-from: type=gha
+          cache-to: type=gha,mode=max

API.en.md

@@ -623,6 +623,7 @@ Reads runtime settings and status, including:
 - `admin` (JWT expiry, default-password warning, etc.)
 - `runtime` (`account_max_inflight`, `account_max_queue`, `global_max_inflight`)
 - `toolcall` / `responses` / `embeddings`
+- `auto_delete` (`sessions`)
 - `claude_mapping` / `model_aliases`
 - `env_backed`, `needs_vercel_sync`
 
@@ -635,6 +636,7 @@ Hot-updates runtime settings. Supported fields:
 - `toolcall.mode` / `toolcall.early_emit_confidence`
 - `responses.store_ttl_seconds`
 - `embeddings.provider`
+- `auto_delete.sessions`
 - `claude_mapping`
 - `model_aliases`
 

API.md

@@ -628,6 +628,7 @@ data: {"type":"message_stop"}
 - `admin`（JWT 过期、默认密码告警等）
 - `runtime`（`account_max_inflight`、`account_max_queue`、`global_max_inflight`）
 - `toolcall` / `responses` / `embeddings`
+- `auto_delete`（`sessions`）
 - `claude_mapping` / `model_aliases`
 - `env_backed`、`needs_vercel_sync`
 
@@ -640,6 +641,7 @@ data: {"type":"message_stop"}
 - `toolcall.mode` / `toolcall.early_emit_confidence`
 - `responses.store_ttl_seconds`
 - `embeddings.provider`
+- `auto_delete.sessions`
 - `claude_mapping`
 - `model_aliases`
 

CONTRIBUTING.en.md

@@ -99,7 +99,7 @@ ds2api/
 ├── api/
 │   ├── index.go             # Vercel Serverless Go entry
 │   ├── chat-stream.js       # Vercel Node.js stream relay
-│   └── helpers/             # Node.js helper modules
+│   └── (rewrite targets in vercel.json)
 ├── internal/
 │   ├── account/             # Account pool and concurrency queue
 │   ├── adapter/
@@ -112,6 +112,7 @@ ds2api/
 │   ├── compat/              # Compatibility helpers
 │   ├── config/              # Config loading and hot-reload
 │   ├── deepseek/            # DeepSeek client, PoW WASM
+│   ├── js/                  # Node runtime stream/compat logic
 │   ├── devcapture/          # Dev packet capture
 │   ├── format/              # Output formatting
 │   ├── prompt/              # Prompt building
@@ -123,7 +124,9 @@ ds2api/
 │   └── webui/               # WebUI static hosting
 ├── webui/                   # React WebUI source
 │   └── src/
-│       ├── components/      # Components
+│       ├── app/             # Routing, auth, config state
+│       ├── features/        # Feature modules
+│       ├── components/      # Shared components
 │       └── locales/         # Language packs
 ├── scripts/                 # Build and test scripts
 ├── static/admin/            # WebUI build output (not committed)

CONTRIBUTING.md

@@ -99,7 +99,7 @@ ds2api/
 ├── api/
 │   ├── index.go             # Vercel Serverless Go 入口
 │   ├── chat-stream.js       # Vercel Node.js 流式转发
-│   └── helpers/             # Node.js 辅助模块
+│   └── (rewrite targets in vercel.json)
 ├── internal/
 │   ├── account/             # 账号池与并发队列
 │   ├── adapter/
@@ -112,6 +112,7 @@ ds2api/
 │   ├── compat/              # 兼容性辅助
 │   ├── config/              # 配置加载与热更新
 │   ├── deepseek/            # DeepSeek 客户端、PoW WASM
+│   ├── js/                  # Node 运行时流式/兼容逻辑
 │   ├── devcapture/          # 开发抓包
 │   ├── format/              # 输出格式化
 │   ├── prompt/              # Prompt 构建
@@ -123,7 +124,9 @@ ds2api/
 │   └── webui/               # WebUI 静态托管
 ├── webui/                   # React WebUI 源码
 │   └── src/
-│       ├── components/      # 组件
+│       ├── app/             # 路由、鉴权、配置状态
+│       ├── features/        # 业务功能模块
+│       ├── components/      # 通用组件
 │       └── locales/         # 语言包
 ├── scripts/                 # 构建与测试脚本
 ├── static/admin/            # WebUI 构建产物（不提交）

DEPLOY.en.md

@@ -113,12 +113,8 @@ go build -o ds2api ./cmd/ds2api
 # Copy env template
 cp .env.example .env
 
-# Generate single-line Base64 from config.json
-DS2API_CONFIG_JSON="$(base64 < config.json | tr -d '\n')"
-
-# Edit .env and set:
+# Edit .env and set at least:
 #   DS2API_ADMIN_KEY=your-admin-key
-#   DS2API_CONFIG_JSON=${DS2API_CONFIG_JSON}
 
 # Start
 docker-compose up -d
@@ -185,6 +181,7 @@ Notes:
 
 - **Port**: DS2API listens on `5001` by default; the template sets `PORT=5001`.
 - **Persistent config**: the template mounts `/data` and sets `DS2API_CONFIG_PATH=/data/config.json`. After importing config in Admin UI, it will be written and persisted to this path.
+- **Build version**: Zeabur / regular `docker build` does not require `BUILD_VERSION` by default. The image prefers that build arg when provided, and automatically falls back to the repo-root `VERSION` file when it is absent.
 - **First login**: after deployment, open `/admin` and login with `DS2API_ADMIN_KEY` shown in Zeabur env/template instructions (recommended: rotate to a strong secret after first login).
 
 ---
@@ -366,7 +363,7 @@ Each archive includes:
 
 - `ds2api` executable (`ds2api.exe` on Windows)
 - `static/admin/` (built WebUI assets)
-- `sha3_wasm_bg.7b9ca65ddd.wasm`
+- `sha3_wasm_bg.7b9ca65ddd.wasm` (optional; binary has embedded fallback)
 - `config.example.json`, `.env.example`
 - `README.MD`, `README.en.md`, `LICENSE`
 
@@ -455,7 +452,9 @@ server {
 ```bash
 # Copy compiled binary and related files to target directory
 sudo mkdir -p /opt/ds2api
-sudo cp ds2api config.json sha3_wasm_bg.7b9ca65ddd.wasm /opt/ds2api/
+sudo cp ds2api config.json /opt/ds2api/
+# Optional: if you want to use an external WASM file (override embedded one)
+# sudo cp sha3_wasm_bg.7b9ca65ddd.wasm /opt/ds2api/
 sudo cp -r static/admin /opt/ds2api/static/admin
 ```
 

DEPLOY.md

@@ -113,12 +113,8 @@ go build -o ds2api ./cmd/ds2api
 # 复制环境变量模板
 cp .env.example .env
 
-# 从 config.json 生成单行 Base64
-DS2API_CONFIG_JSON="$(base64 < config.json | tr -d '\n')"
-
-# 编辑 .env（请改成你的强密码），设置：
+# 编辑 .env（请改成你的强密码），至少设置：
 #   DS2API_ADMIN_KEY=your-admin-key
-#   DS2API_CONFIG_JSON=${DS2API_CONFIG_JSON}
 
 # 启动
 docker-compose up -d
@@ -185,6 +181,7 @@ healthcheck:
 
 - **端口**：服务默认监听 `5001`，模板会固定设置 `PORT=5001`。
 - **配置持久化**：模板挂载卷 `/data`，并设置 `DS2API_CONFIG_PATH=/data/config.json`；在管理台导入配置后，会写入并持久化到该路径。
+- **构建版本号**：Zeabur / 普通 `docker build` 默认不需要传 `BUILD_VERSION`；镜像会优先使用该构建参数，未提供时自动回退到仓库根目录的 `VERSION` 文件。
 - **首次登录**：部署完成后访问 `/admin`，使用 Zeabur 环境变量/模板指引中的 `DS2API_ADMIN_KEY` 登录（建议首次登录后自行更换为强密码）。
 
 ---
@@ -366,7 +363,7 @@ No Output Directory named "public" found after the Build completed.
 
 - `ds2api` 可执行文件（Windows 为 `ds2api.exe`）
 - `static/admin/`（WebUI 构建产物）
-- `sha3_wasm_bg.7b9ca65ddd.wasm`
+- `sha3_wasm_bg.7b9ca65ddd.wasm`（可选；程序内置 embed fallback）
 - `config.example.json`、`.env.example`
 - `README.MD`、`README.en.md`、`LICENSE`
 
@@ -455,7 +452,9 @@ server {
 ```bash
 # 将编译好的二进制文件和相关文件复制到目标目录
 sudo mkdir -p /opt/ds2api
-sudo cp ds2api config.json sha3_wasm_bg.7b9ca65ddd.wasm /opt/ds2api/
+sudo cp ds2api config.json /opt/ds2api/
+# 可选：若你希望使用外置 WASM 文件（覆盖内置版本）
+# sudo cp sha3_wasm_bg.7b9ca65ddd.wasm /opt/ds2api/
 sudo cp -r static/admin /opt/ds2api/static/admin
 ```
 

Dockerfile

@@ -10,19 +10,24 @@ FROM golang:1.24 AS go-builder
 WORKDIR /app
 ARG TARGETOS
 ARG TARGETARCH
+ARG BUILD_VERSION
 COPY go.mod go.sum* ./
 RUN go mod download
 COPY . .
 RUN set -eux; \
     GOOS="${TARGETOS:-$(go env GOOS)}"; \
     GOARCH="${TARGETARCH:-$(go env GOARCH)}"; \
-    CGO_ENABLED=0 GOOS="${GOOS}" GOARCH="${GOARCH}" go build -o /out/ds2api ./cmd/ds2api
+    BUILD_VERSION_RESOLVED="${BUILD_VERSION:-}"; \
+    if [ -z "${BUILD_VERSION_RESOLVED}" ] && [ -f VERSION ]; then BUILD_VERSION_RESOLVED="$(cat VERSION | tr -d "[:space:]")"; fi; \
+    CGO_ENABLED=0 GOOS="${GOOS}" GOARCH="${GOARCH}" go build -ldflags="-s -w -X ds2api/internal/version.BuildVersion=${BUILD_VERSION_RESOLVED}" -o /out/ds2api ./cmd/ds2api
 
 FROM busybox:1.36.1-musl AS busybox-tools
 
 FROM debian:bookworm-slim AS runtime-base
 WORKDIR /app
-COPY --from=go-builder /etc/ssl/certs/ca-certificates.crt /etc/ssl/certs/ca-certificates.crt
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends ca-certificates \
+    && rm -rf /var/lib/apt/lists/*
 COPY --from=busybox-tools /bin/busybox /usr/local/bin/busybox
 EXPOSE 5001
 CMD ["/usr/local/bin/ds2api"]

README.MD

@@ -106,6 +106,14 @@ flowchart LR
 可通过配置中的 `claude_mapping` 或 `claude_model_mapping` 覆盖映射关系。
 另外，`/anthropic/v1/models` 现已包含 Claude 1.x/2.x/3.x/4.x 历史模型 ID 与常见别名，便于旧客户端直接兼容。
 
+
+#### Claude Code 接入避坑（实测）
+
+- `ANTHROPIC_BASE_URL` 推荐直接指向 DS2API 根地址（例如 `http://127.0.0.1:5001`），Claude Code 会请求 `/v1/messages?beta=true`。
+- `ANTHROPIC_API_KEY` 需要与 `config.json` 中 `keys` 一致；建议同时保留常规 key 与 `sk-ant-*` 形态 key，兼容不同客户端校验习惯。
+- 若系统设置了代理，建议对 DS2API 地址配置 `NO_PROXY=127.0.0.1,localhost,<你的主机IP>`，避免本地回环请求被代理拦截。
+- 如遇“工具调用输出成文本、未执行”问题，请升级到包含 Claude 工具调用多格式解析（JSON/XML/ANTML/invoke）的版本。
+
 ### Gemini 接口
 
 Gemini 适配器将模型名通过 `model_aliases` 或内置规则映射到 DeepSeek 原生模型，支持 `generateContent` 和 `streamGenerateContent` 两种调用方式，并完整支持 Tool Calling（`functionDeclarations` → `functionCall` 输出）。
@@ -152,17 +160,13 @@ go run ./cmd/ds2api
 # 1. 准备环境变量文件
 cp .env.example .env
 
-# 2. 从 config.json 生成 DS2API_CONFIG_JSON（单行 Base64）
-DS2API_CONFIG_JSON="$(base64 < config.json | tr -d '\n')"
-
-# 3. 编辑 .env，设置：
+# 2. 编辑 .env（至少设置 DS2API_ADMIN_KEY）
 #    DS2API_ADMIN_KEY=请替换为强密码
-#    DS2API_CONFIG_JSON=${DS2API_CONFIG_JSON}
 
-# 4. 启动
+# 3. 启动
 docker-compose up -d
 
-# 5. 查看日志
+# 4. 查看日志
 docker-compose logs -f
 ```
 
@@ -174,6 +178,8 @@ docker-compose logs -f
 2. 部署完成后访问 `/admin`，使用 Zeabur 环境变量/模板指引中的 `DS2API_ADMIN_KEY` 登录。
 3. 在管理台导入/编辑配置（会写入并持久化到 `/data/config.json`）。
 
+说明：Zeabur 使用仓库内 `Dockerfile` 直接构建时，不需要额外传入 `BUILD_VERSION`；镜像会优先读取该构建参数，未提供时自动回退到仓库根目录的 `VERSION` 文件。
+
 ### 方式三：Vercel 部署
 
 1. Fork 仓库到自己的 GitHub
@@ -238,13 +244,11 @@ cp opencode.json.example opencode.json
   "accounts": [
     {
       "email": "user@example.com",
-      "password": "your-password",
-      "token": ""
+      "password": "your-password"
     },
     {
       "mobile": "12345678901",
-      "password": "your-password",
-      "token": ""
+      "password": "your-password"
     }
   ],
   "model_aliases": {
@@ -265,7 +269,7 @@ cp opencode.json.example opencode.json
   "embeddings": {
     "provider": "deterministic"
   },
-  "claude_model_mapping": {
+  "claude_mapping": {
     "fast": "deepseek-chat",
     "slow": "deepseek-reasoner"
   },
@@ -276,21 +280,25 @@ cp opencode.json.example opencode.json
     "account_max_inflight": 2,
     "account_max_queue": 0,
     "global_max_inflight": 0
+  },
+  "auto_delete": {
+    "sessions": false
   }
 }
 ```
 
 - `keys`：API 访问密钥列表，客户端通过 `Authorization: Bearer <key>` 鉴权
 - `accounts`：DeepSeek 账号列表，支持 `email` 或 `mobile` 登录
-- `token`：留空则首次请求时自动登录获取；也可预填已有 token
+- `token`：配置文件中即使填写也会在加载时被清空（不会从 `config.json` 读取 token）；实际 token 仅在运行时内存中维护并自动刷新
 - `model_aliases`：常见模型名（如 GPT/Codex/Claude）到 DeepSeek 模型的映射
 - `compat.wide_input_strict_output`：建议保持 `true`（当前实现默认宽进严出）
 - `toolcall`：固定采用特征匹配 + 高置信早发策略
 - `responses.store_ttl_seconds`：`/v1/responses/{id}` 的内存缓存 TTL
 - `embeddings.provider`：embedding 提供方（当前内置 `deterministic/mock/builtin`）
-- `claude_model_mapping`：字典中 `fast`/`slow` 后缀映射到对应 DeepSeek 模型
+- `claude_mapping`：字典中 `fast`/`slow` 后缀映射到对应 DeepSeek 模型（兼容读取 `claude_model_mapping`）
 - `admin`：管理后台设置（JWT 过期时间、密码哈希等），可通过 Admin Settings API 热更新
-- `runtime`：运行时参数（并发限制、队列大小），可通过 Admin Settings API 热更新
+- `runtime`：运行时参数（并发限制、队列大小），可通过 Admin Settings API 热更新；`account_max_queue=0`/`global_max_inflight=0` 表示按推荐值自动计算
+- `auto_delete.sessions`：是否在请求结束后自动清理 DeepSeek 会话（默认 `false`，可在 Settings 热更新）
 
 ### 环境变量
 
@@ -389,7 +397,7 @@ ds2api/
 ├── api/
 │   ├── index.go             # Vercel Serverless Go 入口
 │   ├── chat-stream.js       # Vercel Node.js 流式转发
-│   └── helpers/             # Node.js 辅助模块
+│   └── (rewrite targets in vercel.json)
 ├── internal/
 │   ├── account/             # 账号池与并发队列
 │   ├── adapter/
@@ -402,6 +410,7 @@ ds2api/
 │   ├── compat/              # 兼容性辅助
 │   ├── config/              # 配置加载与热更新
 │   ├── deepseek/            # DeepSeek API 客户端、PoW WASM
+│   ├── js/                  # Node 运行时流式处理与兼容逻辑
 │   ├── devcapture/          # 开发抓包模块
 │   ├── format/              # 输出格式化
 │   ├── prompt/              # Prompt 构建
@@ -412,7 +421,9 @@ ds2api/
 │   └── webui/               # WebUI 静态文件托管与自动构建
 ├── webui/                   # React WebUI 源码（Vite + Tailwind）
 │   └── src/
-│       ├── components/      # AccountManager / ApiTester / BatchImport / VercelSync / Login / LandingPage
+│       ├── app/             # 路由、鉴权、配置状态管理
+│       ├── features/        # 业务功能模块（account/settings/vercel/apiTester）
+│       ├── components/      # 登录/落地页等通用组件
 │       └── locales/         # 中英文语言包（zh.json / en.json）
 ├── scripts/
 │   └── build-webui.sh       # WebUI 手动构建脚本
@@ -468,6 +479,23 @@ go run ./cmd/ds2api-tests \
 npm ci --prefix webui && npm run build --prefix webui
 ```
 
+## 测试
+
+详细测试指南请参阅 [TESTING.md](TESTING.md)。
+
+### 快速测试命令
+
+```bash
+# 运行所有单元测试
+go test ./...
+
+# 运行 tool calls 相关测试（调试工具调用问题）
+go test -v -run 'TestParseToolCalls|TestRepair' ./internal/util/
+
+# 运行端到端测试
+./tests/scripts/run-live.sh
+```
+
 ## Release 自动构建（GitHub Actions）
 
 工作流文件：`.github/workflows/release-artifacts.yml`
@@ -475,7 +503,7 @@ npm ci --prefix webui && npm run build --prefix webui
 - **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`、`darwin/amd64`、`darwin/arm64`、`windows/amd64`）+ `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、配置示例、README、LICENSE
+- **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）、配置示例、README、LICENSE
 
 ## 免责声明
 

README.en.md

@@ -106,6 +106,14 @@ flowchart LR
 Override mapping via `claude_mapping` or `claude_model_mapping` in config.
 In addition, `/anthropic/v1/models` now includes historical Claude 1.x/2.x/3.x/4.x IDs and common aliases for legacy client compatibility.
 
+
+#### Claude Code integration pitfalls (validated)
+
+- Set `ANTHROPIC_BASE_URL` to the DS2API root URL (for example `http://127.0.0.1:5001`). Claude Code sends requests to `/v1/messages?beta=true`.
+- `ANTHROPIC_API_KEY` must match an entry in `keys` from `config.json`. Keeping both a regular key and an `sk-ant-*` style key improves client compatibility.
+- If your environment has proxy variables, set `NO_PROXY=127.0.0.1,localhost,<your_host_ip>` for DS2API to avoid proxy interception of local traffic.
+- If tool calls are rendered as plain text and not executed, upgrade to a build that includes multi-format Claude tool-call parsing (JSON/XML/ANTML/invoke).
+
 ### Gemini Endpoint
 
 The Gemini adapter maps model names to DeepSeek native models via `model_aliases` or built-in heuristics, supporting both `generateContent` and `streamGenerateContent` call patterns with full Tool Calling support (`functionDeclarations` → `functionCall` output).
@@ -152,17 +160,13 @@ Default URL: `http://localhost:5001`
 # 1. Prepare env file
 cp .env.example .env
 
-# 2. Generate DS2API_CONFIG_JSON from config.json (single-line Base64)
-DS2API_CONFIG_JSON="$(base64 < config.json | tr -d '\n')"
-
-# 3. Edit .env and set:
+# 2. Edit .env (at least set DS2API_ADMIN_KEY)
 #    DS2API_ADMIN_KEY=replace-with-a-strong-secret
-#    DS2API_CONFIG_JSON=${DS2API_CONFIG_JSON}
 
-# 4. Start
+# 3. Start
 docker-compose up -d
 
-# 5. View logs
+# 4. View logs
 docker-compose logs -f
 ```
 
@@ -174,6 +178,8 @@ Rebuild after updates: `docker-compose up -d --build`
 2. After deployment, open `/admin` and login with `DS2API_ADMIN_KEY` shown in Zeabur env/template instructions.
 3. Import / edit config in Admin UI (it will be written and persisted to `/data/config.json`).
 
+Note: when Zeabur builds directly from the repo `Dockerfile`, you do not need to pass `BUILD_VERSION`. The image prefers that build arg when provided, and automatically falls back to the repo-root `VERSION` file when it is absent.
+
 ### Option 3: Vercel
 
 1. Fork this repo to your GitHub account
@@ -238,13 +244,11 @@ cp opencode.json.example opencode.json
   "accounts": [
     {
       "email": "user@example.com",
-      "password": "your-password",
-      "token": ""
+      "password": "your-password"
     },
     {
       "mobile": "12345678901",
-      "password": "your-password",
-      "token": ""
+      "password": "your-password"
     }
   ],
   "model_aliases": {
@@ -265,7 +269,7 @@ cp opencode.json.example opencode.json
   "embeddings": {
     "provider": "deterministic"
   },
-  "claude_model_mapping": {
+  "claude_mapping": {
     "fast": "deepseek-chat",
     "slow": "deepseek-reasoner"
   },
@@ -276,21 +280,25 @@ cp opencode.json.example opencode.json
     "account_max_inflight": 2,
     "account_max_queue": 0,
     "global_max_inflight": 0
+  },
+  "auto_delete": {
+    "sessions": false
   }
 }
 ```
 
 - `keys`: API access keys; clients authenticate via `Authorization: Bearer <key>`
 - `accounts`: DeepSeek account list, supports `email` or `mobile` login
-- `token`: Leave empty for auto-login on first request; or pre-fill an existing token
+- `token`: Even if set in `config.json`, it is cleared during load (DS2API does not read persisted tokens from config); runtime tokens are maintained/refreshed in memory only
 - `model_aliases`: Map common model names (GPT/Codex/Claude) to DeepSeek models
 - `compat.wide_input_strict_output`: Keep `true` (current default policy)
 - `toolcall`: Fixed to feature matching + high-confidence early emit
 - `responses.store_ttl_seconds`: In-memory TTL for `/v1/responses/{id}`
 - `embeddings.provider`: Embeddings provider (`deterministic/mock/builtin` built-in)
-- `claude_model_mapping`: Maps `fast`/`slow` suffixes to corresponding DeepSeek models
+- `claude_mapping`: Maps `fast`/`slow` suffixes to corresponding DeepSeek models (still compatible with `claude_model_mapping`)
 - `admin`: Admin panel settings (JWT expiry, password hash, etc.), hot-reloadable via Admin Settings API
-- `runtime`: Runtime parameters (concurrency limits, queue sizes), hot-reloadable via Admin Settings API
+- `runtime`: Runtime parameters (concurrency limits, queue sizes), hot-reloadable via Admin Settings API; `account_max_queue=0`/`global_max_inflight=0` means auto-calculate from recommended values
+- `auto_delete.sessions`: Whether to auto-delete DeepSeek sessions after request completion (default `false`, hot-reloadable via Settings)
 
 ### Environment Variables
 
@@ -390,7 +398,7 @@ ds2api/
 ├── api/
 │   ├── index.go             # Vercel Serverless Go entry
 │   ├── chat-stream.js       # Vercel Node.js stream relay
-│   └── helpers/             # Node.js helper modules
+│   └── (rewrite targets in vercel.json)
 ├── internal/
 │   ├── account/             # Account pool and concurrency queue
 │   ├── adapter/
@@ -403,6 +411,7 @@ ds2api/
 │   ├── compat/              # Compatibility helpers
 │   ├── config/              # Config loading and hot-reload
 │   ├── deepseek/            # DeepSeek API client, PoW WASM
+│   ├── js/                  # Node runtime stream/compat logic
 │   ├── devcapture/          # Dev packet capture module
 │   ├── format/              # Output formatting
 │   ├── prompt/              # Prompt construction
@@ -413,7 +422,9 @@ ds2api/
 │   └── webui/               # WebUI static file serving and auto-build
 ├── webui/                   # React WebUI source (Vite + Tailwind)
 │   └── src/
-│       ├── components/      # AccountManager / ApiTester / BatchImport / VercelSync / Login / LandingPage
+│       ├── app/             # Routing, auth, config state
+│       ├── features/        # Feature modules (account/settings/vercel/apiTester)
+│       ├── components/      # Shared UI pieces (login/landing, etc.)
 │       └── locales/         # Language packs (zh.json / en.json)
 ├── scripts/
 │   └── build-webui.sh       # Manual WebUI build script
@@ -476,7 +487,7 @@ 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`, `darwin/amd64`, `darwin/arm64`, `windows/amd64`) + `sha256sums.txt`
 - **Container publishing**: GHCR only (`ghcr.io/cjackhwang/ds2api`)
-- **Each archive includes**: `ds2api` executable, `static/admin`, WASM file, config template, README, LICENSE
+- **Each archive includes**: `ds2api` executable, `static/admin`, WASM file (with embedded fallback support), config template, README, LICENSE
 
 ## Disclaimer
 

TESTING.md

@@ -51,7 +51,7 @@ DS2API 提供两个层级的测试：
 1. **Preflight 检查**：
    - `go test ./... -count=1`（单元测试）
    - `./tests/scripts/check-node-split-syntax.sh`（Node 拆分模块语法门禁）
-   - `node --test`（如仓库存在 Node 单测文件时执行；当前默认以 Go 测试 + Node 语法门禁为主）
+   - `node --test tests/node/stream-tool-sieve.test.js tests/node/chat-stream.test.js tests/node/js_compat_test.js`
    - `npm run build --prefix webui`（WebUI 构建检查）
 
 2. **隔离启动**：复制 `config.json` 到临时目录，启动独立服务进程
@@ -173,6 +173,50 @@ rg "<trace_id>" artifacts/testsuite/<run_id>/server.log
 go test ./...
 ```
 
+### 运行特定模块的单元测试
+
+```bash
+# 运行 tool calls 相关测试（推荐用于调试 tool call 解析问题）
+go test -v -run 'TestParseToolCalls|TestRepair' ./internal/util/
+
+# 运行单个测试用例
+go test -v -run TestParseToolCallsWithDeepSeekHallucination ./internal/util/
+
+# 运行 format 相关测试
+go test -v ./internal/format/...
+
+# 运行 adapter 相关测试
+go test -v ./internal/adapter/openai/...
+```
+
+### 调试 Tool Call 问题 | Debugging Tool Call Issues
+
+当遇到 DeepSeek 工具调用解析问题时，可以使用以下方法：
+
+```bash
+# 1. 运行 tool calls 相关的所有测试
+go test -v -run 'TestParseToolCalls|TestRepair' ./internal/util/
+
+# 2. 查看测试输出中的详细调试信息
+go test -v -run TestParseToolCallsWithDeepSeekHallucination ./internal/util/ 2>&1
+
+# 3. 检查具体测试用例的修复效果
+# 测试用例位于 internal/util/toolcalls_test.go，包含：
+# - TestParseToolCallsWithDeepSeekHallucination: DeepSeek 典型幻觉输出
+# - TestRepairLooseJSONWithNestedObjects: 嵌套对象的方括号修复
+# - TestParseToolCallsWithMixedWindowsPaths: Windows 路径处理
+```
+
+### 运行 Node.js 测试
+
+```bash
+# 运行 Node 测试
+node --test tests/node/stream-tool-sieve.test.js
+
+# 或使用脚本
+./tests/scripts/run-unit-node.sh
+```
+
 ### 跑端到端测试（跳过 preflight）
 
 ```bash

VERSION

@@ -1 +1 @@
-0.1.0
+2.3.7

config.example.json

@@ -9,20 +9,17 @@
     {
       "_comment": "邮箱登录方式",
       "email": "example1@example.com",
-      "password": "your-password-1",
-      "token": ""
+      "password": "your-password-1"
     },
     {
       "_comment": "邮箱登录方式 - 账号2",
       "email": "example2@example.com",
-      "password": "your-password-2",
-      "token": ""
+      "password": "your-password-2"
     },
     {
       "_comment": "手机号登录方式（中国大陆）",
       "mobile": "12345678901",
-      "password": "your-password-3",
-      "token": ""
+      "password": "your-password-3"
     }
   ],
   "model_aliases": {
@@ -43,8 +40,19 @@
   "embeddings": {
     "provider": "deterministic"
   },
-  "claude_model_mapping": {
+  "claude_mapping": {
     "fast": "deepseek-chat",
     "slow": "deepseek-reasoner"
+  },
+  "admin": {
+    "jwt_expire_hours": 24
+  },
+  "runtime": {
+    "account_max_inflight": 2,
+    "account_max_queue": 0,
+    "global_max_inflight": 0
+  },
+  "auto_delete": {
+    "sessions": false
   }
 }

docs/toolcall-semantics.md

@@ -0,0 +1,41 @@
+# Tool call parsing semantics (Go canonical spec)
+
+This document defines the cross-runtime contract for `ParseToolCallsDetailed` / `parseToolCallsDetailed`.
+
+## Output contract
+
+- `calls`: accepted tool calls with normalized tool names.
+- `sawToolCallSyntax`: true when tool-call-like syntax is detected (`tool_calls`, `<tool_call>`, `<function_call>`, `<invoke>`) or a valid call is parsed.
+- `rejectedByPolicy`: true when parser extracted call syntax but all calls are rejected by allow-list policy.
+- `rejectedToolNames`: de-duplicated rejected tool names in first-seen order.
+
+## Parse pipeline
+
+1. Strip fenced code blocks for non-standalone parsing.
+2. Build candidates from:
+   - full text,
+   - fenced JSON snippets,
+   - extracted JSON objects around `tool_calls`,
+   - first `{` to last `}` object slice.
+3. Parse each candidate in order:
+   - JSON payload parser (`tool_calls`, list, single call object),
+   - XML/Markup parser (`<tool_call>`, `<function_call>`, `<invoke>`; supports attributes + nested fields),
+   - Text KV fallback parser (`function.name: <name>` ... `function.arguments: {json}`).
+4. Stop at first candidate that yields at least one call.
+
+## Name normalization policy
+
+When matching parsed names against configured tools:
+
+1. exact match,
+2. case-insensitive match,
+3. namespace tail match (`a.b.c` => `c`),
+4. loose alnum match (remove non `[a-z0-9]`, compare).
+
+## Standalone mode
+
+Standalone mode (`ParseStandaloneToolCallsDetailed`) parses the whole input directly (no candidate slicing), while still applying:
+
+- example-context guard,
+- JSON then markup fallback,
+- the same allow-list normalization policy.

internal/account/pool_test.go

@@ -194,7 +194,7 @@ func TestPoolAccountConcurrencyAliasEnv(t *testing.T) {
 	}
 }
 
-func TestPoolSupportsTokenOnlyAccount(t *testing.T) {
+func TestPoolDropsLegacyTokenOnlyAccountOnLoad(t *testing.T) {
 	t.Setenv("DS2API_ACCOUNT_MAX_INFLIGHT", "1")
 	t.Setenv("DS2API_CONFIG_JSON", `{
 		"keys":["k1"],
@@ -203,19 +203,15 @@ func TestPoolSupportsTokenOnlyAccount(t *testing.T) {
 
 	pool := NewPool(config.LoadStore())
 	status := pool.Status()
-	if got, ok := status["total"].(int); !ok || got != 1 {
+	if got, ok := status["total"].(int); !ok || got != 0 {
 		t.Fatalf("unexpected total in pool status: %#v", status["total"])
 	}
-	if got, ok := status["available"].(int); !ok || got != 1 {
+	if got, ok := status["available"].(int); !ok || got != 0 {
 		t.Fatalf("unexpected available in pool status: %#v", status["available"])
 	}
 
-	acc, ok := pool.Acquire("", nil)
-	if !ok {
-		t.Fatalf("expected acquire success for token-only account")
-	}
-	if acc.Token != "token-only-account" {
-		t.Fatalf("unexpected token on acquired account: %q", acc.Token)
+	if _, ok := pool.Acquire("", nil); ok {
+		t.Fatalf("expected acquire to fail for token-only account")
 	}
 }
 

internal/adapter/claude/handler_stream_test.go

@@ -315,3 +315,83 @@ func asString(v any) string {
 	s, _ := v.(string)
 	return s
 }
+
+func TestHandleClaudeStreamRealtimeToolSafetyAcrossStructuredFormats(t *testing.T) {
+	tests := []struct {
+		name    string
+		payload string
+	}{
+		{name: "xml_tool_call", payload: `<tool_call><tool_name>Bash</tool_name><parameters><command>pwd</command></parameters></tool_call>`},
+		{name: "xml_json_tool_call", payload: `<tool_call>{"tool":"Bash","params":{"command":"pwd"}}</tool_call>`},
+		{name: "nested_tool_tag_style", payload: `<tool_call><tool name="Bash"><command>pwd</command></tool></tool_call>`},
+		{name: "function_tag_style", payload: `<function_call>Bash</function_call><function parameter name="command">pwd</function parameter>`},
+		{name: "antml_argument_style", payload: `<antml:function_calls><antml:function_call id="1" name="Bash"><antml:argument name="command">pwd</antml:argument></antml:function_call></antml:function_calls>`},
+		{name: "antml_function_attr_parameters", payload: `<antml:function_calls><antml:function_call id="1" function="Bash"><antml:parameters>{"command":"pwd"}</antml:parameters></antml:function_call></antml:function_calls>`},
+		{name: "invoke_parameter_style", payload: `<function_calls><invoke name="Bash"><parameter name="command">pwd</parameter></invoke></function_calls>`},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			h := &Handler{}
+			resp := makeClaudeSSEHTTPResponse(
+				`data: {"p":"response/content","v":"`+strings.ReplaceAll(tc.payload, `"`, `\"`)+`"}`,
+				`data: [DONE]`,
+			)
+			rec := httptest.NewRecorder()
+			req := httptest.NewRequest(http.MethodPost, "/anthropic/v1/messages", nil)
+
+			h.handleClaudeStreamRealtime(rec, req, resp, "claude-sonnet-4-5", []any{map[string]any{"role": "user", "content": "use tool"}}, false, false, []string{"Bash"})
+
+			frames := parseClaudeFrames(t, rec.Body.String())
+			foundToolUse := false
+			for _, f := range findClaudeFrames(frames, "content_block_start") {
+				contentBlock, _ := f.Payload["content_block"].(map[string]any)
+				if contentBlock["type"] == "tool_use" {
+					foundToolUse = true
+					break
+				}
+			}
+			if !foundToolUse {
+				t.Fatalf("expected tool_use block for format %s, body=%s", tc.name, rec.Body.String())
+			}
+		})
+	}
+}
+
+func TestHandleClaudeStreamRealtimePromotesUnclosedFencedToolExample(t *testing.T) {
+	h := &Handler{}
+	resp := makeClaudeSSEHTTPResponse(
+		"data: {\"p\":\"response/content\",\"v\":\"Here is an example:\\n```json\\n{\\\"tool_calls\\\":[{\\\"name\\\":\\\"Bash\\\",\\\"input\\\":{\\\"command\\\":\\\"pwd\\\"}}]}\"}",
+		"data: {\"p\":\"response/content\",\"v\":\"\\n```\\nDo not execute it.\"}",
+		`data: [DONE]`,
+	)
+	rec := httptest.NewRecorder()
+	req := httptest.NewRequest(http.MethodPost, "/anthropic/v1/messages", nil)
+
+	h.handleClaudeStreamRealtime(rec, req, resp, "claude-sonnet-4-5", []any{map[string]any{"role": "user", "content": "show example only"}}, false, false, []string{"Bash"})
+
+	frames := parseClaudeFrames(t, rec.Body.String())
+	foundToolUse := false
+	for _, f := range findClaudeFrames(frames, "content_block_start") {
+		contentBlock, _ := f.Payload["content_block"].(map[string]any)
+		if contentBlock["type"] == "tool_use" {
+			foundToolUse = true
+			break
+		}
+	}
+	if !foundToolUse {
+		t.Fatalf("expected tool_use for fenced example, body=%s", rec.Body.String())
+	}
+
+	foundToolStop := false
+	for _, f := range findClaudeFrames(frames, "message_delta") {
+		delta, _ := f.Payload["delta"].(map[string]any)
+		if delta["stop_reason"] == "tool_use" {
+			foundToolStop = true
+			break
+		}
+	}
+	if !foundToolStop {
+		t.Fatalf("expected stop_reason=tool_use, body=%s", rec.Body.String())
+	}
+}

internal/adapter/claude/handler_util_test.go

@@ -125,8 +125,11 @@ func TestBuildClaudeToolPromptSingleTool(t *testing.T) {
 	if !containsStr(prompt, "Search the web") {
 		t.Fatalf("expected description in prompt")
 	}
-	if !containsStr(prompt, "tool_calls") {
-		t.Fatalf("expected tool_calls instruction in prompt")
+	if !containsStr(prompt, "tool_use") {
+		t.Fatalf("expected tool_use instruction in prompt")
+	}
+	if containsStr(prompt, "tool_calls") {
+		t.Fatalf("expected prompt to avoid tool_calls JSON instruction")
 	}
 }
 

internal/adapter/claude/handler_utils.go

@@ -51,7 +51,7 @@ func buildClaudeToolPrompt(tools []any) string {
 		parts = append(parts, fmt.Sprintf("Tool: %s\nDescription: %s\nParameters: %s", name, desc, schema))
 	}
 	parts = append(parts,
-		"When you need to use tools, you can call multiple tools in one response. Output ONLY JSON like {\"tool_calls\":[{\"name\":\"tool\",\"input\":{}}]}",
+		"When you need a tool, respond with Claude-native tool use (tool_use) using the provided tool schema. Do not print tool-call JSON in text.",
 		"History markers in conversation: [TOOL_CALL_HISTORY]...[/TOOL_CALL_HISTORY] are your previous tool calls; [TOOL_RESULT_HISTORY]...[/TOOL_RESULT_HISTORY] are runtime tool outputs, not user input.",
 		"After a valid [TOOL_RESULT_HISTORY], continue with final answer instead of repeating the same call unless required fields are still missing.",
 	)

internal/adapter/claude/stream_runtime_core.go

@@ -8,6 +8,7 @@ import (
 
 	"ds2api/internal/sse"
 	streamengine "ds2api/internal/stream"
+	"ds2api/internal/util"
 )
 
 type claudeStreamRuntime struct {
@@ -116,6 +117,18 @@ func (s *claudeStreamRuntime) onParsed(parsed sse.LineResult) streamengine.Parse
 
 		s.text.WriteString(p.Text)
 		if s.bufferToolContent {
+			if hasUnclosedCodeFence(s.text.String()) {
+				continue
+			}
+			detected := util.ParseToolCalls(s.text.String(), s.toolNames)
+			if len(detected) > 0 {
+				s.finalize("tool_use")
+				return streamengine.ParsedDecision{
+					ContentSeen: true,
+					Stop:        true,
+					StopReason:  streamengine.StopReason("tool_use_detected"),
+				}
+			}
 			continue
 		}
 		s.closeThinkingBlock()
@@ -144,3 +157,7 @@ func (s *claudeStreamRuntime) onParsed(parsed sse.LineResult) streamengine.Parse
 
 	return streamengine.ParsedDecision{ContentSeen: contentSeen}
 }
+
+func hasUnclosedCodeFence(text string) bool {
+	return strings.Count(text, "```")%2 == 1
+}

internal/adapter/openai/chat_stream_runtime.go

@@ -98,11 +98,11 @@ func (s *chatStreamRuntime) sendDone() {
 func (s *chatStreamRuntime) finalize(finishReason string) {
 	finalThinking := s.thinking.String()
 	finalText := s.text.String()
-	detected := util.ParseStandaloneToolCalls(finalText, s.toolNames)
-	if len(detected) > 0 && !s.toolCallsDoneEmitted {
+	detected := util.ParseStandaloneToolCallsDetailed(finalText, s.toolNames)
+	if len(detected.Calls) > 0 && !s.toolCallsDoneEmitted {
 		finishReason = "tool_calls"
 		delta := map[string]any{
-			"tool_calls": formatFinalStreamToolCallsWithStableIDs(detected, s.streamToolCallIDs),
+			"tool_calls": formatFinalStreamToolCallsWithStableIDs(detected.Calls, s.streamToolCallIDs),
 		}
 		if !s.firstChunkSent {
 			delta["role"] = "assistant"
@@ -158,7 +158,7 @@ func (s *chatStreamRuntime) finalize(finishReason string) {
 		}
 	}
 
-	if len(detected) > 0 || s.toolCallsEmitted {
+	if len(detected.Calls) > 0 || s.toolCallsEmitted {
 		finishReason = "tool_calls"
 	}
 	s.sendChunk(openaifmt.BuildChatStreamChunk(

internal/adapter/openai/deps.go

@@ -19,6 +19,7 @@ type DeepSeekCaller interface {
 	CreateSession(ctx context.Context, a *auth.RequestAuth, maxAttempts int) (string, error)
 	GetPow(ctx context.Context, a *auth.RequestAuth, maxAttempts int) (string, error)
 	CallCompletion(ctx context.Context, a *auth.RequestAuth, payload map[string]any, powResp string, maxAttempts int) (*http.Response, error)
+	DeleteAllSessionsForToken(ctx context.Context, token string) error
 }
 
 type ConfigReader interface {
@@ -28,6 +29,7 @@ type ConfigReader interface {
 	ToolcallEarlyEmitConfidence() string
 	ResponsesStoreTTLSeconds() int
 	EmbeddingsProvider() string
+	AutoDeleteSessions() bool
 }
 
 var _ AuthResolver = (*auth.Resolver)(nil)

internal/adapter/openai/deps_injection_test.go

@@ -19,6 +19,7 @@ func (m mockOpenAIConfig) ToolcallMode() string                { return m.toolMo
 func (m mockOpenAIConfig) ToolcallEarlyEmitConfidence() string { return m.earlyEmit }
 func (m mockOpenAIConfig) ResponsesStoreTTLSeconds() int       { return m.responsesTTL }
 func (m mockOpenAIConfig) EmbeddingsProvider() string          { return m.embedProv }
+func (m mockOpenAIConfig) AutoDeleteSessions() bool            { return false }
 
 func TestNormalizeOpenAIChatRequestWithConfigInterface(t *testing.T) {
 	cfg := mockOpenAIConfig{

internal/adapter/openai/handler_chat.go

@@ -35,7 +35,25 @@ func (h *Handler) ChatCompletions(w http.ResponseWriter, r *http.Request) {
 		writeOpenAIError(w, status, detail)
 		return
 	}
-	defer h.Auth.Release(a)
+	defer func() {
+		// 自动删除会话（同步）
+		// 必须在 Release 之前同步删除，否则：
+		// 1. 异步删除时账号已被 Release
+		// 2. 新请求可能获取到同一账号并开始使用
+		// 3. 异步删除仍在进行，会截断新请求正在使用的会话
+		if h.Store.AutoDeleteSessions() && a.DeepSeekToken != "" {
+			deleteCtx, cancel := context.WithTimeout(r.Context(), 10*time.Second)
+			defer cancel()
+			err := h.DS.DeleteAllSessionsForToken(deleteCtx, a.DeepSeekToken)
+			if err != nil {
+				config.Logger.Warn("[auto_delete_sessions] failed", "account", a.AccountID, "error", err)
+			} else {
+				config.Logger.Debug("[auto_delete_sessions] success", "account", a.AccountID)
+			}
+		}
+		h.Auth.Release(a)
+	}()
+
 	r = r.WithContext(auth.WithAuth(r.Context(), a))
 
 	var req map[string]any

internal/adapter/openai/handler_toolcall_format.go

@@ -53,7 +53,7 @@ func injectToolPrompt(messages []map[string]any, tools []any, policy util.ToolCh
 	if len(toolSchemas) == 0 {
 		return messages, names
 	}
-	toolPrompt := "You have access to these tools:\n\n" + strings.Join(toolSchemas, "\n\n") + "\n\nWhen you need to use tools, output ONLY this JSON format (no other text):\n{\"tool_calls\": [{\"name\": \"tool_name\", \"input\": {\"param\": \"value\"}}]}\n\nHistory markers in conversation:\n- [TOOL_CALL_HISTORY]...[/TOOL_CALL_HISTORY] means a tool call you already made earlier.\n- [TOOL_RESULT_HISTORY]...[/TOOL_RESULT_HISTORY] means the runtime returned a tool result (not user input).\n\nIMPORTANT:\n1) If calling tools, output ONLY the JSON. The response must start with { and end with }.\n2) After receiving a tool result, you MUST use it to produce the final answer.\n3) Only call another tool when the previous result is missing required data or returned an error.\n4) Do not repeat a tool call that is already satisfied by an existing [TOOL_RESULT_HISTORY] block."
+	toolPrompt := "You have access to these tools:\n\n" + strings.Join(toolSchemas, "\n\n") + "\n\nWhen you need to use tools, output ONLY a JSON code block like this:\n```json\n{\"tool_calls\": [{\"name\": \"tool_name\", \"input\": {\"param\": \"value\"}}]}\n```\n\n【EXAMPLE】\nUser: Please check the weather in Beijing and Shanghai, and update my todo list.\nAssistant:\n```json\n{\"tool_calls\": [\n  {\"name\": \"get_weather\", \"input\": {\"city\": \"Beijing\"}},\n  {\"name\": \"get_weather\", \"input\": {\"city\": \"Shanghai\"}},\n  {\"name\": \"update_todo\", \"input\": {\"todos\": [{\"content\": \"Buy milk\"}, {\"content\": \"Write report\"}]}}\n]}\n```\n\nHistory markers in conversation:\n- [TOOL_CALL_HISTORY]...[/TOOL_CALL_HISTORY] means a tool call you already made earlier.\n- [TOOL_RESULT_HISTORY]...[/TOOL_RESULT_HISTORY] means the runtime returned a tool result (not user input).\n\nIMPORTANT:\n1) If calling tools, output ONLY the JSON code block. The response must start with ```json and end with ```.\n2) After receiving a tool result, you MUST use it to produce the final answer.\n3) Only call another tool when the previous result is missing required data or returned an error.\n4) Do not repeat a tool call that is already satisfied by an existing [TOOL_RESULT_HISTORY] block.\n5) JSON SYNTAX STRICTLY REQUIRED: All property names MUST be enclosed in double quotes (e.g., \"name\", not name).\n6) ARRAY FORMAT: If providing a list of items, you MUST enclose them in square brackets `[]` (e.g., \"todos\": [{\"item\": \"a\"}, {\"item\": \"b\"}]). DO NOT output comma-separated objects without brackets."
 	if policy.Mode == util.ToolChoiceRequired {
 		toolPrompt += "\n5) For this response, you MUST call at least one tool from the allowed list."
 	}

internal/adapter/openai/handler_toolcall_test.go

@@ -211,7 +211,7 @@ func TestHandleNonStreamUnknownToolNotIntercepted(t *testing.T) {
 	}
 }
 
-func TestHandleNonStreamEmbeddedToolCallExampleRemainsText(t *testing.T) {
+func TestHandleNonStreamEmbeddedToolCallExamplePromotesToolCall(t *testing.T) {
 	h := &Handler{}
 	resp := makeSSEHTTPResponse(
 		`data: {"p":"response/content","v":"下面是示例："}`,
@@ -229,20 +229,21 @@ func TestHandleNonStreamEmbeddedToolCallExampleRemainsText(t *testing.T) {
 	out := decodeJSONBody(t, rec.Body.String())
 	choices, _ := out["choices"].([]any)
 	choice, _ := choices[0].(map[string]any)
-	if choice["finish_reason"] != "stop" {
-		t.Fatalf("expected finish_reason=stop, got %#v", choice["finish_reason"])
+	if choice["finish_reason"] != "tool_calls" {
+		t.Fatalf("expected finish_reason=tool_calls, got %#v", choice["finish_reason"])
 	}
 	msg, _ := choice["message"].(map[string]any)
-	if _, ok := msg["tool_calls"]; ok {
-		t.Fatalf("did not expect tool_calls field for embedded example: %#v", msg["tool_calls"])
+	toolCalls, _ := msg["tool_calls"].([]any)
+	if len(toolCalls) != 1 {
+		t.Fatalf("expected one tool_call field for embedded example: %#v", msg["tool_calls"])
 	}
 	content, _ := msg["content"].(string)
-	if !strings.Contains(content, "下面是示例：") || !strings.Contains(content, "请勿执行。") || !strings.Contains(content, `"tool_calls"`) {
-		t.Fatalf("expected embedded example to remain plain text, got %#v", content)
+	if strings.Contains(content, `"tool_calls"`) {
+		t.Fatalf("expected raw tool_calls json stripped from content, got %#v", content)
 	}
 }
 
-func TestHandleNonStreamFencedToolCallExampleNotIntercepted(t *testing.T) {
+func TestHandleNonStreamFencedToolCallExamplePromotesToolCall(t *testing.T) {
 	h := &Handler{}
 	resp := makeSSEHTTPResponse(
 		"data: {\"p\":\"response/content\",\"v\":\"```json\\n{\\\"tool_calls\\\":[{\\\"name\\\":\\\"search\\\",\\\"input\\\":{\\\"q\\\":\\\"go\\\"}}]}\\n```\"}",
@@ -258,16 +259,17 @@ func TestHandleNonStreamFencedToolCallExampleNotIntercepted(t *testing.T) {
 	out := decodeJSONBody(t, rec.Body.String())
 	choices, _ := out["choices"].([]any)
 	choice, _ := choices[0].(map[string]any)
-	if choice["finish_reason"] != "stop" {
-		t.Fatalf("expected finish_reason=stop, got %#v", choice["finish_reason"])
+	if choice["finish_reason"] != "tool_calls" {
+		t.Fatalf("expected finish_reason=tool_calls, got %#v", choice["finish_reason"])
 	}
 	msg, _ := choice["message"].(map[string]any)
-	if _, ok := msg["tool_calls"]; ok {
-		t.Fatalf("did not expect tool_calls field for fenced example: %#v", msg["tool_calls"])
+	toolCalls, _ := msg["tool_calls"].([]any)
+	if len(toolCalls) != 1 {
+		t.Fatalf("expected one tool_call field for fenced example: %#v", msg["tool_calls"])
 	}
 	content, _ := msg["content"].(string)
-	if !strings.Contains(content, "```json") || !strings.Contains(content, `"tool_calls"`) {
-		t.Fatalf("expected fenced tool example to pass through as text, got %q", content)
+	if strings.Contains(content, `"tool_calls"`) {
+		t.Fatalf("expected raw tool_calls json stripped from content, got %q", content)
 	}
 }
 
@@ -615,7 +617,7 @@ func TestHandleStreamToolCallWithSameChunkTrailingTextRemainsText(t *testing.T)
 	}
 }
 
-func TestHandleStreamFencedToolCallSnippetRemainsText(t *testing.T) {
+func TestHandleStreamFencedToolCallSnippetPromotesToolCall(t *testing.T) {
 	h := &Handler{}
 	resp := makeSSEHTTPResponse(
 		fmt.Sprintf(`data: {"p":"response/content","v":%q}`, "下面是调用示例：\n```json\n"),
@@ -631,8 +633,8 @@ func TestHandleStreamFencedToolCallSnippetRemainsText(t *testing.T) {
 	if !done {
 		t.Fatalf("expected [DONE], body=%s", rec.Body.String())
 	}
-	if streamHasToolCallsDelta(frames) {
-		t.Fatalf("did not expect tool_calls delta for fenced snippet, body=%s", rec.Body.String())
+	if !streamHasToolCallsDelta(frames) {
+		t.Fatalf("expected tool_calls delta for fenced snippet, body=%s", rec.Body.String())
 	}
 	content := strings.Builder{}
 	for _, frame := range frames {
@@ -646,11 +648,11 @@ func TestHandleStreamFencedToolCallSnippetRemainsText(t *testing.T) {
 		}
 	}
 	got := content.String()
-	if !strings.Contains(got, "```json") || !strings.Contains(strings.ToLower(got), "tool_calls") {
-		t.Fatalf("expected fenced tool snippet in content, got=%q", got)
+	if strings.Contains(strings.ToLower(got), "tool_calls") {
+		t.Fatalf("expected raw fenced tool_calls snippet stripped from content, got=%q", got)
 	}
-	if streamFinishReason(frames) != "stop" {
-		t.Fatalf("expected finish_reason=stop, body=%s", rec.Body.String())
+	if streamFinishReason(frames) != "tool_calls" {
+		t.Fatalf("expected finish_reason=tool_calls, body=%s", rec.Body.String())
 	}
 }
 

internal/adapter/openai/message_normalize.go

@@ -6,6 +6,7 @@ import (
 	"strings"
 
 	"ds2api/internal/config"
+	"ds2api/internal/prompt"
 )
 
 func normalizeOpenAIMessagesForPrompt(raw []any, traceID string) []map[string]any {
@@ -77,7 +78,7 @@ func formatAssistantToolCallsForPrompt(msg map[string]any, traceID string) strin
 				args = normalizeOpenAIArgumentsForPrompt(fn["arguments"])
 			}
 			if name == "" {
-				name = "unknown"
+				continue
 			}
 			if args == "" {
 				args = normalizeOpenAIArgumentsForPrompt(call["arguments"])
@@ -132,32 +133,7 @@ func formatToolResultForPrompt(msg map[string]any) string {
 }
 
 func normalizeOpenAIContentForPrompt(v any) string {
-	switch x := v.(type) {
-	case string:
-		return x
-	case []any:
-		parts := make([]string, 0, len(x))
-		for _, item := range x {
-			m, ok := item.(map[string]any)
-			if !ok {
-				continue
-			}
-			t := strings.ToLower(strings.TrimSpace(asString(m["type"])))
-			if t != "text" && t != "output_text" && t != "input_text" {
-				continue
-			}
-			if text := asString(m["text"]); text != "" {
-				parts = append(parts, text)
-				continue
-			}
-			if text := asString(m["content"]); text != "" {
-				parts = append(parts, text)
-			}
-		}
-		return strings.Join(parts, "\n")
-	default:
-		return marshalToPromptString(v)
-	}
+	return prompt.NormalizeContent(v)
 }
 
 func normalizeOpenAIArgumentsForPrompt(v any) string {

internal/adapter/openai/message_normalize_test.go

@@ -194,6 +194,59 @@ func TestNormalizeOpenAIMessagesForPrompt_PreservesConcatenatedToolArguments(t *
 	}
 }
 
+
+func TestNormalizeOpenAIMessagesForPrompt_AssistantToolCallsMissingNameAreDropped(t *testing.T) {
+	raw := []any{
+		map[string]any{
+			"role": "assistant",
+			"tool_calls": []any{
+				map[string]any{
+					"id":   "call_missing_name",
+					"type": "function",
+					"function": map[string]any{
+						"arguments": `{"path":"README.MD"}`,
+					},
+				},
+			},
+		},
+	}
+
+	normalized := normalizeOpenAIMessagesForPrompt(raw, "")
+	if len(normalized) != 0 {
+		t.Fatalf("expected nameless assistant tool_calls to be dropped, got %#v", normalized)
+	}
+}
+
+func TestNormalizeOpenAIMessagesForPrompt_AssistantNilContentDoesNotInjectNullLiteral(t *testing.T) {
+	raw := []any{
+		map[string]any{
+			"role":    "assistant",
+			"content": nil,
+			"tool_calls": []any{
+				map[string]any{
+					"id": "call_screenshot",
+					"function": map[string]any{
+						"name":      "send_file_to_user",
+						"arguments": `{"file_path":"/tmp/a.png"}`,
+					},
+				},
+			},
+		},
+	}
+
+	normalized := normalizeOpenAIMessagesForPrompt(raw, "")
+	if len(normalized) != 1 {
+		t.Fatalf("expected one normalized message, got %d", len(normalized))
+	}
+	content, _ := normalized[0]["content"].(string)
+	if strings.Contains(content, "<｜Assistant｜>null") || strings.HasPrefix(strings.TrimSpace(content), "null") {
+		t.Fatalf("unexpected null literal injected into assistant tool history: %q", content)
+	}
+	if !strings.Contains(content, "function.name: send_file_to_user") {
+		t.Fatalf("expected tool history block preserved, got %q", content)
+	}
+}
+
 func TestNormalizeOpenAIMessagesForPrompt_DeveloperRoleMapsToSystem(t *testing.T) {
 	raw := []any{
 		map[string]any{"role": "developer", "content": "必须先走工具调用"},
@@ -207,3 +260,23 @@ func TestNormalizeOpenAIMessagesForPrompt_DeveloperRoleMapsToSystem(t *testing.T
 		t.Fatalf("expected developer role converted to system, got %#v", normalized[0]["role"])
 	}
 }
+
+func TestNormalizeOpenAIMessagesForPrompt_AssistantArrayContentFallbackWhenTextEmpty(t *testing.T) {
+	raw := []any{
+		map[string]any{
+			"role": "assistant",
+			"content": []any{
+				map[string]any{"type": "text", "text": "", "content": "工具说明文本"},
+			},
+		},
+	}
+
+	normalized := normalizeOpenAIMessagesForPrompt(raw, "")
+	if len(normalized) != 1 {
+		t.Fatalf("expected one normalized message, got %d", len(normalized))
+	}
+	content, _ := normalized[0]["content"].(string)
+	if content != "工具说明文本" {
+		t.Fatalf("expected content fallback text preserved, got %q", content)
+	}
+}

internal/adapter/openai/responses_stream_runtime_toolcalls.go

@@ -94,6 +94,16 @@ func (s *responsesStreamRuntime) closeMessageItem() {
 	outputIndex := s.ensureMessageOutputIndex()
 	text := s.visibleText.String()
 	if s.messagePartAdded {
+		s.sendEvent(
+			"response.output_text.done",
+			openaifmt.BuildResponsesTextDonePayload(
+				s.responseID,
+				itemID,
+				outputIndex,
+				0,
+				text,
+			),
+		)
 		s.sendEvent(
 			"response.content_part.done",
 			openaifmt.BuildResponsesContentPartDonePayload(

internal/adapter/openai/responses_stream_test.go

@@ -226,6 +226,40 @@ func TestHandleResponsesStreamMultiToolCallKeepsNameAndCallIDAligned(t *testing.
 	}
 }
 
+func TestHandleResponsesStreamEmitsOutputTextDoneBeforeContentPartDone(t *testing.T) {
+	h := &Handler{}
+	req := httptest.NewRequest(http.MethodPost, "/v1/responses", nil)
+	rec := httptest.NewRecorder()
+
+	sseLine := func(v string) string {
+		b, _ := json.Marshal(map[string]any{
+			"p": "response/content",
+			"v": v,
+		})
+		return "data: " + string(b) + "\n"
+	}
+
+	streamBody := sseLine("hello") + "data: [DONE]\n"
+	resp := &http.Response{
+		StatusCode: http.StatusOK,
+		Body:       io.NopCloser(strings.NewReader(streamBody)),
+	}
+
+	h.handleResponsesStream(rec, req, resp, "owner-a", "resp_test", "deepseek-chat", "prompt", false, false, nil, util.DefaultToolChoicePolicy(), "")
+	body := rec.Body.String()
+	if !strings.Contains(body, "event: response.output_text.done") {
+		t.Fatalf("expected response.output_text.done payload, body=%s", body)
+	}
+	textDoneIdx := strings.Index(body, "event: response.output_text.done")
+	partDoneIdx := strings.Index(body, "event: response.content_part.done")
+	if textDoneIdx < 0 || partDoneIdx < 0 {
+		t.Fatalf("expected output_text.done + content_part.done, body=%s", body)
+	}
+	if textDoneIdx > partDoneIdx {
+		t.Fatalf("expected output_text.done before content_part.done, body=%s", body)
+	}
+}
+
 func TestHandleResponsesStreamOutputTextDeltaCarriesItemIndexes(t *testing.T) {
 	h := &Handler{}
 	req := httptest.NewRequest(http.MethodPost, "/v1/responses", nil)
@@ -263,7 +297,7 @@ func TestHandleResponsesStreamOutputTextDeltaCarriesItemIndexes(t *testing.T) {
 	}
 }
 
-func TestHandleResponsesStreamThinkingAndMixedToolExampleRemainMessageOnly(t *testing.T) {
+func TestHandleResponsesStreamThinkingAndMixedToolExampleEmitsFunctionCall(t *testing.T) {
 	h := &Handler{}
 	req := httptest.NewRequest(http.MethodPost, "/v1/responses", nil)
 	rec := httptest.NewRecorder()
@@ -299,6 +333,7 @@ func TestHandleResponsesStreamThinkingAndMixedToolExampleRemainMessageOnly(t *te
 	responseObj, _ := completedPayload["response"].(map[string]any)
 	output, _ := responseObj["output"].([]any)
 	hasMessage := false
+	hasFunctionCall := false
 	for _, item := range output {
 		m, _ := item.(map[string]any)
 		if m == nil {
@@ -308,12 +343,15 @@ func TestHandleResponsesStreamThinkingAndMixedToolExampleRemainMessageOnly(t *te
 			hasMessage = true
 		}
 		if asString(m["type"]) == "function_call" {
-			t.Fatalf("did not expect function_call output for mixed prose tool example, output=%#v", output)
+			hasFunctionCall = true
 		}
 	}
 	if !hasMessage {
 		t.Fatalf("expected message output for mixed prose tool example, output=%#v", output)
 	}
+	if !hasFunctionCall {
+		t.Fatalf("expected function_call output for mixed prose tool example, output=%#v", output)
+	}
 }
 
 func TestHandleResponsesStreamToolChoiceNoneRejectsFunctionCall(t *testing.T) {

internal/adapter/openai/stream_status_test.go

@@ -53,6 +53,10 @@ func (m streamStatusDSStub) CallCompletion(_ context.Context, _ *auth.RequestAut
 	return m.resp, nil
 }
 
+func (m streamStatusDSStub) DeleteAllSessionsForToken(_ context.Context, _ string) error {
+	return nil
+}
+
 func makeOpenAISSEHTTPResponse(lines ...string) *http.Response {
 	body := strings.Join(lines, "\n")
 	if !strings.HasSuffix(body, "\n") {
@@ -167,15 +171,15 @@ func TestResponsesNonStreamMixedProseToolPayloadHandlerPath(t *testing.T) {
 		t.Fatalf("decode response failed: %v body=%s", err, rec.Body.String())
 	}
 	outputText, _ := out["output_text"].(string)
-	if outputText == "" {
-		t.Fatalf("expected output_text preserved for mixed prose payload")
+	if outputText != "" {
+		t.Fatalf("expected output_text hidden for mixed prose tool payload, got %q", outputText)
 	}
 	output, _ := out["output"].([]any)
 	if len(output) != 1 {
 		t.Fatalf("expected one output item, got %#v", output)
 	}
 	first, _ := output[0].(map[string]any)
-	if first["type"] != "message" {
-		t.Fatalf("expected message output item, got %#v", output)
+	if first["type"] != "function_call" {
+		t.Fatalf("expected function_call output item, got %#v", output)
 	}
 }

internal/adapter/openai/tool_sieve_core.go

@@ -167,22 +167,22 @@ func findToolSegmentStart(s string) int {
 		return -1
 	}
 	lower := strings.ToLower(s)
-	offset := 0
-	for {
-		keyRel := strings.Index(lower[offset:], "tool_calls")
-		if keyRel < 0 {
-			return -1
+	keywords := []string{"tool_calls", "function.name:", "[tool_call_history]"}
+	bestKeyIdx := -1
+	for _, kw := range keywords {
+		idx := strings.Index(lower, kw)
+		if idx >= 0 && (bestKeyIdx < 0 || idx < bestKeyIdx) {
+			bestKeyIdx = idx
 		}
-		keyIdx := offset + keyRel
-		start := strings.LastIndex(s[:keyIdx], "{")
-		if start < 0 {
-			start = keyIdx
-		}
-		if !insideCodeFence(s[:start]) {
-			return start
-		}
-		offset = keyIdx + len("tool_calls")
 	}
+	if bestKeyIdx < 0 {
+		return -1
+	}
+	start := strings.LastIndex(s[:bestKeyIdx], "{")
+	if start < 0 {
+		start = bestKeyIdx
+	}
+	return start
 }
 
 func consumeToolCapture(state *toolStreamSieveState, toolNames []string) (prefix string, calls []util.ParsedToolCall, suffix string, ready bool) {
@@ -191,13 +191,22 @@ func consumeToolCapture(state *toolStreamSieveState, toolNames []string) (prefix
 		return "", nil, "", false
 	}
 	lower := strings.ToLower(captured)
-	keyIdx := strings.Index(lower, "tool_calls")
+	
+	keyIdx := -1
+	keywords := []string{"tool_calls", "function.name:", "[tool_call_history]"}
+	for _, kw := range keywords {
+		idx := strings.Index(lower, kw)
+		if idx >= 0 && (keyIdx < 0 || idx < keyIdx) {
+			keyIdx = idx
+		}
+	}
+	
 	if keyIdx < 0 {
 		return "", nil, "", false
 	}
 	start := strings.LastIndex(captured[:keyIdx], "{")
 	if start < 0 {
-		return "", nil, "", false
+		start = keyIdx
 	}
 	obj, end, ok := extractJSONObjectFrom(captured, start)
 	if !ok {
@@ -205,9 +214,6 @@ func consumeToolCapture(state *toolStreamSieveState, toolNames []string) (prefix
 	}
 	prefixPart := captured[:start]
 	suffixPart := captured[end:]
-	if insideCodeFence(state.recentTextTail + prefixPart) {
-		return captured, nil, "", true
-	}
 	parsed := util.ParseStandaloneToolCallsDetailed(obj, toolNames)
 	if len(parsed.Calls) == 0 {
 		if parsed.SawToolCallSyntax && parsed.RejectedByPolicy {
@@ -215,6 +221,9 @@ func consumeToolCapture(state *toolStreamSieveState, toolNames []string) (prefix
 			// consume it to avoid leaking raw tool_calls JSON to user content.
 			return prefixPart, nil, suffixPart, true
 		}
+		// If it has obvious keywords but failed to parse even after loose repair,
+		// we still might want to intercept it if it looks like an attempt at tool call.
+		// For now, keep the original logic but rely on loose JSON repair.
 		return captured, nil, "", true
 	}
 	return prefixPart, parsed.Calls, suffixPart, true

internal/adapter/openai/tool_sieve_incremental.go

@@ -19,9 +19,6 @@ func buildIncrementalToolDeltas(state *toolStreamSieveState) []toolCallDelta {
 	if start < 0 {
 		return nil
 	}
-	if insideCodeFence(state.recentTextTail + captured[:start]) {
-		return nil
-	}
 	certainSingle, hasMultiple := classifyToolCallsIncrementalSafety(captured, keyIdx)
 	if hasMultiple {
 		state.disableDeltas = true

internal/admin/deps.go

@@ -27,6 +27,7 @@ type ConfigStore interface {
 	RuntimeAccountMaxInflight() int
 	RuntimeAccountMaxQueue(defaultSize int) int
 	RuntimeGlobalMaxInflight(defaultSize int) int
+	AutoDeleteSessions() bool
 }
 
 type PoolController interface {
@@ -40,6 +41,8 @@ type DeepSeekCaller interface {
 	CreateSession(ctx context.Context, a *auth.RequestAuth, maxAttempts int) (string, error)
 	GetPow(ctx context.Context, a *auth.RequestAuth, maxAttempts int) (string, error)
 	CallCompletion(ctx context.Context, a *auth.RequestAuth, payload map[string]any, powResp string, maxAttempts int) (*http.Response, error)
+	GetSessionCountForToken(ctx context.Context, token string) (*deepseek.SessionStats, error)
+	DeleteAllSessionsForToken(ctx context.Context, token string) error
 }
 
 var _ ConfigStore = (*config.Store)(nil)

internal/admin/handler.go

@@ -31,6 +31,7 @@ func RegisterRoutes(r chi.Router, h *Handler) {
 		pr.Get("/queue/status", h.queueStatus)
 		pr.Post("/accounts/test", h.testSingleAccount)
 		pr.Post("/accounts/test-all", h.testAllAccounts)
+		pr.Post("/accounts/sessions/delete-all", h.deleteAllSessions)
 		pr.Post("/import", h.batchImport)
 		pr.Post("/test", h.testAPI)
 		pr.Post("/vercel/sync", h.syncVercel)
@@ -38,5 +39,6 @@ func RegisterRoutes(r chi.Router, h *Handler) {
 		pr.Get("/export", h.exportConfig)
 		pr.Get("/dev/captures", h.getDevCaptures)
 		pr.Delete("/dev/captures", h.clearDevCaptures)
+		pr.Get("/version", h.getVersion)
 	})
 }

internal/admin/handler_accounts_identifier_test.go

@@ -6,7 +6,6 @@ import (
 	"net/http"
 	"net/http/httptest"
 	"net/url"
-	"strings"
 	"testing"
 
 	"github.com/go-chi/chi/v5"
@@ -26,9 +25,9 @@ func newAdminTestHandler(t *testing.T, raw string) *Handler {
 	}
 }
 
-func TestListAccountsIncludesTokenOnlyIdentifier(t *testing.T) {
+func TestListAccountsUsesEmailIdentifier(t *testing.T) {
 	h := newAdminTestHandler(t, `{
-		"accounts":[{"token":"token-only-account"}]
+		"accounts":[{"email":"u@example.com","password":"pwd"}]
 	}`)
 
 	req := httptest.NewRequest(http.MethodGet, "/admin/accounts?page=1&page_size=10", nil)
@@ -49,38 +48,8 @@ func TestListAccountsIncludesTokenOnlyIdentifier(t *testing.T) {
 	}
 	first, _ := items[0].(map[string]any)
 	identifier, _ := first["identifier"].(string)
-	if identifier == "" {
-		t.Fatalf("expected non-empty identifier: %#v", first)
-	}
-	if !strings.HasPrefix(identifier, "token:") {
-		t.Fatalf("expected token synthetic identifier, got %q", identifier)
-	}
-}
-
-func TestDeleteAccountSupportsTokenOnlyIdentifier(t *testing.T) {
-	h := newAdminTestHandler(t, `{
-		"accounts":[{"token":"token-only-account"}]
-	}`)
-	accounts := h.Store.Accounts()
-	if len(accounts) != 1 {
-		t.Fatalf("expected 1 account, got %d", len(accounts))
-	}
-	id := accounts[0].Identifier()
-	if id == "" {
-		t.Fatal("expected token-only synthetic identifier")
-	}
-
-	r := chi.NewRouter()
-	r.Delete("/admin/accounts/{identifier}", h.deleteAccount)
-	req := httptest.NewRequest(http.MethodDelete, "/admin/accounts/"+url.PathEscape(id), nil)
-	rec := httptest.NewRecorder()
-	r.ServeHTTP(rec, req)
-
-	if rec.Code != http.StatusOK {
-		t.Fatalf("unexpected status: %d body=%s", rec.Code, rec.Body.String())
-	}
-	if got := len(h.Store.Accounts()); got != 0 {
-		t.Fatalf("expected account removed, remaining=%d", got)
+	if identifier != "u@example.com" {
+		t.Fatalf("expected email identifier, got %q", identifier)
 	}
 }
 
@@ -142,11 +111,10 @@ func TestAddAccountRejectsCanonicalMobileDuplicate(t *testing.T) {
 	}
 }
 
-func TestFindAccountByIdentifierSupportsMobileAndTokenOnly(t *testing.T) {
+func TestFindAccountByIdentifierSupportsMobile(t *testing.T) {
 	h := newAdminTestHandler(t, `{
 		"accounts":[
-			{"email":"u@example.com","mobile":"13800138000","password":"pwd"},
-			{"token":"token-only-account"}
+			{"email":"u@example.com","mobile":"13800138000","password":"pwd"}
 		]
 	}`)
 
@@ -165,21 +133,4 @@ func TestFindAccountByIdentifierSupportsMobileAndTokenOnly(t *testing.T) {
 		t.Fatalf("unexpected account by +86 mobile: %#v", accByMobileWithCountryCode)
 	}
 
-	tokenOnlyID := ""
-	for _, acc := range h.Store.Accounts() {
-		if strings.TrimSpace(acc.Email) == "" && strings.TrimSpace(acc.Mobile) == "" {
-			tokenOnlyID = acc.Identifier()
-			break
-		}
-	}
-	if tokenOnlyID == "" {
-		t.Fatal("expected token-only account identifier")
-	}
-	accByTokenOnly, ok := findAccountByIdentifier(h.Store, tokenOnlyID)
-	if !ok {
-		t.Fatalf("expected find by token-only id=%q", tokenOnlyID)
-	}
-	if accByTokenOnly.Token != "token-only-account" {
-		t.Fatalf("unexpected token-only account: %#v", accByTokenOnly)
-	}
 }

internal/admin/handler_accounts_testing.go

@@ -89,7 +89,15 @@ func runAccountTestsConcurrently(accounts []config.Account, maxConcurrency int,
 func (h *Handler) testAccount(ctx context.Context, acc config.Account, model, message string) map[string]any {
 	start := time.Now()
 	identifier := acc.Identifier()
-	result := map[string]any{"account": identifier, "success": false, "response_time": 0, "message": "", "model": model}
+	result := map[string]any{
+		"account":         identifier,
+		"success":         false,
+		"response_time":   0,
+		"message":         "",
+		"model":           model,
+		"session_count":   0,
+		"config_writable": !h.Store.IsEnvBacked(),
+	}
 	defer func() {
 		status := "failed"
 		if ok, _ := result["success"].(bool); ok {
@@ -97,15 +105,14 @@ func (h *Handler) testAccount(ctx context.Context, acc config.Account, model, me
 		}
 		_ = h.Store.UpdateAccountTestStatus(identifier, status)
 	}()
-	token := strings.TrimSpace(acc.Token)
-	if token == "" {
-		newToken, err := h.DS.Login(ctx, acc)
-		if err != nil {
-			result["message"] = "登录失败: " + err.Error()
-			return result
-		}
-		token = newToken
-		_ = h.Store.UpdateAccountToken(acc.Identifier(), token)
+	token, err := h.DS.Login(ctx, acc)
+	if err != nil {
+		result["message"] = "登录失败: " + err.Error()
+		return result
+	}
+	if err := h.Store.UpdateAccountToken(acc.Identifier(), token); err != nil {
+		result["message"] = "登录成功但写入运行时 token 失败: " + err.Error()
+		return result
 	}
 	authCtx := &authn.RequestAuth{UseConfigToken: false, DeepSeekToken: token}
 	sessionID, err := h.DS.CreateSession(ctx, authCtx, 1)
@@ -117,16 +124,26 @@ func (h *Handler) testAccount(ctx context.Context, acc config.Account, model, me
 		}
 		token = newToken
 		authCtx.DeepSeekToken = token
-		_ = h.Store.UpdateAccountToken(acc.Identifier(), token)
+		if err := h.Store.UpdateAccountToken(acc.Identifier(), token); err != nil {
+			result["message"] = "刷新 token 成功但写入运行时 token 失败: " + err.Error()
+			return result
+		}
 		sessionID, err = h.DS.CreateSession(ctx, authCtx, 1)
 		if err != nil {
 			result["message"] = "创建会话失败: " + err.Error()
 			return result
 		}
 	}
+
+	// 获取会话数量
+	sessionStats, sessionErr := h.DS.GetSessionCountForToken(ctx, token)
+	if sessionErr == nil && sessionStats != nil {
+		result["session_count"] = sessionStats.FirstPageCount
+	}
+
 	if strings.TrimSpace(message) == "" {
 		result["success"] = true
-		result["message"] = "API 测试成功（仅会话创建）"
+		result["message"] = "Token 刷新成功（登录与会话创建成功）"
 		result["response_time"] = int(time.Since(start).Milliseconds())
 		return result
 	}
@@ -210,3 +227,45 @@ func (h *Handler) testAPI(w http.ResponseWriter, r *http.Request) {
 	}
 	writeJSON(w, http.StatusOK, map[string]any{"success": false, "status_code": resp.StatusCode, "response": string(body)})
 }
+
+func (h *Handler) deleteAllSessions(w http.ResponseWriter, r *http.Request) {
+	var req map[string]any
+	_ = json.NewDecoder(r.Body).Decode(&req)
+	identifier, _ := req["identifier"].(string)
+	if strings.TrimSpace(identifier) == "" {
+		writeJSON(w, http.StatusBadRequest, map[string]any{"detail": "需要账号标识（identifier / email / mobile）"})
+		return
+	}
+	acc, ok := findAccountByIdentifier(h.Store, identifier)
+	if !ok {
+		writeJSON(w, http.StatusNotFound, map[string]any{"detail": "账号不存在"})
+		return
+	}
+
+	// 每次先登录刷新一次 token，避免使用过期 token。
+	token, err := h.DS.Login(r.Context(), acc)
+	if err != nil {
+		writeJSON(w, http.StatusOK, map[string]any{"success": false, "message": "登录失败: " + err.Error()})
+		return
+	}
+	_ = h.Store.UpdateAccountToken(acc.Identifier(), token)
+
+	// 删除所有会话
+	err = h.DS.DeleteAllSessionsForToken(r.Context(), token)
+	if err != nil {
+		// token 可能过期，尝试重新登录并重试一次
+		newToken, loginErr := h.DS.Login(r.Context(), acc)
+		if loginErr != nil {
+			writeJSON(w, http.StatusOK, map[string]any{"success": false, "message": "删除失败: " + err.Error()})
+			return
+		}
+		token = newToken
+		_ = h.Store.UpdateAccountToken(acc.Identifier(), token)
+		if retryErr := h.DS.DeleteAllSessionsForToken(r.Context(), token); retryErr != nil {
+			writeJSON(w, http.StatusOK, map[string]any{"success": false, "message": "删除失败: " + retryErr.Error()})
+			return
+		}
+	}
+
+	writeJSON(w, http.StatusOK, map[string]any{"success": true, "message": "删除成功"})
+}

internal/admin/handler_accounts_testing_test.go

@@ -1,21 +1,28 @@
 package admin
 
 import (
+	"bytes"
 	"context"
+	"encoding/json"
 	"errors"
 	"net/http"
+	"net/http/httptest"
 	"strings"
 	"testing"
 
 	"ds2api/internal/auth"
 	"ds2api/internal/config"
+	"ds2api/internal/deepseek"
 )
 
 type testingDSMock struct {
-	loginCalls          int
-	createSessionCalls  int
-	getPowCalls         int
-	callCompletionCalls int
+	loginCalls                 int
+	createSessionCalls         int
+	getPowCalls                int
+	callCompletionCalls        int
+	deleteAllSessionsCalls     int
+	deleteAllSessionsError     error
+	deleteAllSessionsErrorOnce bool
 }
 
 func (m *testingDSMock) Login(_ context.Context, _ config.Account) (string, error) {
@@ -38,6 +45,22 @@ func (m *testingDSMock) CallCompletion(_ context.Context, _ *auth.RequestAuth, _
 	return nil, errors.New("should not call CallCompletion in this test")
 }
 
+func (m *testingDSMock) DeleteAllSessionsForToken(_ context.Context, _ string) error {
+	m.deleteAllSessionsCalls++
+	if m.deleteAllSessionsError != nil {
+		err := m.deleteAllSessionsError
+		if m.deleteAllSessionsErrorOnce {
+			m.deleteAllSessionsError = nil
+		}
+		return err
+	}
+	return nil
+}
+
+func (m *testingDSMock) GetSessionCountForToken(_ context.Context, _ string) (*deepseek.SessionStats, error) {
+	return &deepseek.SessionStats{Success: true}, nil
+}
+
 func TestTestAccount_BatchModeOnlyCreatesSession(t *testing.T) {
 	t.Setenv("DS2API_CONFIG_JSON", `{"accounts":[{"email":"batch@example.com","password":"pwd","token":""}]}`)
 	store := config.LoadStore()
@@ -54,7 +77,7 @@ func TestTestAccount_BatchModeOnlyCreatesSession(t *testing.T) {
 		t.Fatalf("expected success=true, got %#v", result)
 	}
 	msg, _ := result["message"].(string)
-	if !strings.Contains(msg, "仅会话创建") {
+	if !strings.Contains(msg, "Token 刷新成功") {
 		t.Fatalf("expected session-only success message, got %q", msg)
 	}
 	if ds.loginCalls != 1 || ds.createSessionCalls != 1 {
@@ -74,3 +97,38 @@ func TestTestAccount_BatchModeOnlyCreatesSession(t *testing.T) {
 		t.Fatalf("expected test status ok, got %q", updated.TestStatus)
 	}
 }
+
+func TestDeleteAllSessions_RetryWithReloginOnDeleteFailure(t *testing.T) {
+	t.Setenv("DS2API_CONFIG_JSON", `{"accounts":[{"email":"batch@example.com","password":"pwd","token":"expired-token"}]}`)
+	store := config.LoadStore()
+	ds := &testingDSMock{deleteAllSessionsError: errors.New("token expired"), deleteAllSessionsErrorOnce: true}
+	h := &Handler{Store: store, DS: ds}
+
+	req := httptest.NewRequest(http.MethodPost, "/delete-all", bytes.NewBufferString(`{"identifier":"batch@example.com"}`))
+	rec := httptest.NewRecorder()
+	h.deleteAllSessions(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("expected status 200, got %d", rec.Code)
+	}
+	var resp map[string]any
+	if err := json.Unmarshal(rec.Body.Bytes(), &resp); err != nil {
+		t.Fatalf("unmarshal response: %v", err)
+	}
+	if ok, _ := resp["success"].(bool); !ok {
+		t.Fatalf("expected success response, got %#v", resp)
+	}
+	if ds.loginCalls != 2 {
+		t.Fatalf("expected initial login plus relogin, got %d", ds.loginCalls)
+	}
+	if ds.deleteAllSessionsCalls != 2 {
+		t.Fatalf("expected delete called twice, got %d", ds.deleteAllSessionsCalls)
+	}
+	updated, ok := store.FindAccount("batch@example.com")
+	if !ok {
+		t.Fatal("expected account")
+	}
+	if updated.Token != "new-token" {
+		t.Fatalf("expected refreshed token persisted, got %q", updated.Token)
+	}
+}

internal/admin/handler_config_import.go

@@ -43,6 +43,7 @@ func (h *Handler) configImport(w http.ResponseWriter, r *http.Request) {
 		writeJSON(w, http.StatusBadRequest, map[string]any{"detail": err.Error()})
 		return
 	}
+	incoming.ClearAccountTokens()
 
 	importedKeys, importedAccounts := 0, 0
 	err = h.Store.Update(func(c *config.Config) error {
@@ -180,6 +181,7 @@ func (h *Handler) configImport(w http.ResponseWriter, r *http.Request) {
 
 func (h *Handler) computeSyncHash() string {
 	snap := h.Store.Snapshot().Clone()
+	snap.ClearAccountTokens()
 	snap.VercelSyncHash = ""
 	snap.VercelSyncTime = 0
 	b, _ := json.Marshal(snap)

internal/admin/handler_config_write.go

@@ -50,9 +50,6 @@ func (h *Handler) updateConfig(w http.ResponseWriter, r *http.Request) {
 					if strings.TrimSpace(acc.Password) == "" {
 						acc.Password = prev.Password
 					}
-					if strings.TrimSpace(acc.Token) == "" {
-						acc.Token = prev.Token
-					}
 				}
 				seen[key] = struct{}{}
 				accounts = append(accounts, acc)

internal/admin/handler_settings_parse.go

@@ -7,15 +7,30 @@ import (
 	"ds2api/internal/config"
 )
 
-func parseSettingsUpdateRequest(req map[string]any) (*config.AdminConfig, *config.RuntimeConfig, *config.ToolcallConfig, *config.ResponsesConfig, *config.EmbeddingsConfig, map[string]string, map[string]string, error) {
+func boolFrom(v any) bool {
+	if v == nil {
+		return false
+	}
+	switch x := v.(type) {
+	case bool:
+		return x
+	case string:
+		return strings.ToLower(strings.TrimSpace(x)) == "true"
+	default:
+		return false
+	}
+}
+
+func parseSettingsUpdateRequest(req map[string]any) (*config.AdminConfig, *config.RuntimeConfig, *config.ToolcallConfig, *config.ResponsesConfig, *config.EmbeddingsConfig, *config.AutoDeleteConfig, map[string]string, map[string]string, error) {
 	var (
-		adminCfg    *config.AdminConfig
-		runtimeCfg  *config.RuntimeConfig
-		toolcallCfg *config.ToolcallConfig
-		respCfg     *config.ResponsesConfig
-		embCfg      *config.EmbeddingsConfig
-		claudeMap   map[string]string
-		aliasMap    map[string]string
+		adminCfg       *config.AdminConfig
+		runtimeCfg     *config.RuntimeConfig
+		toolcallCfg    *config.ToolcallConfig
+		respCfg        *config.ResponsesConfig
+		embCfg         *config.EmbeddingsConfig
+		autoDeleteCfg  *config.AutoDeleteConfig
+		claudeMap      map[string]string
+		aliasMap       map[string]string
 	)
 
 	if raw, ok := req["admin"].(map[string]any); ok {
@@ -23,7 +38,7 @@ func parseSettingsUpdateRequest(req map[string]any) (*config.AdminConfig, *confi
 		if v, exists := raw["jwt_expire_hours"]; exists {
 			n := intFrom(v)
 			if n < 1 || n > 720 {
-				return nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("admin.jwt_expire_hours must be between 1 and 720")
+				return nil, nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("admin.jwt_expire_hours must be between 1 and 720")
 			}
 			cfg.JWTExpireHours = n
 		}
@@ -35,26 +50,26 @@ func parseSettingsUpdateRequest(req map[string]any) (*config.AdminConfig, *confi
 		if v, exists := raw["account_max_inflight"]; exists {
 			n := intFrom(v)
 			if n < 1 || n > 256 {
-				return nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("runtime.account_max_inflight must be between 1 and 256")
+				return nil, nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("runtime.account_max_inflight must be between 1 and 256")
 			}
 			cfg.AccountMaxInflight = n
 		}
 		if v, exists := raw["account_max_queue"]; exists {
 			n := intFrom(v)
 			if n < 1 || n > 200000 {
-				return nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("runtime.account_max_queue must be between 1 and 200000")
+				return nil, nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("runtime.account_max_queue must be between 1 and 200000")
 			}
 			cfg.AccountMaxQueue = n
 		}
 		if v, exists := raw["global_max_inflight"]; exists {
 			n := intFrom(v)
 			if n < 1 || n > 200000 {
-				return nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("runtime.global_max_inflight must be between 1 and 200000")
+				return nil, nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("runtime.global_max_inflight must be between 1 and 200000")
 			}
 			cfg.GlobalMaxInflight = n
 		}
 		if cfg.AccountMaxInflight > 0 && cfg.GlobalMaxInflight > 0 && cfg.GlobalMaxInflight < cfg.AccountMaxInflight {
-			return nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("runtime.global_max_inflight must be >= runtime.account_max_inflight")
+			return nil, nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("runtime.global_max_inflight must be >= runtime.account_max_inflight")
 		}
 		runtimeCfg = cfg
 	}
@@ -67,7 +82,7 @@ func parseSettingsUpdateRequest(req map[string]any) (*config.AdminConfig, *confi
 			case "feature_match", "off":
 				cfg.Mode = mode
 			default:
-				return nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("toolcall.mode must be feature_match or off")
+				return nil, nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("toolcall.mode must be feature_match or off")
 			}
 		}
 		if v, exists := raw["early_emit_confidence"]; exists {
@@ -76,7 +91,7 @@ func parseSettingsUpdateRequest(req map[string]any) (*config.AdminConfig, *confi
 			case "high", "low", "off":
 				cfg.EarlyEmitConfidence = level
 			default:
-				return nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("toolcall.early_emit_confidence must be high, low or off")
+				return nil, nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("toolcall.early_emit_confidence must be high, low or off")
 			}
 		}
 		toolcallCfg = cfg
@@ -87,7 +102,7 @@ func parseSettingsUpdateRequest(req map[string]any) (*config.AdminConfig, *confi
 		if v, exists := raw["store_ttl_seconds"]; exists {
 			n := intFrom(v)
 			if n < 30 || n > 86400 {
-				return nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("responses.store_ttl_seconds must be between 30 and 86400")
+				return nil, nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("responses.store_ttl_seconds must be between 30 and 86400")
 			}
 			cfg.StoreTTLSeconds = n
 		}
@@ -98,9 +113,6 @@ func parseSettingsUpdateRequest(req map[string]any) (*config.AdminConfig, *confi
 		cfg := &config.EmbeddingsConfig{}
 		if v, exists := raw["provider"]; exists {
 			p := strings.TrimSpace(fmt.Sprintf("%v", v))
-			if p == "" {
-				return nil, nil, nil, nil, nil, nil, nil, fmt.Errorf("embeddings.provider cannot be empty")
-			}
 			cfg.Provider = p
 		}
 		embCfg = cfg
@@ -130,5 +142,13 @@ func parseSettingsUpdateRequest(req map[string]any) (*config.AdminConfig, *confi
 		}
 	}
 
-	return adminCfg, runtimeCfg, toolcallCfg, respCfg, embCfg, claudeMap, aliasMap, nil
+	if raw, ok := req["auto_delete"].(map[string]any); ok {
+		cfg := &config.AutoDeleteConfig{}
+		if v, exists := raw["sessions"]; exists {
+			cfg.Sessions = boolFrom(v)
+		}
+		autoDeleteCfg = cfg
+	}
+
+	return adminCfg, runtimeCfg, toolcallCfg, respCfg, embCfg, autoDeleteCfg, claudeMap, aliasMap, nil
 }

internal/admin/handler_settings_read.go

@@ -28,6 +28,7 @@ func (h *Handler) getSettings(w http.ResponseWriter, _ *http.Request) {
 		"toolcall":          snap.Toolcall,
 		"responses":         snap.Responses,
 		"embeddings":        snap.Embeddings,
+		"auto_delete":       snap.AutoDelete,
 		"claude_mapping":    settingsClaudeMapping(snap),
 		"model_aliases":     snap.ModelAliases,
 		"env_backed":        h.Store.IsEnvBacked(),

internal/admin/handler_settings_write.go

@@ -17,7 +17,7 @@ func (h *Handler) updateSettings(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
-	adminCfg, runtimeCfg, toolcallCfg, responsesCfg, embeddingsCfg, claudeMap, aliasMap, err := parseSettingsUpdateRequest(req)
+	adminCfg, runtimeCfg, toolcallCfg, responsesCfg, embeddingsCfg, autoDeleteCfg, claudeMap, aliasMap, err := parseSettingsUpdateRequest(req)
 	if err != nil {
 		writeJSON(w, http.StatusBadRequest, map[string]any{"detail": err.Error()})
 		return
@@ -60,6 +60,9 @@ func (h *Handler) updateSettings(w http.ResponseWriter, r *http.Request) {
 		if embeddingsCfg != nil && strings.TrimSpace(embeddingsCfg.Provider) != "" {
 			c.Embeddings.Provider = strings.TrimSpace(embeddingsCfg.Provider)
 		}
+		if autoDeleteCfg != nil {
+			c.AutoDelete.Sessions = autoDeleteCfg.Sessions
+		}
 		if claudeMap != nil {
 			c.ClaudeMapping = claudeMap
 			c.ClaudeModelMap = nil

internal/admin/handler_version.go

@@ -0,0 +1,75 @@
+package admin
+
+import (
+	"encoding/json"
+	"net/http"
+	"strings"
+	"time"
+
+	"ds2api/internal/version"
+)
+
+const latestReleaseAPI = "https://api.github.com/repos/CJackHwang/ds2api/releases/latest"
+
+type latestReleasePayload struct {
+	TagName     string `json:"tag_name"`
+	HTMLURL     string `json:"html_url"`
+	PublishedAt string `json:"published_at"`
+}
+
+func (h *Handler) getVersion(w http.ResponseWriter, _ *http.Request) {
+	current, source := version.Current()
+	resp := map[string]any{
+		"success":         true,
+		"current_version": current,
+		"current_tag":     version.Tag(current),
+		"source":          source,
+		"checked_at":      time.Now().UTC().Format(time.RFC3339),
+	}
+
+	req, err := http.NewRequest(http.MethodGet, latestReleaseAPI, nil)
+	if err != nil {
+		resp["check_error"] = err.Error()
+		writeJSON(w, http.StatusOK, resp)
+		return
+	}
+	req.Header.Set("Accept", "application/vnd.github+json")
+	req.Header.Set("User-Agent", "ds2api-version-check")
+
+	client := &http.Client{Timeout: 4 * time.Second}
+	r, err := client.Do(req)
+	if err != nil {
+		resp["check_error"] = err.Error()
+		writeJSON(w, http.StatusOK, resp)
+		return
+	}
+	defer r.Body.Close()
+	if r.StatusCode < 200 || r.StatusCode >= 300 {
+		resp["check_error"] = "github api status: " + r.Status
+		writeJSON(w, http.StatusOK, resp)
+		return
+	}
+
+	var data latestReleasePayload
+	if err := json.NewDecoder(r.Body).Decode(&data); err != nil {
+		resp["check_error"] = err.Error()
+		writeJSON(w, http.StatusOK, resp)
+		return
+	}
+
+	latest := strings.TrimSpace(data.TagName)
+	if latest == "" {
+		resp["check_error"] = "missing latest tag"
+		writeJSON(w, http.StatusOK, resp)
+		return
+	}
+	latestVersion := strings.TrimPrefix(latest, "v")
+
+	resp["latest_tag"] = latest
+	resp["latest_version"] = latestVersion
+	resp["release_url"] = data.HTMLURL
+	resp["published_at"] = data.PublishedAt
+	resp["has_update"] = version.Compare(current, latestVersion) < 0
+
+	writeJSON(w, http.StatusOK, resp)
+}

internal/admin/helpers.go

@@ -65,7 +65,6 @@ func toAccount(m map[string]any) config.Account {
 		Email:    email,
 		Mobile:   mobile,
 		Password: fieldString(m, "password"),
-		Token:    fieldString(m, "token"),
 	}
 }
 

internal/admin/helpers_edge_test.go

@@ -188,8 +188,8 @@ func TestToAccountAllFields(t *testing.T) {
 	if acc.Password != "secret" {
 		t.Fatalf("unexpected password: %q", acc.Password)
 	}
-	if acc.Token != "tok123" {
-		t.Fatalf("unexpected token: %q", acc.Token)
+	if acc.Token != "" {
+		t.Fatalf("expected token to be ignored, got %q", acc.Token)
 	}
 }
 

internal/admin/token_runtime_http_test.go

@@ -0,0 +1,109 @@
+package admin
+
+import (
+	"bytes"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+
+	"github.com/go-chi/chi/v5"
+
+	"ds2api/internal/account"
+	"ds2api/internal/config"
+)
+
+func newHTTPAdminHarness(t *testing.T, rawConfig string, ds DeepSeekCaller) http.Handler {
+	t.Helper()
+	t.Setenv("DS2API_CONFIG_JSON", rawConfig)
+	t.Setenv("CONFIG_JSON", "")
+	store := config.LoadStore()
+	h := &Handler{
+		Store: store,
+		Pool:  account.NewPool(store),
+		DS:    ds,
+	}
+	r := chi.NewRouter()
+	RegisterRoutes(r, h)
+	return r
+}
+
+func adminReq(method, path string, body []byte) *http.Request {
+	req := httptest.NewRequest(method, path, bytes.NewReader(body))
+	req.Header.Set("Authorization", "Bearer admin")
+	req.Header.Set("Content-Type", "application/json")
+	return req
+}
+
+func TestConfigImportIgnoresTokenFieldInPayload(t *testing.T) {
+	ds := &testingDSMock{}
+	router := newHTTPAdminHarness(t, `{"accounts":[]}`, ds)
+
+	payload := []byte(`{
+		"mode":"replace",
+		"config":{
+			"accounts":[{"email":"u@example.com","password":"pwd","token":"expired-token"}]
+		}
+	}`)
+	rec := httptest.NewRecorder()
+	router.ServeHTTP(rec, adminReq(http.MethodPost, "/config/import", payload))
+	if rec.Code != http.StatusOK {
+		t.Fatalf("import status=%d body=%s", rec.Code, rec.Body.String())
+	}
+
+	readRec := httptest.NewRecorder()
+	router.ServeHTTP(readRec, adminReq(http.MethodGet, "/config", nil))
+	if readRec.Code != http.StatusOK {
+		t.Fatalf("get config status=%d body=%s", readRec.Code, readRec.Body.String())
+	}
+	var data map[string]any
+	if err := json.Unmarshal(readRec.Body.Bytes(), &data); err != nil {
+		t.Fatalf("decode config response: %v", err)
+	}
+	accounts, _ := data["accounts"].([]any)
+	if len(accounts) != 1 {
+		t.Fatalf("expected one account, got %d", len(accounts))
+	}
+	accountMap, _ := accounts[0].(map[string]any)
+	if hasToken, _ := accountMap["has_token"].(bool); hasToken {
+		t.Fatalf("expected imported token to be ignored, account=%#v", accountMap)
+	}
+}
+
+func TestAccountTestRefreshesRuntimeTokenButExportOmitsToken(t *testing.T) {
+	ds := &testingDSMock{}
+	router := newHTTPAdminHarness(t, `{
+		"accounts":[{"email":"batch@example.com","password":"pwd","token":"stale-token"}]
+	}`, ds)
+
+	rec := httptest.NewRecorder()
+	router.ServeHTTP(rec, adminReq(http.MethodPost, "/accounts/test", []byte(`{"identifier":"batch@example.com"}`)))
+	if rec.Code != http.StatusOK {
+		t.Fatalf("test account status=%d body=%s", rec.Code, rec.Body.String())
+	}
+	var testResp map[string]any
+	if err := json.Unmarshal(rec.Body.Bytes(), &testResp); err != nil {
+		t.Fatalf("decode test response: %v", err)
+	}
+	if ok, _ := testResp["success"].(bool); !ok {
+		t.Fatalf("expected test success, got %#v", testResp)
+	}
+	if ds.loginCalls < 1 {
+		t.Fatalf("expected login to be called at least once, got %d", ds.loginCalls)
+	}
+
+	exportRec := httptest.NewRecorder()
+	router.ServeHTTP(exportRec, adminReq(http.MethodGet, "/config/export", nil))
+	if exportRec.Code != http.StatusOK {
+		t.Fatalf("export status=%d body=%s", exportRec.Code, exportRec.Body.String())
+	}
+	var exportResp map[string]any
+	if err := json.Unmarshal(exportRec.Body.Bytes(), &exportResp); err != nil {
+		t.Fatalf("decode export response: %v", err)
+	}
+	exportJSON, _ := exportResp["json"].(string)
+	if strings.Contains(exportJSON, `"token"`) {
+		t.Fatalf("expected export json to omit tokens, got %s", exportJSON)
+	}
+}

internal/auth/request_test.go

@@ -58,7 +58,7 @@ func TestDetermineWithXAPIKeyManagedKeyAcquiresAccount(t *testing.T) {
 	if auth.AccountID != "acc@example.com" {
 		t.Fatalf("unexpected account id: %q", auth.AccountID)
 	}
-	if auth.DeepSeekToken != "account-token" {
+	if auth.DeepSeekToken != "fresh-token" {
 		t.Fatalf("unexpected account token: %q", auth.DeepSeekToken)
 	}
 	if auth.CallerID == "" {

internal/compat/go_compat_test.go

@@ -73,22 +73,31 @@ func TestGoCompatToolcallFixtures(t *testing.T) {
 		mustLoadJSON(t, fixturePath, &fixture)
 
 		var expected struct {
-			Calls []util.ParsedToolCall `json:"calls"`
+			Calls             []util.ParsedToolCall `json:"calls"`
+			SawToolCallSyntax bool                  `json:"sawToolCallSyntax"`
+			RejectedByPolicy  bool                  `json:"rejectedByPolicy"`
+			RejectedToolNames []string              `json:"rejectedToolNames"`
 		}
 		mustLoadJSON(t, expectedPath, &expected)
 
-		var got []util.ParsedToolCall
+		var got util.ToolCallParseResult
 		switch strings.ToLower(strings.TrimSpace(fixture.Mode)) {
 		case "standalone":
-			got = util.ParseStandaloneToolCalls(fixture.Text, fixture.ToolNames)
+			got = util.ParseStandaloneToolCallsDetailed(fixture.Text, fixture.ToolNames)
 		default:
-			got = util.ParseToolCalls(fixture.Text, fixture.ToolNames)
+			got = util.ParseToolCallsDetailed(fixture.Text, fixture.ToolNames)
 		}
-		if len(got) == 0 && len(expected.Calls) == 0 {
-			continue
+		if got.Calls == nil {
+			got.Calls = []util.ParsedToolCall{}
 		}
-		if !reflect.DeepEqual(got, expected.Calls) {
-			t.Fatalf("toolcall fixture %s mismatch:\n got=%#v\nwant=%#v", name, got, expected.Calls)
+		if got.RejectedToolNames == nil {
+			got.RejectedToolNames = []string{}
+		}
+		if !reflect.DeepEqual(got.Calls, expected.Calls) ||
+			got.SawToolCallSyntax != expected.SawToolCallSyntax ||
+			got.RejectedByPolicy != expected.RejectedByPolicy ||
+			!reflect.DeepEqual(got.RejectedToolNames, expected.RejectedToolNames) {
+			t.Fatalf("toolcall fixture %s mismatch:\n got=%#v\nwant=%#v", name, got, expected)
 		}
 	}
 }

internal/config/account.go

@@ -1,10 +1,6 @@
 package config
 
-import (
-	"crypto/sha256"
-	"encoding/hex"
-	"strings"
-)
+import "strings"
 
 func (a Account) Identifier() string {
 	if strings.TrimSpace(a.Email) != "" {
@@ -13,12 +9,5 @@ func (a Account) Identifier() string {
 	if mobile := NormalizeMobileForStorage(a.Mobile); mobile != "" {
 		return mobile
 	}
-	// Backward compatibility: old configs may contain token-only accounts.
-	// Use a stable non-sensitive synthetic id so they can still join the pool.
-	token := strings.TrimSpace(a.Token)
-	if token == "" {
-		return ""
-	}
-	sum := sha256.Sum256([]byte(token))
-	return "token:" + hex.EncodeToString(sum[:8])
+	return ""
 }

internal/config/codec.go

@@ -47,6 +47,7 @@ func (c Config) MarshalJSON() ([]byte, error) {
 	if strings.TrimSpace(c.Embeddings.Provider) != "" {
 		m["embeddings"] = c.Embeddings
 	}
+	m["auto_delete"] = c.AutoDelete
 	if c.VercelSyncHash != "" {
 		m["_vercel_sync_hash"] = c.VercelSyncHash
 	}
@@ -108,6 +109,10 @@ func (c *Config) UnmarshalJSON(b []byte) error {
 			if err := json.Unmarshal(v, &c.Embeddings); err != nil {
 				return fmt.Errorf("invalid field %q: %w", k, err)
 			}
+		case "auto_delete":
+			if err := json.Unmarshal(v, &c.AutoDelete); err != nil {
+				return fmt.Errorf("invalid field %q: %w", k, err)
+			}
 		case "_vercel_sync_hash":
 			if err := json.Unmarshal(v, &c.VercelSyncHash); err != nil {
 				return fmt.Errorf("invalid field %q: %w", k, err)
@@ -141,6 +146,7 @@ func (c Config) Clone() Config {
 		Toolcall:         c.Toolcall,
 		Responses:        c.Responses,
 		Embeddings:       c.Embeddings,
+		AutoDelete:       c.AutoDelete,
 		VercelSyncHash:   c.VercelSyncHash,
 		VercelSyncTime:   c.VercelSyncTime,
 		AdditionalFields: map[string]any{},

internal/config/config.go

@@ -12,6 +12,7 @@ type Config struct {
 	Toolcall         ToolcallConfig    `json:"toolcall,omitempty"`
 	Responses        ResponsesConfig   `json:"responses,omitempty"`
 	Embeddings       EmbeddingsConfig  `json:"embeddings,omitempty"`
+	AutoDelete       AutoDeleteConfig  `json:"auto_delete"`
 	VercelSyncHash   string            `json:"_vercel_sync_hash,omitempty"`
 	VercelSyncTime   int64             `json:"_vercel_sync_time,omitempty"`
 	AdditionalFields map[string]any    `json:"-"`
@@ -25,6 +26,32 @@ type Account struct {
 	TestStatus string `json:"test_status,omitempty"`
 }
 
+func (c *Config) ClearAccountTokens() {
+	if c == nil {
+		return
+	}
+	for i := range c.Accounts {
+		c.Accounts[i].Token = ""
+	}
+}
+
+// DropInvalidAccounts removes accounts that cannot be addressed by admin APIs
+// (no email and no normalizable mobile). This prevents legacy token-only
+// records from becoming orphaned empty entries after token stripping.
+func (c *Config) DropInvalidAccounts() {
+	if c == nil || len(c.Accounts) == 0 {
+		return
+	}
+	kept := make([]Account, 0, len(c.Accounts))
+	for _, acc := range c.Accounts {
+		if acc.Identifier() == "" {
+			continue
+		}
+		kept = append(kept, acc)
+	}
+	c.Accounts = kept
+}
+
 type CompatConfig struct {
 	WideInputStrictOutput *bool `json:"wide_input_strict_output,omitempty"`
 }
@@ -53,3 +80,7 @@ type ResponsesConfig struct {
 type EmbeddingsConfig struct {
 	Provider string `json:"provider,omitempty"`
 }
+
+type AutoDeleteConfig struct {
+	Sessions bool `json:"sessions"`
+}

internal/config/config_test.go

@@ -2,25 +2,22 @@ package config
 
 import (
 	"encoding/base64"
-	"strings"
+	"os"
 	"testing"
 )
 
-func TestAccountIdentifierFallsBackToTokenHash(t *testing.T) {
+func TestAccountIdentifierRequiresEmailOrMobile(t *testing.T) {
 	acc := Account{Token: "example-token-value"}
 	id := acc.Identifier()
-	if !strings.HasPrefix(id, "token:") {
-		t.Fatalf("expected token-prefixed identifier, got %q", id)
-	}
-	if len(id) != len("token:")+16 {
-		t.Fatalf("unexpected identifier length: %d (%q)", len(id), id)
+	if id != "" {
+		t.Fatalf("expected empty identifier when only token is present, got %q", id)
 	}
 }
 
-func TestStoreFindAccountWithTokenOnlyIdentifier(t *testing.T) {
+func TestLoadStoreClearsTokensFromConfigInput(t *testing.T) {
 	t.Setenv("DS2API_CONFIG_JSON", `{
 		"keys":["k1"],
-		"accounts":[{"token":"token-only-account"}]
+		"accounts":[{"email":"u@example.com","password":"p","token":"token-only-account"}]
 	}`)
 
 	store := LoadStore()
@@ -28,22 +25,62 @@ func TestStoreFindAccountWithTokenOnlyIdentifier(t *testing.T) {
 	if len(accounts) != 1 {
 		t.Fatalf("expected 1 account, got %d", len(accounts))
 	}
-	id := accounts[0].Identifier()
-	if id == "" {
-		t.Fatalf("expected synthetic identifier for token-only account")
-	}
-	found, ok := store.FindAccount(id)
-	if !ok {
-		t.Fatalf("expected FindAccount to locate token-only account by synthetic id")
-	}
-	if found.Token != "token-only-account" {
-		t.Fatalf("unexpected token value: %q", found.Token)
+	if accounts[0].Token != "" {
+		t.Fatalf("expected token to be cleared after loading, got %q", accounts[0].Token)
 	}
 }
 
-func TestStoreUpdateAccountTokenKeepsOldAndNewIdentifierResolvable(t *testing.T) {
+func TestLoadStoreDropsLegacyTokenOnlyAccounts(t *testing.T) {
 	t.Setenv("DS2API_CONFIG_JSON", `{
-		"accounts":[{"token":"old-token"}]
+		"accounts":[
+			{"token":"legacy-token-only"},
+			{"email":"u@example.com","password":"p","token":"runtime-token"}
+		]
+	}`)
+
+	store := LoadStore()
+	accounts := store.Accounts()
+	if len(accounts) != 1 {
+		t.Fatalf("expected token-only account to be dropped, got %d accounts", len(accounts))
+	}
+	if accounts[0].Identifier() != "u@example.com" {
+		t.Fatalf("unexpected remaining account: %#v", accounts[0])
+	}
+	if accounts[0].Token != "" {
+		t.Fatalf("expected persisted token to be cleared, got %q", accounts[0].Token)
+	}
+}
+
+func TestLoadStorePreservesFileBackedTokensForRuntime(t *testing.T) {
+	tmp, err := os.CreateTemp(t.TempDir(), "config-*.json")
+	if err != nil {
+		t.Fatalf("create temp config: %v", err)
+	}
+	defer tmp.Close()
+
+	if _, err := tmp.WriteString(`{
+		"accounts":[{"email":"u@example.com","password":"p","token":"persisted-token"}]
+	}`); err != nil {
+		t.Fatalf("write temp config: %v", err)
+	}
+
+	t.Setenv("DS2API_CONFIG_JSON", "")
+	t.Setenv("CONFIG_JSON", "")
+	t.Setenv("DS2API_CONFIG_PATH", tmp.Name())
+
+	store := LoadStore()
+	accounts := store.Accounts()
+	if len(accounts) != 1 {
+		t.Fatalf("expected 1 account, got %d", len(accounts))
+	}
+	if accounts[0].Token != "persisted-token" {
+		t.Fatalf("expected file-backed token preserved for runtime use, got %q", accounts[0].Token)
+	}
+}
+
+func TestStoreUpdateAccountTokenKeepsIdentifierResolvable(t *testing.T) {
+	t.Setenv("DS2API_CONFIG_JSON", `{
+		"accounts":[{"email":"user@example.com","password":"p"}]
 	}`)
 
 	store := LoadStore()
@@ -52,23 +89,12 @@ func TestStoreUpdateAccountTokenKeepsOldAndNewIdentifierResolvable(t *testing.T)
 		t.Fatalf("expected 1 account, got %d", len(before))
 	}
 	oldID := before[0].Identifier()
-	if oldID == "" {
-		t.Fatal("expected old identifier")
-	}
 	if err := store.UpdateAccountToken(oldID, "new-token"); err != nil {
 		t.Fatalf("update token failed: %v", err)
 	}
 
-	after := store.Accounts()
-	newID := after[0].Identifier()
-	if newID == "" || newID == oldID {
-		t.Fatalf("expected changed identifier, old=%q new=%q", oldID, newID)
-	}
-	if got, ok := store.FindAccount(newID); !ok || got.Token != "new-token" {
-		t.Fatalf("expected find by new identifier")
-	}
 	if got, ok := store.FindAccount(oldID); !ok || got.Token != "new-token" {
-		t.Fatalf("expected find by old identifier alias")
+		t.Fatalf("expected find by stable account identifier")
 	}
 }
 

internal/config/store.go

@@ -39,6 +39,8 @@ func loadConfig() (Config, bool, error) {
 	}
 	if rawCfg != "" {
 		cfg, err := parseConfigString(rawCfg)
+		cfg.ClearAccountTokens()
+		cfg.DropInvalidAccounts()
 		return cfg, true, err
 	}
 
@@ -55,6 +57,7 @@ func loadConfig() (Config, bool, error) {
 	if err := json.Unmarshal(content, &cfg); err != nil {
 		return Config{}, false, err
 	}
+	cfg.DropInvalidAccounts()
 	if IsVercel() {
 		// Vercel filesystem is ephemeral/read-only for runtime writes; avoid save errors.
 		return cfg, true, nil
@@ -161,7 +164,9 @@ func (s *Store) Save() error {
 		Logger.Info("[save_config] source from env, skip write")
 		return nil
 	}
-	b, err := json.MarshalIndent(s.cfg, "", "  ")
+	persistCfg := s.cfg.Clone()
+	persistCfg.ClearAccountTokens()
+	b, err := json.MarshalIndent(persistCfg, "", "  ")
 	if err != nil {
 		return err
 	}
@@ -173,7 +178,9 @@ func (s *Store) saveLocked() error {
 		Logger.Info("[save_config] source from env, skip write")
 		return nil
 	}
-	b, err := json.MarshalIndent(s.cfg, "", "  ")
+	persistCfg := s.cfg.Clone()
+	persistCfg.ClearAccountTokens()
+	b, err := json.MarshalIndent(persistCfg, "", "  ")
 	if err != nil {
 		return err
 	}
@@ -197,7 +204,9 @@ func (s *Store) SetVercelSync(hash string, ts int64) error {
 func (s *Store) ExportJSONAndBase64() (string, string, error) {
 	s.mu.RLock()
 	defer s.mu.RUnlock()
-	b, err := json.Marshal(s.cfg)
+	exportCfg := s.cfg.Clone()
+	exportCfg.ClearAccountTokens()
+	b, err := json.Marshal(exportCfg)
 	if err != nil {
 		return "", "", err
 	}

internal/config/store_accessors.go

@@ -165,3 +165,9 @@ func (s *Store) RuntimeGlobalMaxInflight(defaultSize int) int {
 	}
 	return defaultSize
 }
+
+func (s *Store) AutoDeleteSessions() bool {
+	s.mu.RLock()
+	defer s.mu.RUnlock()
+	return s.cfg.AutoDelete.Sessions
+}

internal/deepseek/client_auth.go

@@ -62,8 +62,8 @@ func (c *Client) CreateSession(ctx context.Context, a *auth.RequestAuth, maxAtte
 			attempts++
 			continue
 		}
-		code := intFrom(resp["code"])
-		if status == http.StatusOK && code == 0 {
+		code, bizCode, msg, bizMsg := extractResponseStatus(resp)
+		if status == http.StatusOK && code == 0 && bizCode == 0 {
 			data, _ := resp["data"].(map[string]any)
 			bizData, _ := data["biz_data"].(map[string]any)
 			sessionID, _ := bizData["id"].(string)
@@ -71,10 +71,9 @@ func (c *Client) CreateSession(ctx context.Context, a *auth.RequestAuth, maxAtte
 				return sessionID, nil
 			}
 		}
-		msg, _ := resp["msg"].(string)
-		config.Logger.Warn("[create_session] failed", "status", status, "code", code, "msg", msg, "use_config_token", a.UseConfigToken, "account", a.AccountID)
+		config.Logger.Warn("[create_session] failed", "status", status, "code", code, "biz_code", bizCode, "msg", msg, "biz_msg", bizMsg, "use_config_token", a.UseConfigToken, "account", a.AccountID)
 		if a.UseConfigToken {
-			if isTokenInvalid(status, code, msg) && !refreshed {
+			if !refreshed && shouldAttemptRefresh(status, code, bizCode, msg, bizMsg) {
 				if c.Auth.RefreshToken(ctx, a) {
 					refreshed = true
 					continue
@@ -96,6 +95,7 @@ func (c *Client) GetPow(ctx context.Context, a *auth.RequestAuth, maxAttempts in
 		maxAttempts = c.maxRetries
 	}
 	attempts := 0
+	refreshed := false
 	for attempts < maxAttempts {
 		headers := c.authHeaders(a.DeepSeekToken)
 		resp, status, err := c.postJSONWithStatus(ctx, c.regular, DeepSeekCreatePowURL, headers, map[string]any{"target_path": "/api/v0/chat/completion"})
@@ -104,8 +104,8 @@ func (c *Client) GetPow(ctx context.Context, a *auth.RequestAuth, maxAttempts in
 			attempts++
 			continue
 		}
-		code := intFrom(resp["code"])
-		if status == http.StatusOK && code == 0 {
+		code, bizCode, msg, bizMsg := extractResponseStatus(resp)
+		if status == http.StatusOK && code == 0 && bizCode == 0 {
 			data, _ := resp["data"].(map[string]any)
 			bizData, _ := data["biz_data"].(map[string]any)
 			challenge, _ := bizData["challenge"].(map[string]any)
@@ -116,15 +116,16 @@ func (c *Client) GetPow(ctx context.Context, a *auth.RequestAuth, maxAttempts in
 			}
 			return BuildPowHeader(challenge, answer)
 		}
-		msg, _ := resp["msg"].(string)
-		config.Logger.Warn("[get_pow] failed", "status", status, "code", code, "msg", msg, "use_config_token", a.UseConfigToken, "account", a.AccountID)
+		config.Logger.Warn("[get_pow] failed", "status", status, "code", code, "biz_code", bizCode, "msg", msg, "biz_msg", bizMsg, "use_config_token", a.UseConfigToken, "account", a.AccountID)
 		if a.UseConfigToken {
-			if isTokenInvalid(status, code, msg) {
+			if !refreshed && shouldAttemptRefresh(status, code, bizCode, msg, bizMsg) {
 				if c.Auth.RefreshToken(ctx, a) {
+					refreshed = true
 					continue
 				}
 			}
 			if c.Auth.SwitchAccount(ctx, a) {
+				refreshed = false
 				attempts++
 				continue
 			}
@@ -143,15 +144,75 @@ func (c *Client) authHeaders(token string) map[string]string {
 	return headers
 }
 
-func isTokenInvalid(status int, code int, msg string) bool {
-	msg = strings.ToLower(msg)
+func isTokenInvalid(status int, code int, bizCode int, msg string, bizMsg string) bool {
+	msg = strings.ToLower(strings.TrimSpace(msg) + " " + strings.TrimSpace(bizMsg))
 	if status == http.StatusUnauthorized || status == http.StatusForbidden {
 		return true
 	}
-	if code == 40001 || code == 40002 || code == 40003 {
+	if code == 40001 || code == 40002 || code == 40003 || bizCode == 40001 || bizCode == 40002 || bizCode == 40003 {
 		return true
 	}
-	return strings.Contains(msg, "token") || strings.Contains(msg, "unauthorized")
+	return strings.Contains(msg, "token") ||
+		strings.Contains(msg, "unauthorized") ||
+		strings.Contains(msg, "expired") ||
+		strings.Contains(msg, "not login") ||
+		strings.Contains(msg, "login required") ||
+		strings.Contains(msg, "invalid jwt")
+}
+
+func shouldAttemptRefresh(status int, code int, bizCode int, msg string, bizMsg string) bool {
+	if isTokenInvalid(status, code, bizCode, msg, bizMsg) {
+		return true
+	}
+	// Some DeepSeek failures come back as HTTP 200/code=0 but with non-zero biz_code.
+	// Only attempt refresh when these biz failures still look auth-related.
+	return status == http.StatusOK &&
+		code == 0 &&
+		bizCode != 0 &&
+		isAuthIndicativeBizFailure(msg, bizMsg)
+}
+
+func isAuthIndicativeBizFailure(msg string, bizMsg string) bool {
+	combined := strings.ToLower(strings.TrimSpace(msg) + " " + strings.TrimSpace(bizMsg))
+	authKeywords := []string{
+		"auth",
+		"authorization",
+		"credential",
+		"expired",
+		"invalid jwt",
+		"jwt",
+		"login",
+		"not login",
+		"session expired",
+		"token",
+		"unauthorized",
+		"登录",
+		"未登录",
+		"认证",
+		"凭证",
+		"会话过期",
+		"令牌",
+	}
+	for _, keyword := range authKeywords {
+		if strings.Contains(combined, keyword) {
+			return true
+		}
+	}
+	return false
+}
+
+func extractResponseStatus(resp map[string]any) (code int, bizCode int, msg string, bizMsg string) {
+	code = intFrom(resp["code"])
+	msg, _ = resp["msg"].(string)
+	data, _ := resp["data"].(map[string]any)
+	bizCode = intFrom(data["biz_code"])
+	bizMsg, _ = data["biz_msg"].(string)
+	if strings.TrimSpace(bizMsg) == "" {
+		if bizData, ok := data["biz_data"].(map[string]any); ok {
+			bizMsg, _ = bizData["msg"].(string)
+		}
+	}
+	return code, bizCode, msg, bizMsg
 }
 
 func normalizeMobileForLogin(raw string) (mobile string, areaCode any) {

internal/deepseek/client_auth_refresh_test.go

@@ -0,0 +1,27 @@
+package deepseek
+
+import "testing"
+
+func TestShouldAttemptRefreshOnTokenInvalidSignal(t *testing.T) {
+	if !shouldAttemptRefresh(401, 0, 0, "unauthorized", "") {
+		t.Fatal("expected refresh when response indicates invalid token")
+	}
+}
+
+func TestShouldAttemptRefreshOnAuthIndicativeBizCodeFailure(t *testing.T) {
+	if !shouldAttemptRefresh(200, 0, 400123, "", "login expired, token invalid") {
+		t.Fatal("expected refresh on auth-indicative biz_code failure")
+	}
+}
+
+func TestShouldAttemptRefreshFalseOnNonAuthBizCodeFailure(t *testing.T) {
+	if shouldAttemptRefresh(200, 0, 400123, "", "session create failed: quota reached") {
+		t.Fatal("did not expect refresh on non-auth biz_code failure")
+	}
+}
+
+func TestShouldAttemptRefreshFalseOnGenericServerError(t *testing.T) {
+	if shouldAttemptRefresh(500, 500, 0, "internal error", "") {
+		t.Fatal("did not expect refresh on generic server error")
+	}
+}

internal/deepseek/client_http_json.go

@@ -62,3 +62,51 @@ func (c *Client) postJSONWithStatus(ctx context.Context, doer trans.Doer, url st
 	}
 	return out, resp.StatusCode, nil
 }
+
+func (c *Client) getJSON(ctx context.Context, doer trans.Doer, url string, headers map[string]string) (map[string]any, error) {
+	body, status, err := c.getJSONWithStatus(ctx, doer, url, headers)
+	if err != nil {
+		return nil, err
+	}
+	if status == 0 {
+		return nil, errors.New("request failed")
+	}
+	return body, nil
+}
+
+func (c *Client) getJSONWithStatus(ctx context.Context, doer trans.Doer, url string, headers map[string]string) (map[string]any, int, error) {
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
+	if err != nil {
+		return nil, 0, err
+	}
+	for k, v := range headers {
+		req.Header.Set(k, v)
+	}
+	resp, err := doer.Do(req)
+	if err != nil {
+		config.Logger.Warn("[deepseek] fingerprint GET request failed, fallback to std transport", "url", url, "error", err)
+		req2, reqErr := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
+		if reqErr != nil {
+			return nil, 0, err
+		}
+		for k, v := range headers {
+			req2.Header.Set(k, v)
+		}
+		resp, err = c.fallback.Do(req2)
+		if err != nil {
+			return nil, 0, err
+		}
+	}
+	defer resp.Body.Close()
+	payloadBytes, err := readResponseBody(resp)
+	if err != nil {
+		return nil, resp.StatusCode, err
+	}
+	out := map[string]any{}
+	if len(payloadBytes) > 0 {
+		if err := json.Unmarshal(payloadBytes, &out); err != nil {
+			config.Logger.Warn("[deepseek] json parse failed", "url", url, "status", resp.StatusCode, "content_encoding", resp.Header.Get("Content-Encoding"), "preview", preview(payloadBytes))
+		}
+	}
+	return out, resp.StatusCode, nil
+}

internal/deepseek/client_session.go

@@ -0,0 +1,256 @@
+package deepseek
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"net/http"
+	"net/url"
+	"strings"
+
+	"ds2api/internal/auth"
+	"ds2api/internal/config"
+)
+
+// SessionInfo 会话信息
+type SessionInfo struct {
+	ID        string  `json:"id"`
+	Title     string  `json:"title"`
+	TitleType string  `json:"title_type"`
+	Pinned    bool    `json:"pinned"`
+	UpdatedAt float64 `json:"updated_at"`
+}
+
+// SessionStats 会话统计结果
+type SessionStats struct {
+	AccountID      string // 账号标识 (email 或 mobile)
+	FirstPageCount int    // 第一页会话数量（当 HasMore 为 true 时，真实总数可能更大）
+	PinnedCount    int    // 置顶会话数量
+	HasMore        bool   // 是否还有更多页
+	Success        bool   // 请求是否成功
+	ErrorMessage   string // 错误信息
+}
+
+// GetSessionCount 获取单个账号的会话数量
+func (c *Client) GetSessionCount(ctx context.Context, a *auth.RequestAuth, maxAttempts int) (*SessionStats, error) {
+	if maxAttempts <= 0 {
+		maxAttempts = c.maxRetries
+	}
+
+	stats := &SessionStats{
+		AccountID: a.AccountID,
+	}
+
+	attempts := 0
+	refreshed := false
+
+	for attempts < maxAttempts {
+		headers := c.authHeaders(a.DeepSeekToken)
+
+		// 构建请求 URL
+		reqURL := DeepSeekFetchSessionURL + "?lte_cursor.pinned=false"
+
+		resp, status, err := c.getJSONWithStatus(ctx, c.regular, reqURL, headers)
+		if err != nil {
+			config.Logger.Warn("[get_session_count] request error", "error", err, "account", a.AccountID)
+			attempts++
+			continue
+		}
+
+		code, bizCode, msg, bizMsg := extractResponseStatus(resp)
+		if status == http.StatusOK && code == 0 && bizCode == 0 {
+			data, _ := resp["data"].(map[string]any)
+			bizData, _ := data["biz_data"].(map[string]any)
+			chatSessions, _ := bizData["chat_sessions"].([]any)
+			hasMore, _ := bizData["has_more"].(bool)
+
+			stats.FirstPageCount = len(chatSessions)
+			stats.HasMore = hasMore
+			stats.Success = true
+
+			// 统计置顶会话数量
+			for _, session := range chatSessions {
+				if s, ok := session.(map[string]any); ok {
+					if pinned, ok := s["pinned"].(bool); ok && pinned {
+						stats.PinnedCount++
+					}
+				}
+			}
+
+			return stats, nil
+		}
+
+		stats.ErrorMessage = fmt.Sprintf("status=%d, code=%d, msg=%s", status, code, msg)
+		config.Logger.Warn("[get_session_count] failed", "status", status, "code", code, "biz_code", bizCode, "msg", msg, "biz_msg", bizMsg, "account", a.AccountID)
+
+		if a.UseConfigToken {
+			if isTokenInvalid(status, code, bizCode, msg, bizMsg) && !refreshed {
+				if c.Auth.RefreshToken(ctx, a) {
+					refreshed = true
+					continue
+				}
+			}
+			if c.Auth.SwitchAccount(ctx, a) {
+				refreshed = false
+				attempts++
+				continue
+			}
+		}
+		attempts++
+	}
+
+	stats.Success = false
+	stats.ErrorMessage = "get session count failed after retries"
+	return stats, errors.New(stats.ErrorMessage)
+}
+
+// GetSessionCountForToken 直接使用 token 获取会话数量（直通模式）
+func (c *Client) GetSessionCountForToken(ctx context.Context, token string) (*SessionStats, error) {
+	headers := c.authHeaders(token)
+	reqURL := DeepSeekFetchSessionURL + "?lte_cursor.pinned=false"
+
+	resp, status, err := c.getJSONWithStatus(ctx, c.regular, reqURL, headers)
+	if err != nil {
+		return nil, err
+	}
+
+	code, bizCode, msg, bizMsg := extractResponseStatus(resp)
+	if status != http.StatusOK || code != 0 || bizCode != 0 {
+		if strings.TrimSpace(bizMsg) != "" {
+			msg = bizMsg
+		}
+		return nil, fmt.Errorf("request failed: status=%d, code=%d, msg=%s", status, code, msg)
+	}
+
+	data, _ := resp["data"].(map[string]any)
+	bizData, _ := data["biz_data"].(map[string]any)
+	chatSessions, _ := bizData["chat_sessions"].([]any)
+	hasMore, _ := bizData["has_more"].(bool)
+
+	stats := &SessionStats{
+		FirstPageCount: len(chatSessions),
+		HasMore:        hasMore,
+		Success:        true,
+	}
+
+	// 统计置顶会话数量
+	for _, session := range chatSessions {
+		if s, ok := session.(map[string]any); ok {
+			if pinned, ok := s["pinned"].(bool); ok && pinned {
+				stats.PinnedCount++
+			}
+		}
+	}
+
+	return stats, nil
+}
+
+// GetSessionCountAll 获取所有账号的会话数量统计
+func (c *Client) GetSessionCountAll(ctx context.Context) []*SessionStats {
+	accounts := c.Store.Accounts()
+	results := make([]*SessionStats, 0, len(accounts))
+
+	for _, acc := range accounts {
+		token := acc.Token
+		accountID := acc.Email
+		if accountID == "" {
+			accountID = acc.Mobile
+		}
+
+		// 如果没有 token，尝试登录获取
+		if token == "" {
+			var err error
+			token, err = c.Login(ctx, acc)
+			if err != nil {
+				results = append(results, &SessionStats{
+					AccountID:    accountID,
+					Success:      false,
+					ErrorMessage: fmt.Sprintf("login failed: %v", err),
+				})
+				continue
+			}
+		}
+
+		stats, err := c.GetSessionCountForToken(ctx, token)
+		if err != nil {
+			results = append(results, &SessionStats{
+				AccountID:    accountID,
+				Success:      false,
+				ErrorMessage: err.Error(),
+			})
+			continue
+		}
+
+		stats.AccountID = accountID
+		results = append(results, stats)
+	}
+
+	return results
+}
+
+// FetchSessionPage 获取会话列表（支持分页）
+func (c *Client) FetchSessionPage(ctx context.Context, a *auth.RequestAuth, cursor string) ([]SessionInfo, bool, error) {
+	headers := c.authHeaders(a.DeepSeekToken)
+
+	// 构建请求 URL
+	params := url.Values{}
+	params.Set("lte_cursor.pinned", "false")
+	if cursor != "" {
+		params.Set("lte_cursor", cursor)
+	}
+	reqURL := DeepSeekFetchSessionURL + "?" + params.Encode()
+
+	resp, status, err := c.getJSONWithStatus(ctx, c.regular, reqURL, headers)
+	if err != nil {
+		return nil, false, err
+	}
+
+	code := intFrom(resp["code"])
+	if status != http.StatusOK || code != 0 {
+		msg, _ := resp["msg"].(string)
+		return nil, false, fmt.Errorf("request failed: status=%d, code=%d, msg=%s", status, code, msg)
+	}
+
+	data, _ := resp["data"].(map[string]any)
+	bizData, _ := data["biz_data"].(map[string]any)
+	chatSessions, _ := bizData["chat_sessions"].([]any)
+	hasMore, _ := bizData["has_more"].(bool)
+
+	sessions := make([]SessionInfo, 0, len(chatSessions))
+	for _, s := range chatSessions {
+		if m, ok := s.(map[string]any); ok {
+			session := SessionInfo{
+				ID:        stringFromMap(m, "id"),
+				Title:     stringFromMap(m, "title"),
+				TitleType: stringFromMap(m, "title_type"),
+				Pinned:    boolFromMap(m, "pinned"),
+				UpdatedAt: floatFromMap(m, "updated_at"),
+			}
+			sessions = append(sessions, session)
+		}
+	}
+
+	return sessions, hasMore, nil
+}
+
+// 辅助函数
+func stringFromMap(m map[string]any, key string) string {
+	if v, ok := m[key].(string); ok {
+		return v
+	}
+	return ""
+}
+
+func boolFromMap(m map[string]any, key string) bool {
+	if v, ok := m[key].(bool); ok {
+		return v
+	}
+	return false
+}
+
+func floatFromMap(m map[string]any, key string) float64 {
+	if v, ok := m[key].(float64); ok {
+		return v
+	}
+	return 0
+}

internal/deepseek/client_session_delete.go

@@ -0,0 +1,155 @@
+package deepseek
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"net/http"
+
+	"ds2api/internal/auth"
+	"ds2api/internal/config"
+)
+
+// DeleteSessionResult 删除会话结果
+type DeleteSessionResult struct {
+	SessionID    string // 会话 ID
+	Success      bool   // 是否成功
+	ErrorMessage string // 错误信息
+}
+
+// DeleteSession 删除单个会话
+func (c *Client) DeleteSession(ctx context.Context, a *auth.RequestAuth, sessionID string, maxAttempts int) (*DeleteSessionResult, error) {
+	if maxAttempts <= 0 {
+		maxAttempts = c.maxRetries
+	}
+
+	result := &DeleteSessionResult{
+		SessionID: sessionID,
+	}
+
+	if sessionID == "" {
+		result.ErrorMessage = "session_id is required"
+		return result, errors.New(result.ErrorMessage)
+	}
+
+	attempts := 0
+	refreshed := false
+
+	for attempts < maxAttempts {
+		headers := c.authHeaders(a.DeepSeekToken)
+
+		payload := map[string]any{
+			"chat_session_id": sessionID,
+		}
+
+		resp, status, err := c.postJSONWithStatus(ctx, c.regular, DeepSeekDeleteSessionURL, headers, payload)
+		if err != nil {
+			config.Logger.Warn("[delete_session] request error", "error", err, "session_id", sessionID)
+			attempts++
+			continue
+		}
+
+		code, bizCode, msg, bizMsg := extractResponseStatus(resp)
+		if status == http.StatusOK && code == 0 && bizCode == 0 {
+			result.Success = true
+			return result, nil
+		}
+
+		result.ErrorMessage = fmt.Sprintf("status=%d, code=%d, msg=%s", status, code, msg)
+		config.Logger.Warn("[delete_session] failed", "status", status, "code", code, "biz_code", bizCode, "msg", msg, "biz_msg", bizMsg, "session_id", sessionID)
+
+		if a.UseConfigToken {
+			if isTokenInvalid(status, code, bizCode, msg, bizMsg) && !refreshed {
+				if c.Auth.RefreshToken(ctx, a) {
+					refreshed = true
+					continue
+				}
+			}
+			if c.Auth.SwitchAccount(ctx, a) {
+				refreshed = false
+				attempts++
+				continue
+			}
+		}
+		attempts++
+	}
+
+	result.Success = false
+	result.ErrorMessage = "delete session failed after retries"
+	return result, errors.New(result.ErrorMessage)
+}
+
+// DeleteSessionForToken 直接使用 token 删除会话（直通模式）
+func (c *Client) DeleteSessionForToken(ctx context.Context, token string, sessionID string) (*DeleteSessionResult, error) {
+	result := &DeleteSessionResult{
+		SessionID: sessionID,
+	}
+
+	if sessionID == "" {
+		result.ErrorMessage = "session_id is required"
+		return result, errors.New(result.ErrorMessage)
+	}
+
+	headers := c.authHeaders(token)
+	payload := map[string]any{
+		"chat_session_id": sessionID,
+	}
+
+	resp, status, err := c.postJSONWithStatus(ctx, c.regular, DeepSeekDeleteSessionURL, headers, payload)
+	if err != nil {
+		result.ErrorMessage = err.Error()
+		return result, err
+	}
+
+	code := intFrom(resp["code"])
+	if status != http.StatusOK || code != 0 {
+		msg, _ := resp["msg"].(string)
+		result.ErrorMessage = fmt.Sprintf("request failed: status=%d, code=%d, msg=%s", status, code, msg)
+		return result, errors.New(result.ErrorMessage)
+	}
+
+	result.Success = true
+	return result, nil
+}
+
+// DeleteAllSessions 删除所有会话（谨慎使用）
+func (c *Client) DeleteAllSessions(ctx context.Context, a *auth.RequestAuth) error {
+	headers := c.authHeaders(a.DeepSeekToken)
+	payload := map[string]any{}
+
+	resp, status, err := c.postJSONWithStatus(ctx, c.regular, DeepSeekDeleteAllSessionsURL, headers, payload)
+	if err != nil {
+		config.Logger.Warn("[delete_all_sessions] request error", "error", err)
+		return err
+	}
+
+	code := intFrom(resp["code"])
+	if status != http.StatusOK || code != 0 {
+		msg, _ := resp["msg"].(string)
+		config.Logger.Warn("[delete_all_sessions] failed", "status", status, "code", code, "msg", msg)
+		return fmt.Errorf("request failed: status=%d, code=%d, msg=%s", status, code, msg)
+	}
+
+	return nil
+}
+
+// DeleteAllSessionsForToken 直接使用 token 删除所有会话（直通模式）
+func (c *Client) DeleteAllSessionsForToken(ctx context.Context, token string) error {
+	headers := c.authHeaders(token)
+	payload := map[string]any{}
+
+	resp, status, err := c.postJSONWithStatus(ctx, c.regular, DeepSeekDeleteAllSessionsURL, headers, payload)
+	if err != nil {
+		config.Logger.Warn("[delete_all_sessions_for_token] request error", "error", err)
+		return err
+	}
+
+	code := intFrom(resp["code"])
+	if status != http.StatusOK || code != 0 {
+		msg, _ := resp["msg"].(string)
+		config.Logger.Warn("[delete_all_sessions_for_token] failed", "status", status, "code", code, "msg", msg)
+		return fmt.Errorf("request failed: status=%d, code=%d, msg=%s", status, code, msg)
+	}
+
+	return nil
+}

internal/deepseek/constants.go

@@ -11,6 +11,9 @@ const (
 	DeepSeekCreateSessionURL = "https://chat.deepseek.com/api/v0/chat_session/create"
 	DeepSeekCreatePowURL     = "https://chat.deepseek.com/api/v0/chat/create_pow_challenge"
 	DeepSeekCompletionURL    = "https://chat.deepseek.com/api/v0/chat/completion"
+	DeepSeekFetchSessionURL  = "https://chat.deepseek.com/api/v0/chat_session/fetch_page"
+	DeepSeekDeleteSessionURL     = "https://chat.deepseek.com/api/v0/chat_session/delete"
+	DeepSeekDeleteAllSessionsURL = "https://chat.deepseek.com/api/v0/chat_session/delete_all"
 )
 
 var defaultBaseHeaders = map[string]string{

internal/format/openai/render_chat.go

@@ -8,15 +8,15 @@ import (
 )
 
 func BuildChatCompletion(completionID, model, finalPrompt, finalThinking, finalText string, toolNames []string) map[string]any {
-	detected := util.ParseStandaloneToolCalls(finalText, toolNames)
+	detected := util.ParseStandaloneToolCallsDetailed(finalText, toolNames)
 	finishReason := "stop"
 	messageObj := map[string]any{"role": "assistant", "content": finalText}
 	if strings.TrimSpace(finalThinking) != "" {
 		messageObj["reasoning_content"] = finalThinking
 	}
-	if len(detected) > 0 {
+	if len(detected.Calls) > 0 {
 		finishReason = "tool_calls"
-		messageObj["tool_calls"] = util.FormatOpenAIToolCalls(detected)
+		messageObj["tool_calls"] = util.FormatOpenAIToolCalls(detected.Calls)
 		messageObj["content"] = nil
 	}
 

internal/format/openai/render_responses.go

@@ -13,12 +13,12 @@ import (
 func BuildResponseObject(responseID, model, finalPrompt, finalThinking, finalText string, toolNames []string) map[string]any {
 	// Strict mode: only standalone, structured tool-call payloads are treated
 	// as executable tool calls.
-	detected := util.ParseStandaloneToolCalls(finalText, toolNames)
+	detected := util.ParseStandaloneToolCallsDetailed(finalText, toolNames)
 	exposedOutputText := finalText
 	output := make([]any, 0, 2)
-	if len(detected) > 0 {
+	if len(detected.Calls) > 0 {
 		exposedOutputText = ""
-		output = append(output, toResponsesFunctionCallItems(detected)...)
+		output = append(output, toResponsesFunctionCallItems(detected.Calls)...)
 	} else {
 		content := make([]any, 0, 2)
 		if finalThinking != "" {

internal/format/openai/render_stream_events.go

@@ -71,6 +71,19 @@ func BuildResponsesTextDeltaPayload(responseID, itemID string, outputIndex, cont
 	}
 }
 
+
+func BuildResponsesTextDonePayload(responseID, itemID string, outputIndex, contentIndex int, text string) map[string]any {
+	return map[string]any{
+		"type":          "response.output_text.done",
+		"id":            responseID,
+		"response_id":   responseID,
+		"item_id":       itemID,
+		"output_index":  outputIndex,
+		"content_index": contentIndex,
+		"text":          text,
+	}
+}
+
 func BuildResponsesReasoningDeltaPayload(responseID, delta string) map[string]any {
 	return map[string]any{
 		"type":        "response.reasoning.delta",

internal/format/openai/render_test.go

@@ -45,7 +45,7 @@ func TestBuildResponseObjectToolCallsFollowChatShape(t *testing.T) {
 	}
 }
 
-func TestBuildResponseObjectTreatsMixedProseToolPayloadAsText(t *testing.T) {
+func TestBuildResponseObjectPromotesMixedProseToolPayloadToFunctionCall(t *testing.T) {
 	obj := BuildResponseObject(
 		"resp_test",
 		"gpt-4o",
@@ -56,20 +56,20 @@ func TestBuildResponseObjectTreatsMixedProseToolPayloadAsText(t *testing.T) {
 	)
 
 	outputText, _ := obj["output_text"].(string)
-	if outputText == "" {
-		t.Fatalf("expected output_text preserved for mixed prose payload")
+	if outputText != "" {
+		t.Fatalf("expected output_text hidden for mixed prose tool payload, got %q", outputText)
 	}
 	output, _ := obj["output"].([]any)
 	if len(output) != 1 {
-		t.Fatalf("expected one message output item, got %#v", obj["output"])
+		t.Fatalf("expected one function_call output item, got %#v", obj["output"])
 	}
 	first, _ := output[0].(map[string]any)
-	if first["type"] != "message" {
-		t.Fatalf("expected message output type, got %#v", first["type"])
+	if first["type"] != "function_call" {
+		t.Fatalf("expected function_call output type, got %#v", first["type"])
 	}
 }
 
-func TestBuildResponseObjectFencedToolPayloadRemainsText(t *testing.T) {
+func TestBuildResponseObjectPromotesFencedToolPayloadToFunctionCall(t *testing.T) {
 	obj := BuildResponseObject(
 		"resp_test",
 		"gpt-4o",
@@ -80,16 +80,16 @@ func TestBuildResponseObjectFencedToolPayloadRemainsText(t *testing.T) {
 	)
 
 	outputText, _ := obj["output_text"].(string)
-	if outputText == "" {
-		t.Fatalf("expected output_text preserved for fenced example")
+	if outputText != "" {
+		t.Fatalf("expected output_text hidden for fenced tool payload, got %q", outputText)
 	}
 	output, _ := obj["output"].([]any)
 	if len(output) != 1 {
-		t.Fatalf("expected one message output item, got %#v", obj["output"])
+		t.Fatalf("expected one function_call output item, got %#v", obj["output"])
 	}
 	first, _ := output[0].(map[string]any)
-	if first["type"] != "message" {
-		t.Fatalf("expected message output type, got %#v", first["type"])
+	if first["type"] != "function_call" {
+		t.Fatalf("expected function_call output type, got %#v", first["type"])
 	}
 }
 

internal/js/chat-stream/toolcall_policy.js

@@ -60,6 +60,9 @@ function formatIncrementalToolCallDeltas(deltas, idStore) {
     if (typeof d.arguments === 'string' && d.arguments !== '') {
       fn.arguments = d.arguments;
     }
+    if (Object.keys(fn).length === 0) {
+      continue;
+    }
     if (Object.keys(fn).length > 0) {
       item.function = fn;
     }

internal/js/chat-stream/vercel_stream.js

@@ -1,33 +1,22 @@
 'use strict';
 
 const {
-  extractToolNames,
   createToolSieveState,
   processToolSieveChunk,
   flushToolSieve,
   parseStandaloneToolCalls,
   formatOpenAIStreamToolCalls,
 } = require('../helpers/stream-tool-sieve');
-const {
-  BASE_HEADERS,
-} = require('../shared/deepseek-constants');
-
-const {
-  writeOpenAIError,
-} = require('./error_shape');
-const {
-  parseChunkForContent,
-  isCitation,
-} = require('./sse_parse');
-const {
-  buildUsage,
-} = require('./token_usage');
+const { BASE_HEADERS } = require('../shared/deepseek-constants');
+const { writeOpenAIError } = require('./error_shape');
+const { parseChunkForContent, isCitation } = require('./sse_parse');
+const { buildUsage } = require('./token_usage');
 const {
   resolveToolcallPolicy,
+  formatIncrementalToolCallDeltas,
+  filterIncrementalToolCallDeltasByAllowed,
 } = require('./toolcall_policy');
-const {
-  createChatCompletionEmitter,
-} = require('./stream_emitter');
+const { createChatCompletionEmitter } = require('./stream_emitter');
 const {
   asString,
   isAbortError,
@@ -57,6 +46,7 @@ async function handleVercelStream(req, res, rawBody, payload) {
   const searchEnabled = toBool(prep.body.search_enabled);
   const toolPolicy = resolveToolcallPolicy(prep.body, payload.tools);
   const toolNames = toolPolicy.toolNames;
+  const emitEarlyToolDeltas = toolPolicy.emitEarlyToolDeltas;
 
   if (!model || !leaseID || !deepseekToken || !powHeader || !completionPayload) {
     writeOpenAIError(res, 500, 'invalid vercel prepare response');
@@ -132,6 +122,7 @@ async function handleVercelStream(req, res, rawBody, payload) {
     const toolSieveState = createToolSieveState();
     let toolCallsEmitted = false;
     const streamToolCallIDs = new Map();
+    const streamToolNames = new Map();
     const decoder = new TextDecoder();
     reader = completionRes.body.getReader();
     let buffered = '';
@@ -255,6 +246,18 @@ async function handleVercelStream(req, res, rawBody, payload) {
               }
               const events = processToolSieveChunk(toolSieveState, p.text, toolNames);
               for (const evt of events) {
+                if (evt.type === 'tool_call_deltas') {
+                  if (!emitEarlyToolDeltas) {
+                    continue;
+                  }
+                  const filtered = filterIncrementalToolCallDeltasByAllowed(evt.deltas, toolNames, streamToolNames);
+                  const formatted = formatIncrementalToolCallDeltas(filtered, streamToolCallIDs);
+                  if (formatted.length > 0) {
+                    toolCallsEmitted = true;
+                    sendDeltaFrame({ tool_calls: formatted });
+                  }
+                  continue;
+                }
                 if (evt.type === 'tool_calls') {
                   toolCallsEmitted = true;
                   sendDeltaFrame({ tool_calls: formatOpenAIStreamToolCalls(evt.calls, streamToolCallIDs) });

internal/js/helpers/stream-tool-sieve/parse.js

@@ -2,28 +2,33 @@
 
 const {
   toStringSafe,
-  looksLikeToolExampleContext,
 } = require('./state');
 const {
-  stripFencedCodeBlocks,
   buildToolCallCandidates,
   parseToolCallsPayload,
+  parseMarkupToolCalls,
+  parseTextKVToolCalls,
 } = require('./parse_payload');
 
+const TOOL_NAME_LOOSE_PATTERN = /[^a-z0-9]+/g;
+
 function extractToolNames(tools) {
   if (!Array.isArray(tools) || tools.length === 0) {
     return [];
   }
   const out = [];
+  const seen = new Set();
   for (const t of tools) {
     if (!t || typeof t !== 'object') {
       continue;
     }
     const fn = t.function && typeof t.function === 'object' ? t.function : t;
     const name = toStringSafe(fn.name);
-    // Keep parity with Go injectToolPrompt: object tools without name still
-    // enter tool mode via fallback name "unknown".
-    out.push(name || 'unknown');
+    if (!name || seen.has(name)) {
+      continue;
+    }
+    seen.add(name);
+    out.push(name);
   }
   return out;
 }
@@ -34,26 +39,36 @@ function parseToolCalls(text, toolNames) {
 
 function parseToolCallsDetailed(text, toolNames) {
   const result = emptyParseResult();
-  if (!toStringSafe(text)) {
+  const normalized = toStringSafe(text);
+  if (!normalized) {
     return result;
   }
-  const sanitized = stripFencedCodeBlocks(text);
-  if (!toStringSafe(sanitized)) {
-    return result;
-  }
-  result.sawToolCallSyntax = sanitized.toLowerCase().includes('tool_calls');
+  result.sawToolCallSyntax = looksLikeToolCallSyntax(normalized);
 
-  const candidates = buildToolCallCandidates(sanitized);
+  const candidates = buildToolCallCandidates(normalized);
   let parsed = [];
   for (const c of candidates) {
     parsed = parseToolCallsPayload(c);
+    if (parsed.length === 0) {
+      parsed = parseMarkupToolCalls(c);
+    }
+    if (parsed.length === 0) {
+      parsed = parseTextKVToolCalls(c);
+    }
     if (parsed.length > 0) {
       result.sawToolCallSyntax = true;
       break;
     }
   }
   if (parsed.length === 0) {
-    return result;
+    parsed = parseMarkupToolCalls(normalized);
+    if (parsed.length === 0) {
+      parsed = parseTextKVToolCalls(normalized);
+      if (parsed.length === 0) {
+        return result;
+      }
+    }
+    result.sawToolCallSyntax = true;
   }
 
   const filtered = filterToolCallsDetailed(parsed, toolNames);
@@ -73,17 +88,29 @@ function parseStandaloneToolCallsDetailed(text, toolNames) {
   if (!trimmed) {
     return result;
   }
-  if (looksLikeToolExampleContext(trimmed)) {
-    return result;
+  result.sawToolCallSyntax = looksLikeToolCallSyntax(trimmed);
+  const candidates = buildToolCallCandidates(trimmed);
+  let parsed = [];
+  for (const c of candidates) {
+    parsed = parseToolCallsPayload(c);
+    if (parsed.length === 0) {
+      parsed = parseMarkupToolCalls(c);
+    }
+    if (parsed.length === 0) {
+      parsed = parseTextKVToolCalls(c);
+    }
+    if (parsed.length > 0) {
+      break;
+    }
   }
-  result.sawToolCallSyntax = trimmed.toLowerCase().includes('tool_calls');
-  if (!trimmed.startsWith('{') && !trimmed.startsWith('[')) {
-    return result;
-  }
-
-  const parsed = parseToolCallsPayload(trimmed);
   if (parsed.length === 0) {
-    return result;
+    parsed = parseMarkupToolCalls(trimmed);
+    if (parsed.length === 0) {
+      parsed = parseTextKVToolCalls(trimmed);
+      if (parsed.length === 0) {
+        return result;
+      }
+    }
   }
 
   result.sawToolCallSyntax = true;
@@ -146,7 +173,7 @@ function filterToolCallsDetailed(parsed, toolNames) {
     if (allowed.has(tc.name)) {
       matchedName = tc.name;
     } else {
-      matchedName = allowedCanonical.get(tc.name.toLowerCase()) || '';
+      matchedName = resolveAllowedToolName(tc.name, allowed, allowedCanonical);
     }
     if (!matchedName) {
       if (!seenRejected.has(tc.name)) {
@@ -163,6 +190,46 @@ function filterToolCallsDetailed(parsed, toolNames) {
   return { calls, rejectedToolNames: rejected };
 }
 
+function resolveAllowedToolName(name, allowed, allowedCanonical) {
+  const normalizedName = toStringSafe(name).trim();
+  if (!normalizedName) {
+    return '';
+  }
+  if (allowed.has(normalizedName)) {
+    return normalizedName;
+  }
+  const lower = normalizedName.toLowerCase();
+  if (allowedCanonical.has(lower)) {
+    return allowedCanonical.get(lower);
+  }
+  const idx = lower.lastIndexOf('.');
+  if (idx >= 0 && idx < lower.length - 1) {
+    const tail = lower.slice(idx + 1);
+    if (allowedCanonical.has(tail)) {
+      return allowedCanonical.get(tail);
+    }
+  }
+  const loose = lower.replace(TOOL_NAME_LOOSE_PATTERN, '');
+  if (!loose) {
+    return '';
+  }
+  for (const [candidateLower, canonical] of allowedCanonical.entries()) {
+    if (candidateLower.replace(TOOL_NAME_LOOSE_PATTERN, '') === loose) {
+      return canonical;
+    }
+  }
+  return '';
+}
+
+function looksLikeToolCallSyntax(text) {
+  const lower = toStringSafe(text).toLowerCase();
+  return lower.includes('tool_calls')
+    || lower.includes('<tool_call')
+    || lower.includes('<function_call')
+    || lower.includes('<invoke')
+    || lower.includes('function.name:');
+}
+
 module.exports = {
   extractToolNames,
   parseToolCalls,

internal/js/helpers/stream-tool-sieve/parse_payload.js

@@ -1,6 +1,24 @@
 'use strict';
 
 const TOOL_CALL_PATTERN = /\{\s*["']tool_calls["']\s*:\s*\[(.*?)\]\s*\}/s;
+const TOOL_CALL_MARKUP_BLOCK_PATTERN = /<(?:[a-z0-9_:-]+:)?(tool_call|function_call|invoke)\b([^>]*)>([\s\S]*?)<\/(?:[a-z0-9_:-]+:)?\1>/gi;
+const TOOL_CALL_MARKUP_SELFCLOSE_PATTERN = /<(?:[a-z0-9_:-]+:)?invoke\b([^>]*)\/>/gi;
+const TOOL_CALL_MARKUP_KV_PATTERN = /<(?:[a-z0-9_:-]+:)?([a-z0-9_.-]+)\b[^>]*>([\s\S]*?)<\/(?:[a-z0-9_:-]+:)?\1>/gi;
+const TOOL_CALL_MARKUP_ATTR_PATTERN = /(name|function|tool)\s*=\s*"([^"]+)"/i;
+const TOOL_CALL_MARKUP_NAME_PATTERNS = [
+  /<(?:[a-z0-9_:-]+:)?name\b[^>]*>([\s\S]*?)<\/(?:[a-z0-9_:-]+:)?name>/i,
+  /<(?:[a-z0-9_:-]+:)?function\b[^>]*>([\s\S]*?)<\/(?:[a-z0-9_:-]+:)?function>/i,
+];
+const TOOL_CALL_MARKUP_ARGS_PATTERNS = [
+  /<(?:[a-z0-9_:-]+:)?input\b[^>]*>([\s\S]*?)<\/(?:[a-z0-9_:-]+:)?input>/i,
+  /<(?:[a-z0-9_:-]+:)?arguments\b[^>]*>([\s\S]*?)<\/(?:[a-z0-9_:-]+:)?arguments>/i,
+  /<(?:[a-z0-9_:-]+:)?argument\b[^>]*>([\s\S]*?)<\/(?:[a-z0-9_:-]+:)?argument>/i,
+  /<(?:[a-z0-9_:-]+:)?parameters\b[^>]*>([\s\S]*?)<\/(?:[a-z0-9_:-]+:)?parameters>/i,
+  /<(?:[a-z0-9_:-]+:)?parameter\b[^>]*>([\s\S]*?)<\/(?:[a-z0-9_:-]+:)?parameter>/i,
+  /<(?:[a-z0-9_:-]+:)?args\b[^>]*>([\s\S]*?)<\/(?:[a-z0-9_:-]+:)?args>/i,
+  /<(?:[a-z0-9_:-]+:)?params\b[^>]*>([\s\S]*?)<\/(?:[a-z0-9_:-]+:)?params>/i,
+];
+const TEXT_KV_NAME_PATTERN = /function\.name:\s*([a-zA-Z0-9_.-]+)/gi;
 
 const {
   toStringSafe,
@@ -103,6 +121,153 @@ function parseToolCallsPayload(payload) {
   return one ? [one] : [];
 }
 
+function parseMarkupToolCalls(text) {
+  const raw = toStringSafe(text).trim();
+  if (!raw) {
+    return [];
+  }
+  const out = [];
+  for (const m of raw.matchAll(TOOL_CALL_MARKUP_BLOCK_PATTERN)) {
+    const parsed = parseMarkupSingleToolCall(toStringSafe(m[2]).trim(), toStringSafe(m[3]).trim());
+    if (parsed) {
+      out.push(parsed);
+    }
+  }
+  for (const m of raw.matchAll(TOOL_CALL_MARKUP_SELFCLOSE_PATTERN)) {
+    const parsed = parseMarkupSingleToolCall(toStringSafe(m[1]).trim(), '');
+    if (parsed) {
+      out.push(parsed);
+    }
+  }
+  return out;
+}
+
+function parseTextKVToolCalls(text) {
+  const raw = toStringSafe(text);
+  if (!raw) {
+    return [];
+  }
+  const out = [];
+  const matches = [...raw.matchAll(TEXT_KV_NAME_PATTERN)];
+  if (matches.length === 0) {
+    return out;
+  }
+  for (let i = 0; i < matches.length; i += 1) {
+    const match = matches[i];
+    const name = toStringSafe(match[1]).trim();
+    if (!name) {
+      continue;
+    }
+    const nameEnd = match.index + toStringSafe(match[0]).length;
+    const searchEnd = i + 1 < matches.length ? matches[i + 1].index : raw.length;
+    const searchArea = raw.slice(nameEnd, searchEnd);
+    const argIdx = searchArea.indexOf('function.arguments:');
+    if (argIdx < 0) {
+      continue;
+    }
+    const argStart = nameEnd + argIdx + 'function.arguments:'.length;
+    const bracePos = raw.slice(argStart, searchEnd).indexOf('{');
+    if (bracePos < 0) {
+      continue;
+    }
+    const objStart = argStart + bracePos;
+    const obj = extractJSONObjectFrom(raw, objStart);
+    if (!obj.ok) {
+      continue;
+    }
+    out.push({
+      name,
+      input: parseToolCallInput(raw.slice(objStart, obj.end)),
+    });
+  }
+  return out;
+}
+
+function parseMarkupSingleToolCall(attrs, inner) {
+  const embedded = parseToolCallsPayload(inner);
+  if (embedded.length > 0) {
+    return embedded[0];
+  }
+  let name = '';
+  const attrMatch = attrs.match(TOOL_CALL_MARKUP_ATTR_PATTERN);
+  if (attrMatch && attrMatch[2]) {
+    name = toStringSafe(attrMatch[2]).trim();
+  }
+  if (!name) {
+    name = stripTagText(findMarkupTagValue(inner, TOOL_CALL_MARKUP_NAME_PATTERNS));
+  }
+  if (!name) {
+    return null;
+  }
+
+  let input = {};
+  const argsRaw = findMarkupTagValue(inner, TOOL_CALL_MARKUP_ARGS_PATTERNS);
+  if (argsRaw) {
+    input = parseMarkupInput(argsRaw);
+  } else {
+    const kv = parseMarkupKVObject(inner);
+    if (Object.keys(kv).length > 0) {
+      input = kv;
+    }
+  }
+  return { name, input };
+}
+
+function parseMarkupInput(raw) {
+  const s = toStringSafe(raw).trim();
+  if (!s) {
+    return {};
+  }
+  const parsed = parseToolCallInput(s);
+  if (parsed && typeof parsed === 'object' && !Array.isArray(parsed) && Object.keys(parsed).length > 0) {
+    return parsed;
+  }
+  const kv = parseMarkupKVObject(s);
+  if (Object.keys(kv).length > 0) {
+    return kv;
+  }
+  return { _raw: stripTagText(s) };
+}
+
+function parseMarkupKVObject(text) {
+  const raw = toStringSafe(text).trim();
+  if (!raw) {
+    return {};
+  }
+  const out = {};
+  for (const m of raw.matchAll(TOOL_CALL_MARKUP_KV_PATTERN)) {
+    const key = toStringSafe(m[1]).trim();
+    if (!key) {
+      continue;
+    }
+    const valueRaw = stripTagText(m[2]);
+    if (!valueRaw) {
+      continue;
+    }
+    try {
+      out[key] = JSON.parse(valueRaw);
+    } catch (_err) {
+      out[key] = valueRaw;
+    }
+  }
+  return out;
+}
+
+function stripTagText(text) {
+  return toStringSafe(text).replace(/<[^>]+>/g, ' ').trim();
+}
+
+function findMarkupTagValue(text, patterns) {
+  const source = toStringSafe(text);
+  for (const p of patterns) {
+    const m = source.match(p);
+    if (m && m[1]) {
+      return toStringSafe(m[1]);
+    }
+  }
+  return '';
+}
+
 function parseToolCallList(v) {
   if (!Array.isArray(v)) {
     return [];
@@ -193,4 +358,6 @@ module.exports = {
   stripFencedCodeBlocks,
   buildToolCallCandidates,
   parseToolCallsPayload,
+  parseMarkupToolCalls,
+  parseTextKVToolCalls,
 };

internal/js/helpers/stream-tool-sieve/sieve.js

@@ -21,22 +21,14 @@ function processToolSieveChunk(state, chunk, toolNames) {
   }
   const events = [];
 
-  if (Array.isArray(state.pendingToolCalls) && state.pendingToolCalls.length > 0) {
-    const pending = state.pending || '';
-    if (pending.trim() !== '') {
-      const content = (state.pendingToolRaw || '') + pending;
-      state.pending = '';
-      state.pendingToolRaw = '';
-      state.pendingToolCalls = [];
-      noteText(state, content);
-      events.push({ type: 'text', text: content });
-    } else {
-      return events;
-    }
-  }
-
   // eslint-disable-next-line no-constant-condition
   while (true) {
+    if (Array.isArray(state.pendingToolCalls) && state.pendingToolCalls.length > 0) {
+      events.push({ type: 'tool_calls', calls: state.pendingToolCalls });
+      state.pendingToolRaw = '';
+      state.pendingToolCalls = [];
+      continue;
+    }
     if (state.capturing) {
       if (state.pending) {
         state.capture += state.pending;
@@ -54,6 +46,9 @@ function processToolSieveChunk(state, chunk, toolNames) {
       if (Array.isArray(consumed.calls) && consumed.calls.length > 0) {
         state.pendingToolRaw = captured;
         state.pendingToolCalls = consumed.calls;
+        if (consumed.suffix) {
+          state.pending = consumed.suffix + state.pending;
+        }
         continue;
       }
       if (consumed.prefix) {
@@ -173,19 +168,34 @@ function findToolSegmentStart(s) {
     return -1;
   }
   const lower = s.toLowerCase();
+  const keywords = ['tool_calls', 'function.name:', '[tool_call_history]'];
   let offset = 0;
   // eslint-disable-next-line no-constant-condition
   while (true) {
-    const keyIdx = lower.indexOf('tool_calls', offset);
-    if (keyIdx < 0) {
+    let bestKeyIdx = -1;
+    let matchedKeyword = '';
+
+    for (const kw of keywords) {
+      const idx = lower.indexOf(kw, offset);
+      if (idx >= 0) {
+        if (bestKeyIdx < 0 || idx < bestKeyIdx) {
+          bestKeyIdx = idx;
+          matchedKeyword = kw;
+        }
+      }
+    }
+
+    if (bestKeyIdx < 0) {
       return -1;
     }
+
+    const keyIdx = bestKeyIdx;
     const start = s.slice(0, keyIdx).lastIndexOf('{');
     const candidateStart = start >= 0 ? start : keyIdx;
     if (!insideCodeFence(s.slice(0, candidateStart))) {
       return candidateStart;
     }
-    offset = keyIdx + 'tool_calls'.length;
+    offset = keyIdx + matchedKeyword.length;
   }
 }
 
@@ -195,20 +205,28 @@ function consumeToolCapture(state, toolNames) {
     return { ready: false, prefix: '', calls: [], suffix: '' };
   }
   const lower = captured.toLowerCase();
-  const keyIdx = lower.indexOf('tool_calls');
+  
+  let keyIdx = -1;
+  const keywords = ['tool_calls', 'function.name:', '[tool_call_history]'];
+  for (const kw of keywords) {
+    const idx = lower.indexOf(kw);
+    if (idx >= 0 && (keyIdx < 0 || idx < keyIdx)) {
+      keyIdx = idx;
+    }
+  }
+  
   if (keyIdx < 0) {
     return { ready: false, prefix: '', calls: [], suffix: '' };
   }
   const start = captured.slice(0, keyIdx).lastIndexOf('{');
-  if (start < 0) {
-    return { ready: false, prefix: '', calls: [], suffix: '' };
-  }
-  const obj = extractJSONObjectFrom(captured, start);
+  const actualStart = start >= 0 ? start : keyIdx;
+  
+  const obj = extractJSONObjectFrom(captured, actualStart);
   if (!obj.ok) {
     return { ready: false, prefix: '', calls: [], suffix: '' };
   }
 
-  const prefixPart = captured.slice(0, start);
+  const prefixPart = captured.slice(0, actualStart);
   const suffixPart = captured.slice(obj.end);
 
   if (insideCodeFence((state.recentTextTail || '') + prefixPart)) {
@@ -220,16 +238,7 @@ function consumeToolCapture(state, toolNames) {
     };
   }
 
-  if ((state.recentTextTail || '').trim() !== '' || prefixPart.trim() !== '' || suffixPart.trim() !== '') {
-    return {
-      ready: true,
-      prefix: captured,
-      calls: [],
-      suffix: '',
-    };
-  }
-
-  const parsed = parseStandaloneToolCallsDetailed(captured.slice(start, obj.end), toolNames);
+  const parsed = parseStandaloneToolCallsDetailed(captured.slice(actualStart, obj.end), toolNames);
   if (!Array.isArray(parsed.calls) || parsed.calls.length === 0) {
     if (parsed.sawToolCallSyntax && parsed.rejectedByPolicy) {
       return {

internal/prompt/messages.go

@@ -51,6 +51,9 @@ func MessagesPrepare(messages []map[string]any) string {
 }
 
 func NormalizeContent(v any) string {
+	if v == nil {
+		return ""
+	}
 	switch x := v.(type) {
 	case string:
 		return x
@@ -64,11 +67,11 @@ func NormalizeContent(v any) string {
 			typeStr, _ := m["type"].(string)
 			typeStr = strings.ToLower(strings.TrimSpace(typeStr))
 			if typeStr == "text" || typeStr == "output_text" || typeStr == "input_text" {
-				if txt, ok := m["text"].(string); ok {
+				if txt, ok := m["text"].(string); ok && txt != "" {
 					parts = append(parts, txt)
 					continue
 				}
-				if txt, ok := m["content"].(string); ok {
+				if txt, ok := m["content"].(string); ok && txt != "" {
 					parts = append(parts, txt)
 				}
 			}

internal/prompt/messages_test.go

@@ -0,0 +1,32 @@
+package prompt
+
+import "testing"
+
+func TestNormalizeContentNilReturnsEmpty(t *testing.T) {
+	if got := NormalizeContent(nil); got != "" {
+		t.Fatalf("expected empty string for nil content, got %q", got)
+	}
+}
+
+func TestMessagesPrepareNilContentNoNullLiteral(t *testing.T) {
+	messages := []map[string]any{
+		{"role": "assistant", "content": nil},
+		{"role": "user", "content": "ok"},
+	}
+	got := MessagesPrepare(messages)
+	if got == "" {
+		t.Fatalf("expected non-empty output")
+	}
+	if got == "null" {
+		t.Fatalf("expected no null literal output, got %q", got)
+	}
+}
+
+func TestNormalizeContentArrayFallsBackToContentWhenTextEmpty(t *testing.T) {
+	got := NormalizeContent([]any{
+		map[string]any{"type": "text", "text": "", "content": "from-content"},
+	})
+	if got != "from-content" {
+		t.Fatalf("expected fallback to content when text is empty, got %q", got)
+	}
+}

internal/util/toolcalls_candidates.go

@@ -20,7 +20,7 @@ func buildToolCallCandidates(text string) []string {
 		}
 	}
 
-	// best-effort extraction around "tool_calls" key in mixed text payloads.
+	// best-effort extraction around tool call keywords in mixed text payloads.
 	candidates = append(candidates, extractToolCallObjects(trimmed)...)
 
 	// best-effort object slice: from first '{' to last '}'
@@ -57,25 +57,65 @@ func extractToolCallObjects(text string) []string {
 	lower := strings.ToLower(text)
 	out := []string{}
 	offset := 0
+	keywords := []string{"tool_calls", "function.name:", "[tool_call_history]"}
 	for {
-		idx := strings.Index(lower[offset:], "tool_calls")
-		if idx < 0 {
+		bestIdx := -1
+		matchedKeyword := ""
+		for _, kw := range keywords {
+			idx := strings.Index(lower[offset:], kw)
+			if idx >= 0 {
+				absIdx := offset + idx
+				if bestIdx < 0 || absIdx < bestIdx {
+					bestIdx = absIdx
+					matchedKeyword = kw
+				}
+			}
+		}
+
+		if bestIdx < 0 {
 			break
 		}
-		idx += offset
-		start := strings.LastIndex(text[:idx], "{")
-		for start >= 0 {
+
+		idx := bestIdx
+		// Avoid backtracking too far to prevent OOM on malicious or very long strings
+		searchLimit := idx - 2000
+		if searchLimit < offset {
+			searchLimit = offset
+		}
+		
+		start := strings.LastIndex(text[searchLimit:idx], "{")
+		if start >= 0 {
+			start += searchLimit
+		}
+		
+		if start < 0 {
+			offset = idx + len(matchedKeyword)
+			continue
+		}
+
+		foundObj := false
+		for start >= searchLimit {
 			candidate, end, ok := extractJSONObject(text, start)
 			if ok {
 				// Move forward to avoid repeatedly matching the same object.
 				offset = end
 				out = append(out, strings.TrimSpace(candidate))
+				foundObj = true
 				break
 			}
-			start = strings.LastIndex(text[:start], "{")
+			// Try previous '{'
+			if start > searchLimit {
+				prevStart := strings.LastIndex(text[searchLimit:start], "{")
+				if prevStart >= 0 {
+					start = searchLimit + prevStart
+					continue
+				}
+			}
+			break
 		}
-		if start < 0 {
-			offset = idx + len("tool_calls")
+		
+		if !foundObj {
+			offset = idx + len(matchedKeyword)
 		}
 	}
 	return out
@@ -88,7 +128,12 @@ func extractJSONObject(text string, start int) (string, int, bool) {
 	depth := 0
 	quote := byte(0)
 	escaped := false
-	for i := start; i < len(text); i++ {
+	// Limit scan length to avoid OOM on unclosed objects
+	maxLen := start + 50000
+	if maxLen > len(text) {
+		maxLen = len(text)
+	}
+	for i := start; i < maxLen; i++ {
 		ch := text[i]
 		if quote != 0 {
 			if escaped {

internal/util/toolcalls_input_parse.go

@@ -0,0 +1,108 @@
+package util
+
+import (
+	"encoding/json"
+	"strings"
+	"unicode"
+)
+
+func parseToolCallInput(v any) map[string]any {
+	switch x := v.(type) {
+	case nil:
+		return map[string]any{}
+	case map[string]any:
+		return x
+	case string:
+		raw := strings.TrimSpace(x)
+		if raw == "" {
+			return map[string]any{}
+		}
+		var parsed map[string]any
+		if err := json.Unmarshal([]byte(raw), &parsed); err == nil && parsed != nil {
+			repairPathLikeControlChars(parsed)
+			return parsed
+		}
+		// Try to repair invalid backslashes (common in Windows paths output by models)
+		repaired := repairInvalidJSONBackslashes(raw)
+		if repaired != raw {
+			if err := json.Unmarshal([]byte(repaired), &parsed); err == nil && parsed != nil {
+				repairPathLikeControlChars(parsed)
+				return parsed
+			}
+		}
+		// Try to repair loose JSON in string argument as well
+		repairedLoose := RepairLooseJSON(raw)
+		if repairedLoose != raw {
+			if err := json.Unmarshal([]byte(repairedLoose), &parsed); err == nil && parsed != nil {
+				repairPathLikeControlChars(parsed)
+				return parsed
+			}
+		}
+		return map[string]any{"_raw": raw}
+	default:
+		b, err := json.Marshal(x)
+		if err != nil {
+			return map[string]any{}
+		}
+		var parsed map[string]any
+		if err := json.Unmarshal(b, &parsed); err == nil && parsed != nil {
+			return parsed
+		}
+		return map[string]any{}
+	}
+}
+
+func repairPathLikeControlChars(m map[string]any) {
+	for k, v := range m {
+		switch vv := v.(type) {
+		case map[string]any:
+			repairPathLikeControlChars(vv)
+		case []any:
+			for _, item := range vv {
+				if child, ok := item.(map[string]any); ok {
+					repairPathLikeControlChars(child)
+				}
+			}
+		case string:
+			if isPathLikeKey(k) && containsControlRune(vv) {
+				m[k] = escapeControlRunes(vv)
+			}
+		}
+	}
+}
+
+func isPathLikeKey(key string) bool {
+	k := strings.ToLower(strings.TrimSpace(key))
+	return strings.Contains(k, "path") || strings.Contains(k, "file")
+}
+
+func containsControlRune(s string) bool {
+	for _, r := range s {
+		if unicode.IsControl(r) {
+			return true
+		}
+	}
+	return false
+}
+
+func escapeControlRunes(s string) string {
+	var b strings.Builder
+	b.Grow(len(s) + 8)
+	for _, r := range s {
+		switch r {
+		case '\b':
+			b.WriteString(`\b`)
+		case '\f':
+			b.WriteString(`\f`)
+		case '\n':
+			b.WriteString(`\n`)
+		case '\r':
+			b.WriteString(`\r`)
+		case '\t':
+			b.WriteString(`\t`)
+		default:
+			b.WriteRune(r)
+		}
+	}
+	return b.String()
+}

internal/util/toolcalls_json_repair.go

@@ -0,0 +1,79 @@
+package util
+
+import (
+	"regexp"
+	"strings"
+)
+
+func repairInvalidJSONBackslashes(s string) string {
+	if !strings.Contains(s, "\\") {
+		return s
+	}
+	var out strings.Builder
+	out.Grow(len(s) + 10)
+	runes := []rune(s)
+	for i := 0; i < len(runes); i++ {
+		if runes[i] == '\\' {
+			if i+1 < len(runes) {
+				next := runes[i+1]
+				switch next {
+				case '"', '\\', '/', 'b', 'f', 'n', 'r', 't':
+					out.WriteRune('\\')
+					out.WriteRune(next)
+					i++
+					continue
+				case 'u':
+					if i+5 < len(runes) {
+						isHex := true
+						for j := 1; j <= 4; j++ {
+							r := runes[i+1+j]
+							if !((r >= '0' && r <= '9') || (r >= 'a' && r <= 'f') || (r >= 'A' && r <= 'F')) {
+								isHex = false
+								break
+							}
+						}
+						if isHex {
+							out.WriteRune('\\')
+							out.WriteRune('u')
+							for j := 1; j <= 4; j++ {
+								out.WriteRune(runes[i+1+j])
+							}
+							i += 5
+							continue
+						}
+					}
+				}
+			}
+			// Not a valid escape sequence, double it
+			out.WriteString("\\\\")
+		} else {
+			out.WriteRune(runes[i])
+		}
+	}
+	return out.String()
+}
+
+var unquotedKeyPattern = regexp.MustCompile(`([{,]\s*)([a-zA-Z_][a-zA-Z0-9_]*)\s*:`)
+
+// missingArrayBracketsPattern identifies a sequence of two or more JSON objects separated by commas
+// that immediately follow a colon, which indicates a missing array bracket `[` `]`.
+// E.g., "key": {"a": 1}, {"b": 2} -> "key": [{"a": 1}, {"b": 2}]
+// NOTE: The pattern uses (?:[^{}]|\{[^{}]*\})* to support single-level nested {} objects,
+// which handles cases like {"content": "x", "input": {"q": "y"}}
+var missingArrayBracketsPattern = regexp.MustCompile(`(:\s*)(\{(?:[^{}]|\{[^{}]*\})*\}(?:\s*,\s*\{(?:[^{}]|\{[^{}]*\})*\})+)`)
+
+func RepairLooseJSON(s string) string {
+	s = strings.TrimSpace(s)
+	if s == "" {
+		return s
+	}
+	// 1. Replace unquoted keys: {key: -> {"key":
+	s = unquotedKeyPattern.ReplaceAllString(s, `$1"$2":`)
+
+	// 2. Heuristic: Fix missing array brackets for list of objects
+	// e.g., : {obj1}, {obj2} -> : [{obj1}, {obj2}]
+	// This specifically addresses DeepSeek's "list hallucination"
+	s = missingArrayBracketsPattern.ReplaceAllString(s, `$1[$2]`)
+
+	return s
+}

internal/util/toolcalls_markup.go

@@ -0,0 +1,161 @@
+package util
+
+import (
+	"encoding/json"
+	"regexp"
+	"strings"
+)
+
+var toolCallMarkupTagNames = []string{"tool_call", "function_call", "invoke"}
+var toolCallMarkupTagPatternByName = map[string]*regexp.Regexp{
+	"tool_call":     regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?tool_call\b([^>]*)>(.*?)</(?:[a-z0-9_:-]+:)?tool_call>`),
+	"function_call": regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?function_call\b([^>]*)>(.*?)</(?:[a-z0-9_:-]+:)?function_call>`),
+	"invoke":        regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?invoke\b([^>]*)>(.*?)</(?:[a-z0-9_:-]+:)?invoke>`),
+}
+var toolCallMarkupSelfClosingPattern = regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?invoke\b([^>]*)/>`)
+var toolCallMarkupKVPattern = regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?([a-z0-9_\-.]+)\b[^>]*>(.*?)</(?:[a-z0-9_:-]+:)?([a-z0-9_\-.]+)>`)
+var toolCallMarkupAttrPattern = regexp.MustCompile(`(?is)(name|function|tool)\s*=\s*"([^"]+)"`)
+var anyTagPattern = regexp.MustCompile(`(?is)<[^>]+>`)
+var toolCallMarkupNameTagNames = []string{"name", "function"}
+var toolCallMarkupNamePatternByTag = map[string]*regexp.Regexp{
+	"name":     regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?name\b[^>]*>(.*?)</(?:[a-z0-9_:-]+:)?name>`),
+	"function": regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?function\b[^>]*>(.*?)</(?:[a-z0-9_:-]+:)?function>`),
+}
+var toolCallMarkupArgsTagNames = []string{"input", "arguments", "argument", "parameters", "parameter", "args", "params"}
+var toolCallMarkupArgsPatternByTag = map[string]*regexp.Regexp{
+	"input":      regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?input\b[^>]*>(.*?)</(?:[a-z0-9_:-]+:)?input>`),
+	"arguments":  regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?arguments\b[^>]*>(.*?)</(?:[a-z0-9_:-]+:)?arguments>`),
+	"argument":   regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?argument\b[^>]*>(.*?)</(?:[a-z0-9_:-]+:)?argument>`),
+	"parameters": regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?parameters\b[^>]*>(.*?)</(?:[a-z0-9_:-]+:)?parameters>`),
+	"parameter":  regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?parameter\b[^>]*>(.*?)</(?:[a-z0-9_:-]+:)?parameter>`),
+	"args":       regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?args\b[^>]*>(.*?)</(?:[a-z0-9_:-]+:)?args>`),
+	"params":     regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?params\b[^>]*>(.*?)</(?:[a-z0-9_:-]+:)?params>`),
+}
+
+func parseMarkupToolCalls(text string) []ParsedToolCall {
+	trimmed := strings.TrimSpace(text)
+	if trimmed == "" {
+		return nil
+	}
+
+	out := make([]ParsedToolCall, 0)
+	for _, tagName := range toolCallMarkupTagNames {
+		pattern := toolCallMarkupTagPatternByName[tagName]
+		for _, m := range pattern.FindAllStringSubmatch(trimmed, -1) {
+			if len(m) < 3 {
+				continue
+			}
+			attrs := strings.TrimSpace(m[1])
+			inner := strings.TrimSpace(m[2])
+			if parsed := parseMarkupSingleToolCall(attrs, inner); parsed.Name != "" {
+				out = append(out, parsed)
+			}
+		}
+	}
+	for _, m := range toolCallMarkupSelfClosingPattern.FindAllStringSubmatch(trimmed, -1) {
+		if len(m) < 2 {
+			continue
+		}
+		if parsed := parseMarkupSingleToolCall(strings.TrimSpace(m[1]), ""); parsed.Name != "" {
+			out = append(out, parsed)
+		}
+	}
+	if len(out) == 0 {
+		return nil
+	}
+	return out
+}
+
+func parseMarkupSingleToolCall(attrs string, inner string) ParsedToolCall {
+	if parsed := parseToolCallsPayload(inner); len(parsed) > 0 {
+		return parsed[0]
+	}
+
+	name := ""
+	if m := toolCallMarkupAttrPattern.FindStringSubmatch(attrs); len(m) >= 3 {
+		name = strings.TrimSpace(m[2])
+	}
+	if name == "" {
+		name = findMarkupTagValue(inner, toolCallMarkupNameTagNames, toolCallMarkupNamePatternByTag)
+	}
+	if name == "" {
+		return ParsedToolCall{}
+	}
+
+	input := map[string]any{}
+	if argsRaw := findMarkupTagValue(inner, toolCallMarkupArgsTagNames, toolCallMarkupArgsPatternByTag); argsRaw != "" {
+		input = parseMarkupInput(argsRaw)
+	} else if kv := parseMarkupKVObject(inner); len(kv) > 0 {
+		input = kv
+	}
+	return ParsedToolCall{Name: name, Input: input}
+}
+
+func parseMarkupInput(raw string) map[string]any {
+	raw = strings.TrimSpace(raw)
+	if raw == "" {
+		return map[string]any{}
+	}
+	if parsed := parseToolCallInput(raw); len(parsed) > 0 {
+		return parsed
+	}
+	if kv := parseMarkupKVObject(raw); len(kv) > 0 {
+		return kv
+	}
+	return map[string]any{"_raw": stripTagText(raw)}
+}
+
+func parseMarkupKVObject(text string) map[string]any {
+	matches := toolCallMarkupKVPattern.FindAllStringSubmatch(strings.TrimSpace(text), -1)
+	if len(matches) == 0 {
+		return nil
+	}
+	out := map[string]any{}
+	for _, m := range matches {
+		if len(m) < 4 {
+			continue
+		}
+		key := strings.TrimSpace(m[1])
+		endKey := strings.TrimSpace(m[3])
+		if key == "" {
+			continue
+		}
+		if !strings.EqualFold(key, endKey) {
+			continue
+		}
+		value := strings.TrimSpace(stripTagText(m[2]))
+		if value == "" {
+			continue
+		}
+		var jsonValue any
+		if json.Unmarshal([]byte(value), &jsonValue) == nil {
+			out[key] = jsonValue
+			continue
+		}
+		out[key] = value
+	}
+	if len(out) == 0 {
+		return nil
+	}
+	return out
+}
+
+func stripTagText(text string) string {
+	return strings.TrimSpace(anyTagPattern.ReplaceAllString(text, ""))
+}
+
+func findMarkupTagValue(text string, tagNames []string, patternByTag map[string]*regexp.Regexp) string {
+	for _, tag := range tagNames {
+		pattern := patternByTag[tag]
+		if pattern == nil {
+			continue
+		}
+		if m := pattern.FindStringSubmatch(text); len(m) >= 2 {
+			value := strings.TrimSpace(m[1])
+			if value != "" {
+				return value
+			}
+		}
+	}
+	return ""
+}

internal/util/toolcalls_name_match.go

@@ -0,0 +1,33 @@
+package util
+
+import (
+	"regexp"
+	"strings"
+)
+
+var toolNameLoosePattern = regexp.MustCompile(`[^a-z0-9]+`)
+
+func resolveAllowedToolNameWithLooseMatch(name string, allowed map[string]struct{}, allowedCanonical map[string]string) string {
+	if _, ok := allowed[name]; ok {
+		return name
+	}
+	lower := strings.ToLower(strings.TrimSpace(name))
+	if canonical, ok := allowedCanonical[lower]; ok {
+		return canonical
+	}
+	if idx := strings.LastIndex(lower, "."); idx >= 0 && idx < len(lower)-1 {
+		if canonical, ok := allowedCanonical[lower[idx+1:]]; ok {
+			return canonical
+		}
+	}
+	loose := toolNameLoosePattern.ReplaceAllString(lower, "")
+	if loose == "" {
+		return ""
+	}
+	for candidateLower, canonical := range allowedCanonical {
+		if toolNameLoosePattern.ReplaceAllString(candidateLower, "") == loose {
+			return canonical
+		}
+	}
+	return ""
+}

internal/util/toolcalls_parse.go

@@ -2,12 +2,9 @@ package util
 
 import (
 	"encoding/json"
-	"regexp"
 	"strings"
 )
 
-var toolNameLoosePattern = regexp.MustCompile(`[^a-z0-9]+`)
-
 type ParsedToolCall struct {
 	Name  string         `json:"name"`
 	Input map[string]any `json:"input"`
@@ -19,7 +16,6 @@ type ToolCallParseResult struct {
 	RejectedByPolicy  bool
 	RejectedToolNames []string
 }
-
 func ParseToolCalls(text string, availableToolNames []string) []ParsedToolCall {
 	return ParseToolCallsDetailed(text, availableToolNames).Calls
 }
@@ -29,23 +25,36 @@ func ParseToolCallsDetailed(text string, availableToolNames []string) ToolCallPa
 	if strings.TrimSpace(text) == "" {
 		return result
 	}
-	text = stripFencedCodeBlocks(text)
-	if strings.TrimSpace(text) == "" {
-		return result
-	}
-	result.SawToolCallSyntax = strings.Contains(strings.ToLower(text), "tool_calls")
+	result.SawToolCallSyntax = looksLikeToolCallSyntax(text)
 
 	candidates := buildToolCallCandidates(text)
 	var parsed []ParsedToolCall
 	for _, candidate := range candidates {
-		if tc := parseToolCallsPayload(candidate); len(tc) > 0 {
+		tc := parseToolCallsPayload(candidate)
+		if len(tc) == 0 {
+			tc = parseXMLToolCalls(candidate)
+		}
+		if len(tc) == 0 {
+			tc = parseMarkupToolCalls(candidate)
+		}
+		if len(tc) == 0 {
+			tc = parseTextKVToolCalls(candidate)
+		}
+		if len(tc) > 0 {
 			parsed = tc
 			result.SawToolCallSyntax = true
 			break
 		}
 	}
 	if len(parsed) == 0 {
-		return result
+		parsed = parseXMLToolCalls(text)
+		if len(parsed) == 0 {
+			parsed = parseTextKVToolCalls(text)
+			if len(parsed) == 0 {
+				return result
+			}
+		}
+		result.SawToolCallSyntax = true
 	}
 
 	calls, rejectedNames := filterToolCallsDetailed(parsed, availableToolNames)
@@ -54,7 +63,6 @@ func ParseToolCallsDetailed(text string, availableToolNames []string) ToolCallPa
 	result.RejectedByPolicy = len(rejectedNames) > 0 && len(calls) == 0
 	return result
 }
-
 func ParseStandaloneToolCalls(text string, availableToolNames []string) []ParsedToolCall {
 	return ParseStandaloneToolCallsDetailed(text, availableToolNames).Calls
 }
@@ -65,28 +73,42 @@ func ParseStandaloneToolCallsDetailed(text string, availableToolNames []string)
 	if trimmed == "" {
 		return result
 	}
-	if looksLikeToolExampleContext(trimmed) {
-		return result
-	}
-	result.SawToolCallSyntax = strings.Contains(strings.ToLower(trimmed), "tool_calls")
-	candidates := []string{trimmed}
+	result.SawToolCallSyntax = looksLikeToolCallSyntax(trimmed)
+	candidates := buildToolCallCandidates(trimmed)
+	var parsed []ParsedToolCall
 	for _, candidate := range candidates {
 		candidate = strings.TrimSpace(candidate)
 		if candidate == "" {
 			continue
 		}
-		if !strings.HasPrefix(candidate, "{") && !strings.HasPrefix(candidate, "[") {
-			continue
+		parsed = parseToolCallsPayload(candidate)
+		if len(parsed) == 0 {
+			parsed = parseXMLToolCalls(candidate)
 		}
-		if parsed := parseToolCallsPayload(candidate); len(parsed) > 0 {
-			result.SawToolCallSyntax = true
-			calls, rejectedNames := filterToolCallsDetailed(parsed, availableToolNames)
-			result.Calls = calls
-			result.RejectedToolNames = rejectedNames
-			result.RejectedByPolicy = len(rejectedNames) > 0 && len(calls) == 0
-			return result
+		if len(parsed) == 0 {
+			parsed = parseMarkupToolCalls(candidate)
+		}
+		if len(parsed) == 0 {
+			parsed = parseTextKVToolCalls(candidate)
+		}
+		if len(parsed) > 0 {
+			break
 		}
 	}
+	if len(parsed) == 0 {
+		parsed = parseXMLToolCalls(trimmed)
+		if len(parsed) == 0 {
+			parsed = parseTextKVToolCalls(trimmed)
+			if len(parsed) == 0 {
+				return result
+			}
+		}
+	}
+	result.SawToolCallSyntax = true
+	calls, rejectedNames := filterToolCallsDetailed(parsed, availableToolNames)
+	result.Calls = calls
+	result.RejectedToolNames = rejectedNames
+	result.RejectedByPolicy = len(rejectedNames) > 0 && len(calls) == 0
 	return result
 }
 
@@ -106,27 +128,32 @@ func filterToolCallsDetailed(parsed []ParsedToolCall, availableToolNames []strin
 	}
 	if len(allowed) == 0 {
 		rejectedSet := map[string]struct{}{}
+		rejected := make([]string, 0, len(parsed))
 		for _, tc := range parsed {
 			if tc.Name == "" {
 				continue
 			}
+			if _, ok := rejectedSet[tc.Name]; ok {
+				continue
+			}
 			rejectedSet[tc.Name] = struct{}{}
-		}
-		rejected := make([]string, 0, len(rejectedSet))
-		for name := range rejectedSet {
-			rejected = append(rejected, name)
+			rejected = append(rejected, tc.Name)
 		}
 		return nil, rejected
 	}
 	out := make([]ParsedToolCall, 0, len(parsed))
 	rejectedSet := map[string]struct{}{}
+	rejected := make([]string, 0)
 	for _, tc := range parsed {
 		if tc.Name == "" {
 			continue
 		}
 		matchedName := resolveAllowedToolName(tc.Name, allowed, allowedCanonical)
 		if matchedName == "" {
-			rejectedSet[tc.Name] = struct{}{}
+			if _, ok := rejectedSet[tc.Name]; !ok {
+				rejectedSet[tc.Name] = struct{}{}
+				rejected = append(rejected, tc.Name)
+			}
 			continue
 		}
 		tc.Name = matchedName
@@ -135,42 +162,23 @@ func filterToolCallsDetailed(parsed []ParsedToolCall, availableToolNames []strin
 		}
 		out = append(out, tc)
 	}
-	rejected := make([]string, 0, len(rejectedSet))
-	for name := range rejectedSet {
-		rejected = append(rejected, name)
-	}
 	return out, rejected
 }
 
 func resolveAllowedToolName(name string, allowed map[string]struct{}, allowedCanonical map[string]string) string {
-	if _, ok := allowed[name]; ok {
-		return name
-	}
-	lower := strings.ToLower(strings.TrimSpace(name))
-	if canonical, ok := allowedCanonical[lower]; ok {
-		return canonical
-	}
-	if idx := strings.LastIndex(lower, "."); idx >= 0 && idx < len(lower)-1 {
-		if canonical, ok := allowedCanonical[lower[idx+1:]]; ok {
-			return canonical
-		}
-	}
-	loose := toolNameLoosePattern.ReplaceAllString(lower, "")
-	if loose == "" {
-		return ""
-	}
-	for candidateLower, canonical := range allowedCanonical {
-		if toolNameLoosePattern.ReplaceAllString(candidateLower, "") == loose {
-			return canonical
-		}
-	}
-	return ""
+	return resolveAllowedToolNameWithLooseMatch(name, allowed, allowedCanonical)
 }
 
 func parseToolCallsPayload(payload string) []ParsedToolCall {
 	var decoded any
 	if err := json.Unmarshal([]byte(payload), &decoded); err != nil {
-		return nil
+		// Try to repair backslashes first! Because LLMs often mix these two problems.
+		repaired := repairInvalidJSONBackslashes(payload)
+		// Try loose repair on top of that
+		repaired = RepairLooseJSON(repaired)
+		if err := json.Unmarshal([]byte(repaired), &decoded); err != nil {
+			return nil
+		}
 	}
 	switch v := decoded.(type) {
 	case map[string]any:
@@ -186,6 +194,15 @@ func parseToolCallsPayload(payload string) []ParsedToolCall {
 	return nil
 }
 
+func looksLikeToolCallSyntax(text string) bool {
+	lower := strings.ToLower(text)
+	return strings.Contains(lower, "tool_calls") ||
+		strings.Contains(lower, "<tool_call") ||
+		strings.Contains(lower, "<function_call") ||
+		strings.Contains(lower, "<invoke") ||
+		strings.Contains(lower, "function.name:")
+}
+
 func parseToolCallList(v any) []ParsedToolCall {
 	items, ok := v.([]any)
 	if !ok {
@@ -238,32 +255,3 @@ func parseToolCallItem(m map[string]any) (ParsedToolCall, bool) {
 		Input: parseToolCallInput(inputRaw),
 	}, true
 }
-
-func parseToolCallInput(v any) map[string]any {
-	switch x := v.(type) {
-	case nil:
-		return map[string]any{}
-	case map[string]any:
-		return x
-	case string:
-		raw := strings.TrimSpace(x)
-		if raw == "" {
-			return map[string]any{}
-		}
-		var parsed map[string]any
-		if err := json.Unmarshal([]byte(raw), &parsed); err == nil && parsed != nil {
-			return parsed
-		}
-		return map[string]any{"_raw": raw}
-	default:
-		b, err := json.Marshal(x)
-		if err != nil {
-			return map[string]any{}
-		}
-		var parsed map[string]any
-		if err := json.Unmarshal(b, &parsed); err == nil && parsed != nil {
-			return parsed
-		}
-		return map[string]any{}
-	}
-}

internal/util/toolcalls_parse_markup.go

@@ -0,0 +1,235 @@
+package util
+
+import (
+	"encoding/json"
+	"encoding/xml"
+	"regexp"
+	"strings"
+)
+
+var xmlToolCallPattern = regexp.MustCompile(`(?is)<tool_call>\s*(.*?)\s*</tool_call>`)
+var functionCallPattern = regexp.MustCompile(`(?is)<function_call>\s*([^<]+?)\s*</function_call>`)
+var functionParamPattern = regexp.MustCompile(`(?is)<function\s+parameter\s+name="([^"]+)"\s*>\s*(.*?)\s*</function\s+parameter>`)
+var antmlFunctionCallPattern = regexp.MustCompile(`(?is)<(?:[a-z0-9_]+:)?function_call[^>]*(?:name|function)="([^"]+)"[^>]*>\s*(.*?)\s*</(?:[a-z0-9_]+:)?function_call>`)
+var antmlArgumentPattern = regexp.MustCompile(`(?is)<(?:[a-z0-9_]+:)?argument\s+name="([^"]+)"\s*>\s*(.*?)\s*</(?:[a-z0-9_]+:)?argument>`)
+var antmlParametersPattern = regexp.MustCompile(`(?is)<(?:[a-z0-9_]+:)?parameters\s*>\s*(\{.*?\})\s*</(?:[a-z0-9_]+:)?parameters>`)
+var invokeCallPattern = regexp.MustCompile(`(?is)<invoke\s+name="([^"]+)"\s*>(.*?)</invoke>`)
+var invokeParamPattern = regexp.MustCompile(`(?is)<parameter\s+name="([^"]+)"\s*>\s*(.*?)\s*</parameter>`)
+
+func parseXMLToolCalls(text string) []ParsedToolCall {
+	matches := xmlToolCallPattern.FindAllString(text, -1)
+	out := make([]ParsedToolCall, 0, len(matches)+1)
+	for _, block := range matches {
+		call, ok := parseSingleXMLToolCall(block)
+		if !ok {
+			continue
+		}
+		out = append(out, call)
+	}
+	if len(out) > 0 {
+		return out
+	}
+	if call, ok := parseFunctionCallTagStyle(text); ok {
+		return []ParsedToolCall{call}
+	}
+	if calls := parseAntmlFunctionCallStyles(text); len(calls) > 0 {
+		return calls
+	}
+	if call, ok := parseInvokeFunctionCallStyle(text); ok {
+		return []ParsedToolCall{call}
+	}
+	return nil
+}
+
+func parseSingleXMLToolCall(block string) (ParsedToolCall, bool) {
+	inner := strings.TrimSpace(block)
+	inner = strings.TrimPrefix(inner, "<tool_call>")
+	inner = strings.TrimSuffix(inner, "</tool_call>")
+	inner = strings.TrimSpace(inner)
+	if strings.HasPrefix(inner, "{") {
+		var payload map[string]any
+		if err := json.Unmarshal([]byte(inner), &payload); err == nil {
+			name := strings.TrimSpace(asString(payload["tool"]))
+			if name == "" {
+				name = strings.TrimSpace(asString(payload["tool_name"]))
+			}
+			if name != "" {
+				input := map[string]any{}
+				if params, ok := payload["params"].(map[string]any); ok {
+					input = params
+				} else if params, ok := payload["parameters"].(map[string]any); ok {
+					input = params
+				}
+				return ParsedToolCall{Name: name, Input: input}, true
+			}
+		}
+	}
+
+	dec := xml.NewDecoder(strings.NewReader(block))
+	name := ""
+	params := map[string]any{}
+	inParams := false
+	inTool := false
+	for {
+		tok, err := dec.Token()
+		if err != nil {
+			break
+		}
+		switch t := tok.(type) {
+		case xml.StartElement:
+			tag := strings.ToLower(t.Name.Local)
+			switch tag {
+			case "tool":
+				inTool = true
+				for _, attr := range t.Attr {
+					if strings.EqualFold(strings.TrimSpace(attr.Name.Local), "name") && strings.TrimSpace(name) == "" {
+						name = strings.TrimSpace(attr.Value)
+					}
+				}
+			case "parameters":
+				inParams = true
+			case "tool_name", "name":
+				var v string
+				if err := dec.DecodeElement(&v, &t); err == nil && strings.TrimSpace(v) != "" {
+					name = strings.TrimSpace(v)
+				}
+			case "input", "arguments", "argument", "args", "params":
+				var v string
+				if err := dec.DecodeElement(&v, &t); err == nil && strings.TrimSpace(v) != "" {
+					if parsed := parseToolCallInput(strings.TrimSpace(v)); len(parsed) > 0 {
+						for k, vv := range parsed {
+							params[k] = vv
+						}
+					}
+				}
+			default:
+				if inParams || inTool {
+					var v string
+					if err := dec.DecodeElement(&v, &t); err == nil {
+						params[t.Name.Local] = strings.TrimSpace(v)
+					}
+				}
+			}
+		case xml.EndElement:
+			tag := strings.ToLower(t.Name.Local)
+			if tag == "parameters" {
+				inParams = false
+			}
+			if tag == "tool" {
+				inTool = false
+			}
+		}
+	}
+	if strings.TrimSpace(name) == "" {
+		return ParsedToolCall{}, false
+	}
+	return ParsedToolCall{Name: strings.TrimSpace(name), Input: params}, true
+}
+
+func parseFunctionCallTagStyle(text string) (ParsedToolCall, bool) {
+	m := functionCallPattern.FindStringSubmatch(text)
+	if len(m) < 2 {
+		return ParsedToolCall{}, false
+	}
+	name := strings.TrimSpace(m[1])
+	if name == "" {
+		return ParsedToolCall{}, false
+	}
+	input := map[string]any{}
+	for _, pm := range functionParamPattern.FindAllStringSubmatch(text, -1) {
+		if len(pm) < 3 {
+			continue
+		}
+		key := strings.TrimSpace(pm[1])
+		val := strings.TrimSpace(pm[2])
+		if key != "" {
+			input[key] = val
+		}
+	}
+	return ParsedToolCall{Name: name, Input: input}, true
+}
+
+func parseAntmlFunctionCallStyles(text string) []ParsedToolCall {
+	matches := antmlFunctionCallPattern.FindAllStringSubmatch(text, -1)
+	if len(matches) == 0 {
+		return nil
+	}
+	out := make([]ParsedToolCall, 0, len(matches))
+	for _, m := range matches {
+		if call, ok := parseSingleAntmlFunctionCallMatch(m); ok {
+			out = append(out, call)
+		}
+	}
+	if len(out) == 0 {
+		return nil
+	}
+	return out
+}
+
+func parseSingleAntmlFunctionCallMatch(m []string) (ParsedToolCall, bool) {
+	if len(m) < 3 {
+		return ParsedToolCall{}, false
+	}
+	name := strings.TrimSpace(m[1])
+	if name == "" {
+		return ParsedToolCall{}, false
+	}
+	body := strings.TrimSpace(m[2])
+	input := map[string]any{}
+	if strings.HasPrefix(body, "{") {
+		if err := json.Unmarshal([]byte(body), &input); err == nil {
+			return ParsedToolCall{Name: name, Input: input}, true
+		}
+	}
+	if pm := antmlParametersPattern.FindStringSubmatch(body); len(pm) >= 2 {
+		if err := json.Unmarshal([]byte(strings.TrimSpace(pm[1])), &input); err == nil {
+			return ParsedToolCall{Name: name, Input: input}, true
+		}
+	}
+	for _, am := range antmlArgumentPattern.FindAllStringSubmatch(body, -1) {
+		if len(am) < 3 {
+			continue
+		}
+		k := strings.TrimSpace(am[1])
+		v := strings.TrimSpace(am[2])
+		if k != "" {
+			input[k] = v
+		}
+	}
+	return ParsedToolCall{Name: name, Input: input}, true
+}
+
+func parseInvokeFunctionCallStyle(text string) (ParsedToolCall, bool) {
+	m := invokeCallPattern.FindStringSubmatch(text)
+	if len(m) < 3 {
+		return ParsedToolCall{}, false
+	}
+	name := strings.TrimSpace(m[1])
+	if name == "" {
+		return ParsedToolCall{}, false
+	}
+	input := map[string]any{}
+	for _, pm := range invokeParamPattern.FindAllStringSubmatch(m[2], -1) {
+		if len(pm) < 3 {
+			continue
+		}
+		k := strings.TrimSpace(pm[1])
+		v := strings.TrimSpace(pm[2])
+		if k != "" {
+			input[k] = v
+		}
+	}
+	if len(input) == 0 {
+		if argsRaw := findMarkupTagValue(m[2], toolCallMarkupArgsTagNames, toolCallMarkupArgsPatternByTag); argsRaw != "" {
+			input = parseMarkupInput(argsRaw)
+		} else if kv := parseMarkupKVObject(m[2]); len(kv) > 0 {
+			input = kv
+		}
+	}
+	return ParsedToolCall{Name: name, Input: input}, true
+}
+
+func asString(v any) string {
+	s, _ := v.(string)
+	return s
+}

internal/util/toolcalls_test.go

@@ -1,6 +1,9 @@
 package util
 
-import "testing"
+import (
+	"strings"
+	"testing"
+)
 
 func TestParseToolCalls(t *testing.T) {
 	text := `prefix {"tool_calls":[{"name":"search","input":{"q":"golang"}}]} suffix`
@@ -19,8 +22,8 @@ func TestParseToolCalls(t *testing.T) {
 func TestParseToolCallsFromFencedJSON(t *testing.T) {
 	text := "I will call tools now\n```json\n{\"tool_calls\":[{\"name\":\"search\",\"input\":{\"q\":\"news\"}}]}\n```"
 	calls := ParseToolCalls(text, []string{"search"})
-	if len(calls) != 0 {
-		t.Fatalf("expected fenced tool_call example to be ignored, got %#v", calls)
+	if len(calls) != 1 {
+		t.Fatalf("expected fenced tool_call payload to be parsed, got %#v", calls)
 	}
 }
 
@@ -96,10 +99,10 @@ func TestFormatOpenAIToolCalls(t *testing.T) {
 	}
 }
 
-func TestParseStandaloneToolCallsOnlyMatchesStandalonePayload(t *testing.T) {
+func TestParseStandaloneToolCallsSupportsMixedProsePayload(t *testing.T) {
 	mixed := `这里是示例：{"tool_calls":[{"name":"search","input":{"q":"go"}}]}`
-	if calls := ParseStandaloneToolCalls(mixed, []string{"search"}); len(calls) != 0 {
-		t.Fatalf("expected standalone parser to ignore mixed prose, got %#v", calls)
+	if calls := ParseStandaloneToolCalls(mixed, []string{"search"}); len(calls) != 1 {
+		t.Fatalf("expected standalone parser to parse mixed prose payload, got %#v", calls)
 	}
 
 	standalone := `{"tool_calls":[{"name":"search","input":{"q":"go"}}]}`
@@ -109,10 +112,10 @@ func TestParseStandaloneToolCallsOnlyMatchesStandalonePayload(t *testing.T) {
 	}
 }
 
-func TestParseStandaloneToolCallsIgnoresFencedCodeBlock(t *testing.T) {
+func TestParseStandaloneToolCallsParsesFencedCodeBlock(t *testing.T) {
 	fenced := "```json\n{\"tool_calls\":[{\"name\":\"search\",\"input\":{\"q\":\"go\"}}]}\n```"
-	if calls := ParseStandaloneToolCalls(fenced, []string{"search"}); len(calls) != 0 {
-		t.Fatalf("expected fenced tool_call example to be ignored, got %#v", calls)
+	if calls := ParseStandaloneToolCalls(fenced, []string{"search"}); len(calls) != 1 {
+		t.Fatalf("expected fenced tool_call payload to be parsed, got %#v", calls)
 	}
 }
 
@@ -137,3 +140,380 @@ func TestParseToolCallsAllowsPunctuationVariantToolName(t *testing.T) {
 		t.Fatalf("expected canonical tool name read_file, got %q", calls[0].Name)
 	}
 }
+
+func TestParseToolCallsSupportsClaudeXMLToolCall(t *testing.T) {
+	text := `<tool_call><tool_name>Bash</tool_name><parameters><command>pwd</command><description>show cwd</description></parameters></tool_call>`
+	calls := ParseToolCalls(text, []string{"bash"})
+	if len(calls) != 1 {
+		t.Fatalf("expected 1 call, got %#v", calls)
+	}
+	if calls[0].Name != "bash" {
+		t.Fatalf("expected canonical tool name bash, got %q", calls[0].Name)
+	}
+	if calls[0].Input["command"] != "pwd" {
+		t.Fatalf("expected command argument, got %#v", calls[0].Input)
+	}
+}
+
+func TestParseToolCallsDetailedMarksXMLToolCallSyntax(t *testing.T) {
+	text := `<tool_call><tool_name>Bash</tool_name><parameters><command>pwd</command></parameters></tool_call>`
+	res := ParseToolCallsDetailed(text, []string{"bash"})
+	if !res.SawToolCallSyntax {
+		t.Fatalf("expected SawToolCallSyntax=true, got %#v", res)
+	}
+	if len(res.Calls) != 1 {
+		t.Fatalf("expected one parsed call, got %#v", res)
+	}
+}
+
+func TestParseToolCallsSupportsClaudeXMLJSONToolCall(t *testing.T) {
+	text := `<tool_call>{"tool":"Bash","params":{"command":"pwd","description":"show cwd"}}</tool_call>`
+	calls := ParseToolCalls(text, []string{"bash"})
+	if len(calls) != 1 {
+		t.Fatalf("expected 1 call, got %#v", calls)
+	}
+	if calls[0].Name != "bash" {
+		t.Fatalf("expected canonical tool name bash, got %q", calls[0].Name)
+	}
+	if calls[0].Input["command"] != "pwd" {
+		t.Fatalf("expected command argument, got %#v", calls[0].Input)
+	}
+}
+
+func TestParseToolCallsSupportsFunctionCallTagStyle(t *testing.T) {
+	text := `<function_call>Bash</function_call><function parameter name="command">ls -la</function parameter><function parameter name="description">list</function parameter>`
+	calls := ParseToolCalls(text, []string{"bash"})
+	if len(calls) != 1 {
+		t.Fatalf("expected 1 call, got %#v", calls)
+	}
+	if calls[0].Name != "bash" {
+		t.Fatalf("expected canonical tool name bash, got %q", calls[0].Name)
+	}
+	if calls[0].Input["command"] != "ls -la" {
+		t.Fatalf("expected command argument, got %#v", calls[0].Input)
+	}
+}
+
+func TestParseToolCallsSupportsAntmlFunctionCallStyle(t *testing.T) {
+	text := `<antml:function_calls><antml:function_call name="Bash">{"command":"pwd","description":"x"}</antml:function_call></antml:function_calls>`
+	calls := ParseToolCalls(text, []string{"bash"})
+	if len(calls) != 1 {
+		t.Fatalf("expected 1 call, got %#v", calls)
+	}
+	if calls[0].Name != "bash" {
+		t.Fatalf("expected canonical tool name bash, got %q", calls[0].Name)
+	}
+	if calls[0].Input["command"] != "pwd" {
+		t.Fatalf("expected command argument, got %#v", calls[0].Input)
+	}
+}
+
+func TestParseToolCallsSupportsAntmlArgumentStyle(t *testing.T) {
+	text := `<antml:function_calls><antml:function_call id="1" name="Bash"><antml:argument name="command">pwd</antml:argument><antml:argument name="description">x</antml:argument></antml:function_call></antml:function_calls>`
+	calls := ParseToolCalls(text, []string{"bash"})
+	if len(calls) != 1 {
+		t.Fatalf("expected 1 call, got %#v", calls)
+	}
+	if calls[0].Name != "bash" {
+		t.Fatalf("expected canonical tool name bash, got %q", calls[0].Name)
+	}
+	if calls[0].Input["command"] != "pwd" {
+		t.Fatalf("expected command argument, got %#v", calls[0].Input)
+	}
+}
+
+func TestParseToolCallsSupportsInvokeFunctionCallStyle(t *testing.T) {
+	text := `<function_calls><invoke name="Bash"><parameter name="command">pwd</parameter><parameter name="description">d</parameter></invoke></function_calls>`
+	calls := ParseToolCalls(text, []string{"bash"})
+	if len(calls) != 1 {
+		t.Fatalf("expected 1 call, got %#v", calls)
+	}
+	if calls[0].Name != "bash" {
+		t.Fatalf("expected canonical tool name bash, got %q", calls[0].Name)
+	}
+	if calls[0].Input["command"] != "pwd" {
+		t.Fatalf("expected command argument, got %#v", calls[0].Input)
+	}
+}
+
+func TestParseToolCallsSupportsNestedToolTagStyle(t *testing.T) {
+	text := `<tool_call><tool name="Bash"><command>pwd</command><description>show cwd</description></tool></tool_call>`
+	calls := ParseToolCalls(text, []string{"bash"})
+	if len(calls) != 1 {
+		t.Fatalf("expected 1 call, got %#v", calls)
+	}
+	if calls[0].Name != "bash" {
+		t.Fatalf("expected canonical tool name bash, got %q", calls[0].Name)
+	}
+	if calls[0].Input["command"] != "pwd" {
+		t.Fatalf("expected command argument, got %#v", calls[0].Input)
+	}
+}
+
+func TestParseToolCallsSupportsAntmlFunctionAttributeWithParametersTag(t *testing.T) {
+	text := `<antml:function_calls><antml:function_call id="x" function="Bash"><antml:parameters>{"command":"pwd"}</antml:parameters></antml:function_call></antml:function_calls>`
+	calls := ParseToolCalls(text, []string{"bash"})
+	if len(calls) != 1 {
+		t.Fatalf("expected 1 call, got %#v", calls)
+	}
+	if calls[0].Name != "bash" {
+		t.Fatalf("expected canonical tool name bash, got %q", calls[0].Name)
+	}
+	if calls[0].Input["command"] != "pwd" {
+		t.Fatalf("expected command argument, got %#v", calls[0].Input)
+	}
+}
+
+func TestParseToolCallsSupportsMultipleAntmlFunctionCalls(t *testing.T) {
+	text := `<antml:function_calls><antml:function_call id="1" function="Bash"><antml:parameters>{"command":"pwd"}</antml:parameters></antml:function_call><antml:function_call id="2" function="Read"><antml:parameters>{"file_path":"README.md"}</antml:parameters></antml:function_call></antml:function_calls>`
+	calls := ParseToolCalls(text, []string{"bash", "read"})
+	if len(calls) != 2 {
+		t.Fatalf("expected 2 calls, got %#v", calls)
+	}
+	if calls[0].Name != "bash" || calls[1].Name != "read" {
+		t.Fatalf("expected canonical names [bash read], got %#v", calls)
+	}
+}
+
+func TestParseToolCallsDoesNotAcceptMismatchedMarkupTags(t *testing.T) {
+	text := `<tool_call><name>read_file</function><arguments>{"path":"README.md"}</arguments></tool_call>`
+	calls := ParseToolCalls(text, []string{"read_file"})
+	if len(calls) != 0 {
+		t.Fatalf("expected mismatched tags to be rejected, got %#v", calls)
+	}
+}
+
+func TestRepairInvalidJSONBackslashes(t *testing.T) {
+	tests := []struct {
+		input    string
+		expected string
+	}{
+		{`{"path": "C:\Users\name"}`, `{"path": "C:\\Users\name"}`},
+		{`{"cmd": "cd D:\git_codes"}`, `{"cmd": "cd D:\\git_codes"}`},
+		{`{"text": "line1\nline2"}`, `{"text": "line1\nline2"}`},
+		{`{"path": "D:\\back\\slash"}`, `{"path": "D:\\back\\slash"}`},
+		{`{"unicode": "\u2705"}`, `{"unicode": "\u2705"}`},
+		{`{"invalid_u": "\u123"}`, `{"invalid_u": "\\u123"}`},
+	}
+
+	for _, tt := range tests {
+		got := repairInvalidJSONBackslashes(tt.input)
+		if got != tt.expected {
+			t.Errorf("repairInvalidJSONBackslashes(%s) = %s; want %s", tt.input, got, tt.expected)
+		}
+	}
+}
+
+func TestRepairLooseJSON(t *testing.T) {
+	tests := []struct {
+		input    string
+		expected string
+	}{
+		{`{tool_calls: [{"name": "search", "input": {"q": "go"}}]}`, `{"tool_calls": [{"name": "search", "input": {"q": "go"}}]}`},
+		{`{name: "search", input: {q: "go"}}`, `{"name": "search", "input": {"q": "go"}}`},
+	}
+
+	for _, tt := range tests {
+		got := RepairLooseJSON(tt.input)
+		if got != tt.expected {
+			t.Errorf("RepairLooseJSON(%s) = %s; want %s", tt.input, got, tt.expected)
+		}
+	}
+}
+
+func TestParseToolCallsWithUnquotedKeys(t *testing.T) {
+	text := `这里是列表：{tool_calls: [{"name": "todowrite", "input": {"todos": "test"}}]}`
+	availableTools := []string{"todowrite"}
+
+	parsed := ParseToolCalls(text, availableTools)
+	if len(parsed) != 1 {
+		t.Fatalf("expected 1 tool call, got %d", len(parsed))
+	}
+	if parsed[0].Name != "todowrite" {
+		t.Errorf("expected tool todowrite, got %s", parsed[0].Name)
+	}
+}
+
+func TestParseToolCallsWithInvalidBackslashes(t *testing.T) {
+	// DeepSeek sometimes outputs Windows paths with single backslashes in JSON strings
+	// Note: using raw string to simulate what AI actually sends in the stream
+	text := `好的，执行以下命令：{"name": "execute_command", "input": "{\"command\": \"cd D:\git_codes && dir\"}"}`
+	availableTools := []string{"execute_command"}
+
+	parsed := ParseToolCalls(text, availableTools)
+	// If standard JSON fails, buildToolCallCandidates should still extract the object,
+	// and parseToolCallsPayload should repair it.
+	if len(parsed) != 1 {
+		// If it still fails, let's see why
+		candidates := buildToolCallCandidates(text)
+		t.Logf("Candidates: %v", candidates)
+		t.Fatalf("expected 1 tool call, got %d", len(parsed))
+	}
+
+	cmd, ok := parsed[0].Input["command"].(string)
+	if !ok {
+		t.Fatalf("expected command string in input, got %v", parsed[0].Input)
+	}
+
+	expected := "cd D:\\git_codes && dir"
+	if cmd != expected {
+		t.Errorf("expected command %q, got %q", expected, cmd)
+	}
+}
+
+func TestParseToolCallsWithDeepSeekHallucination(t *testing.T) {
+	// 模拟 DeepSeek 典型的幻觉输出：未加引号的键名 + 包含 Windows 路径的嵌套 JSON 字符串 + 漏掉列表的方括号
+	text := `检测到实施意图——实现经典算法。需在misc/目录创建Python文件。
+关键约束:
+1. Windows UTF-8编码处理
+2. 必须用绝对路径导入
+3. 禁止write覆盖已有文件（misc/目录允许创建新文件）
+将任务分解并委托：
+- 研究8皇后算法模式（并行探索）
+- 实现带可视化输出的解决方案（unspecified-high）
+先创建todo列表追踪步骤。
+{tool_calls: [{"name": "todowrite", "input": {"todos": {"content": "研究8皇后问题算法模式（回溯法）和输出格式", "status": "pending", "priority": "high"}, {"content": "在misc/目录创建8皇后Python脚本，包含完整解决方案和可视化输出", "status": "pending", "priority": "high"}, {"content": "验证脚本正确性（运行测试）", "status": "pending", "priority": "medium"}}}]}`
+
+	availableTools := []string{"todowrite"}
+	parsed := ParseToolCalls(text, availableTools)
+
+	if len(parsed) != 1 {
+		cands := buildToolCallCandidates(text)
+		for i, c := range cands {
+			t.Logf("CAND %d: %s", i, c)
+			repaired := RepairLooseJSON(c)
+			t.Logf("  REPAIRED: %s", repaired)
+		}
+		t.Fatalf("expected 1 tool call, got %d. Candidates: %v", len(parsed), buildToolCallCandidates(text))
+	}
+
+	if parsed[0].Name != "todowrite" {
+		t.Errorf("expected tool name 'todowrite', got %q", parsed[0].Name)
+	}
+
+	todos, ok := parsed[0].Input["todos"].([]any)
+	if !ok {
+		t.Fatalf("expected 'todos' to be parsed as a list, got %T: %#v", parsed[0].Input["todos"], parsed[0].Input["todos"])
+	}
+	if len(todos) != 3 {
+		t.Errorf("expected 3 todo items, got %d", len(todos))
+	}
+}
+
+func TestParseToolCallsWithMixedWindowsPaths(t *testing.T) {
+	// 更复杂的案例：嵌套 JSON 字符串中的反斜杠未转义
+	text := `关键约束: 1. Windows UTF-8编码处理 2. 必须用绝对路径导入 D:\git_codes\ds2api\misc
+{tool_calls: [{"name": "write_file", "input": "{\"path\": \"D:\\git_codes\\ds2api\\misc\\queens.py\", \"content\": \"print('hello')\"}"}]}`
+
+	availableTools := []string{"write_file"}
+	parsed := ParseToolCalls(text, availableTools)
+
+	if len(parsed) != 1 {
+		t.Fatalf("expected 1 tool call from mixed text with paths, got %d", len(parsed))
+	}
+
+	path, _ := parsed[0].Input["path"].(string)
+	// 在解析后的 Go map 中，反斜杠应该被还原
+	if !strings.Contains(path, "D:\\git_codes") && !strings.Contains(path, "D:/git_codes") {
+		t.Errorf("expected path to contain Windows style separators, got %q", path)
+	}
+}
+
+func TestParseToolCallInputRepairsControlCharsInPath(t *testing.T) {
+	in := `{"path":"D:\tmp\new\readme.txt","content":"line1\nline2"}`
+	parsed := parseToolCallInput(in)
+
+	path, ok := parsed["path"].(string)
+	if !ok {
+		t.Fatalf("expected path string in parsed input, got %#v", parsed["path"])
+	}
+	if path != `D:\tmp\new\readme.txt` {
+		t.Fatalf("expected repaired windows path, got %q", path)
+	}
+
+	content, ok := parsed["content"].(string)
+	if !ok {
+		t.Fatalf("expected content string in parsed input, got %#v", parsed["content"])
+	}
+	if content != "line1\nline2" {
+		t.Fatalf("expected non-path field to keep decoded escapes, got %q", content)
+	}
+}
+
+func TestRepairLooseJSONWithNestedObjects(t *testing.T) {
+	// 测试嵌套对象的修复：DeepSeek 幻觉输出，每个元素内部包含嵌套 {}
+	// 注意：正则只支持单层嵌套，不支持更深层次的嵌套
+	tests := []struct {
+		name     string
+		input    string
+		expected string
+	}{
+		// 1. 单层嵌套对象（核心修复目标）
+		{
+			name:     "单层嵌套 - 2个元素",
+			input:    `"todos": {"content": "研究算法", "input": {"q": "8 queens"}}, {"content": "实现", "input": {"path": "queens.py"}}`,
+			expected: `"todos": [{"content": "研究算法", "input": {"q": "8 queens"}}, {"content": "实现", "input": {"path": "queens.py"}}]`,
+		},
+		// 2. 3个单层嵌套对象
+		{
+			name:     "3个单层嵌套对象",
+			input:    `"items": {"a": {"x":1}}, {"b": {"y":2}}, {"c": {"z":3}}`,
+			expected: `"items": [{"a": {"x":1}}, {"b": {"y":2}}, {"c": {"z":3}}]`,
+		},
+		// 3. 混合嵌套：有些字段是对象，有些是原始值
+		{
+			name:     "混合嵌套 - 对象和原始值混合",
+			input:    `"items": {"name": "test", "config": {"timeout": 30}}, {"name": "test2", "config": {"timeout": 60}}`,
+			expected: `"items": [{"name": "test", "config": {"timeout": 30}}, {"name": "test2", "config": {"timeout": 60}}]`,
+		},
+		// 4. 4个嵌套对象（边界测试）
+		{
+			name:     "4个嵌套对象",
+			input:    `"todos": {"id": 1}, {"id": 2}, {"id": 3}, {"id": 4}`,
+			expected: `"todos": [{"id": 1}, {"id": 2}, {"id": 3}, {"id": 4}]`,
+		},
+		// 5. DeepSeek 典型幻觉：无空格逗号分隔
+		{
+			name:     "无空格逗号分隔",
+			input:    `"results": {"name": "a"}, {"name": "b"}, {"name": "c"}`,
+			expected: `"results": [{"name": "a"}, {"name": "b"}, {"name": "c"}]`,
+		},
+		// 6. 嵌套数组（数组在对象内，不是深层嵌套）
+		{
+			name:     "对象内包含数组",
+			input:    `"data": {"items": [1,2,3]}, {"items": [4,5,6]}`,
+			expected: `"data": [{"items": [1,2,3]}, {"items": [4,5,6]}]`,
+		},
+		// 7. 真实的 DeepSeek 8皇后问题输出
+		{
+			name:     "DeepSeek 8皇后真实输出",
+			input:    `"todos": {"content": "研究8皇后算法", "status": "pending"}, {"content": "实现Python脚本", "status": "pending"}, {"content": "验证结果", "status": "pending"}`,
+			expected: `"todos": [{"content": "研究8皇后算法", "status": "pending"}, {"content": "实现Python脚本", "status": "pending"}, {"content": "验证结果", "status": "pending"}]`,
+		},
+		// 8. 简单无嵌套对象（回归测试）
+		{
+			name:     "简单无嵌套对象",
+			input:    `"items": {"a": 1}, {"b": 2}`,
+			expected: `"items": [{"a": 1}, {"b": 2}]`,
+		},
+		// 9. 更复杂的单层嵌套
+		{
+			name:     "复杂单层嵌套",
+			input:    `"functions": {"name": "execute", "input": {"command": "ls"}}, {"name": "read", "input": {"file": "a.txt"}}`,
+			expected: `"functions": [{"name": "execute", "input": {"command": "ls"}}, {"name": "read", "input": {"file": "a.txt"}}]`,
+		},
+		// 10. 5个嵌套对象
+		{
+			name:     "5个嵌套对象",
+			input:    `"tasks": {"id":1}, {"id":2}, {"id":3}, {"id":4}, {"id":5}`,
+			expected: `"tasks": [{"id":1}, {"id":2}, {"id":3}, {"id":4}, {"id":5}]`,
+		},
+	}
+
+	for _, tt := range tests {
+		got := RepairLooseJSON(tt.input)
+		if got != tt.expected {
+			t.Errorf("[%s] RepairLooseJSON with nested objects:\n  input:    %s\n  got:      %s\n  expected: %s", tt.name, tt.input, got, tt.expected)
+		}
+	}
+}

internal/util/toolcalls_textkv.go

@@ -0,0 +1,55 @@
+package util
+
+import (
+	"regexp"
+	"strings"
+)
+
+var textKVNamePattern = regexp.MustCompile(`(?is)function\.name:\s*([a-zA-Z0-9_\-.]+)`)
+
+func parseTextKVToolCalls(text string) []ParsedToolCall {
+	var out []ParsedToolCall
+	matches := textKVNamePattern.FindAllStringSubmatchIndex(text, -1)
+	if len(matches) == 0 {
+		return nil
+	}
+
+	for i, match := range matches {
+		name := text[match[2]:match[3]]
+
+		offset := match[1]
+		endSearch := len(text)
+		if i+1 < len(matches) {
+			endSearch = matches[i+1][0]
+		}
+
+		searchArea := text[offset:endSearch]
+		argIdx := strings.Index(searchArea, "function.arguments:")
+		if argIdx < 0 {
+			continue
+		}
+
+		startIdx := offset + argIdx + len("function.arguments:")
+		braceIdx := strings.IndexByte(text[startIdx:endSearch], '{')
+		if braceIdx < 0 {
+			continue
+		}
+
+		actualStart := startIdx + braceIdx
+		objJson, _, ok := extractJSONObject(text, actualStart)
+		if !ok {
+			continue
+		}
+
+		input := parseToolCallInput(objJson)
+		out = append(out, ParsedToolCall{
+			Name:  name,
+			Input: input,
+		})
+	}
+
+	if len(out) == 0 {
+		return nil
+	}
+	return out
+}

internal/util/toolcalls_textkv_test.go

@@ -0,0 +1,63 @@
+package util
+
+import (
+	"testing"
+)
+
+func TestParseTextKVToolCalls_Basic(t *testing.T) {
+	text := `
+[TOOL_CALL_HISTORY]
+status: already_called
+origin: assistant
+not_user_input: true
+tool_call_id: call_3fcd15235eb94f7eae3a8de5a9cfa36b
+function.name: execute_command
+function.arguments: {"command":"cd scripts && python check_syntax.py example.py","cwd":null,"timeout":30}
+[/TOOL_CALL_HISTORY]
+
+Some other text thinking...
+`
+	calls := ParseToolCalls(text, []string{"execute_command"})
+	if len(calls) != 1 {
+		t.Fatalf("expected 1 call, got %d", len(calls))
+	}
+	if calls[0].Name != "execute_command" {
+		t.Fatalf("unexpected name: %s", calls[0].Name)
+	}
+	if calls[0].Input["command"] != "cd scripts && python check_syntax.py example.py" {
+		t.Fatalf("unexpected command arg: %v", calls[0].Input["command"])
+	}
+}
+
+func TestParseTextKVToolCalls_Multiple(t *testing.T) {
+	text := `
+function.name: read_file
+function.arguments: {
+	"path": "abc.txt"
+}
+
+function.name: bash
+function.arguments: {"command": "ls"}
+`
+	calls := ParseToolCalls(text, []string{"read_file", "bash"})
+	if len(calls) != 2 {
+		t.Fatalf("expected 2 calls, got %d", len(calls))
+	}
+	if calls[0].Name != "read_file" {
+		t.Fatalf("unexpected 1st name: %s", calls[0].Name)
+	}
+	if calls[1].Name != "bash" {
+		t.Fatalf("unexpected 2nd name: %s", calls[1].Name)
+	}
+}
+
+func TestParseTextKVToolCalls_Standalone(t *testing.T) {
+	text := "function.name: read_file\nfunction.arguments: {\"path\":\"README.md\"}"
+	calls := ParseStandaloneToolCalls(text, []string{"read_file"})
+	if len(calls) != 1 {
+		t.Fatalf("expected 1 call, got %d", len(calls))
+	}
+	if calls[0].Name != "read_file" {
+		t.Fatalf("unexpected name: %s", calls[0].Name)
+	}
+}

internal/util/util_edge_test.go

@@ -409,8 +409,8 @@ func TestParseToolCallsWithFunctionWrapper(t *testing.T) {
 func TestParseStandaloneToolCallsFencedCodeBlock(t *testing.T) {
 	fenced := "Here's an example:\n```json\n{\"tool_calls\":[{\"name\":\"search\",\"input\":{\"q\":\"go\"}}]}\n```\nDon't execute this."
 	calls := ParseStandaloneToolCalls(fenced, []string{"search"})
-	if len(calls) != 0 {
-		t.Fatalf("expected fenced code block ignored, got %d calls", len(calls))
+	if len(calls) != 1 {
+		t.Fatalf("expected fenced code block to be parsed, got %d calls", len(calls))
 	}
 }
 

internal/version/version.go

@@ -0,0 +1,185 @@
+package version
+
+import (
+	"os"
+	"path/filepath"
+	"runtime"
+	"strconv"
+	"strings"
+	"sync"
+)
+
+// BuildVersion can be injected at build time via -ldflags.
+// In release builds it should come from Git tag (e.g. v2.3.5).
+var BuildVersion = ""
+
+var (
+	currentOnce sync.Once
+	currentVal  string
+	sourceVal   string
+)
+
+func Current() (value string, source string) {
+	currentOnce.Do(func() {
+		if build := strings.TrimSpace(BuildVersion); build != "" {
+			currentVal = normalize(build)
+			sourceVal = "build-ldflags"
+			return
+		}
+		if fv := readVersionFile(); fv != "" {
+			currentVal = normalize(fv)
+			sourceVal = "file:VERSION"
+			return
+		}
+
+		if vv := versionFromVercelEnv(); vv != "" {
+			currentVal = vv
+			sourceVal = "env:vercel"
+			return
+		}
+		currentVal = "dev"
+		sourceVal = "default"
+	})
+	return currentVal, sourceVal
+}
+
+func readVersionFile() string {
+	candidates := []string{"VERSION"}
+	if wd, err := os.Getwd(); err == nil {
+		candidates = append(candidates, filepath.Join(wd, "VERSION"))
+	}
+	if _, file, _, ok := runtime.Caller(0); ok {
+		repoRoot := filepath.Clean(filepath.Join(filepath.Dir(file), "../.."))
+		candidates = append(candidates, filepath.Join(repoRoot, "VERSION"))
+	}
+	seen := map[string]struct{}{}
+	for _, c := range candidates {
+		c = filepath.Clean(strings.TrimSpace(c))
+		if c == "" {
+			continue
+		}
+		if _, ok := seen[c]; ok {
+			continue
+		}
+		seen[c] = struct{}{}
+		b, err := os.ReadFile(c)
+		if err != nil {
+			continue
+		}
+		if v := strings.TrimSpace(string(b)); v != "" {
+			return v
+		}
+	}
+	return ""
+}
+
+func normalize(v string) string {
+	v = strings.TrimSpace(v)
+	if v == "" {
+		return ""
+	}
+	return strings.TrimPrefix(v, "v")
+}
+
+func Tag(v string) string {
+	v = normalize(v)
+	if v == "" || v == "dev" {
+		return v
+	}
+	if v[0] < '0' || v[0] > '9' {
+		return v
+	}
+	return "v" + v
+}
+
+func versionFromVercelEnv() string {
+	if tag := normalize(strings.TrimSpace(os.Getenv("VERCEL_GIT_COMMIT_TAG"))); tag != "" {
+		return tag
+	}
+	ref := strings.TrimSpace(os.Getenv("VERCEL_GIT_COMMIT_REF"))
+	sha := strings.TrimSpace(os.Getenv("VERCEL_GIT_COMMIT_SHA"))
+	if len(sha) > 7 {
+		sha = sha[:7]
+	}
+	ref = sanitizeVersionLabel(ref)
+	sha = sanitizeVersionLabel(sha)
+	if ref == "" && sha == "" {
+		return ""
+	}
+	if ref != "" && sha != "" {
+		return "preview-" + ref + "." + sha
+	}
+	if ref != "" {
+		return "preview-" + ref
+	}
+	return "preview-" + sha
+}
+
+func sanitizeVersionLabel(in string) string {
+	in = strings.TrimSpace(strings.ToLower(in))
+	if in == "" {
+		return ""
+	}
+	var b strings.Builder
+	b.Grow(len(in))
+	prevDash := false
+	for i := 0; i < len(in); i++ {
+		c := in[i]
+		if (c >= 'a' && c <= 'z') || (c >= '0' && c <= '9') {
+			b.WriteByte(c)
+			prevDash = false
+			continue
+		}
+		if !prevDash {
+			b.WriteByte('-')
+			prevDash = true
+		}
+	}
+	out := strings.Trim(b.String(), "-")
+	return out
+}
+
+func Compare(a, b string) int {
+	pa := parse(normalize(a))
+	pb := parse(normalize(b))
+	for i := 0; i < 3; i++ {
+		if pa[i] < pb[i] {
+			return -1
+		}
+		if pa[i] > pb[i] {
+			return 1
+		}
+	}
+	return 0
+}
+
+func parse(v string) [3]int {
+	var out [3]int
+	parts := strings.SplitN(v, ".", 4)
+	for i := 0; i < 3 && i < len(parts); i++ {
+		n := readLeadingInt(parts[i])
+		out[i] = n
+	}
+	return out
+}
+
+func readLeadingInt(s string) int {
+	s = strings.TrimSpace(s)
+	if s == "" {
+		return 0
+	}
+	i := 0
+	for ; i < len(s); i++ {
+		if s[i] < '0' || s[i] > '9' {
+			break
+		}
+	}
+	if i == 0 {
+		return 0
+	}
+	n, err := strconv.Atoi(s[:i])
+	if err != nil {
+		return 0
+	}
+	return n
+}

internal/version/version_test.go

@@ -0,0 +1,39 @@
+package version
+
+import "testing"
+
+func TestNormalizeAndTag(t *testing.T) {
+	if got := normalize("v2.3.5"); got != "2.3.5" {
+		t.Fatalf("normalize failed: %q", got)
+	}
+	if got := Tag("2.3.5"); got != "v2.3.5" {
+		t.Fatalf("tag failed: %q", got)
+	}
+}
+
+func TestCompare(t *testing.T) {
+	if Compare("2.3.5", "2.3.5") != 0 {
+		t.Fatal("expected equal")
+	}
+	if Compare("2.3.5", "2.3.6") >= 0 {
+		t.Fatal("expected less")
+	}
+	if Compare("v2.10.0", "2.3.9") <= 0 {
+		t.Fatal("expected greater")
+	}
+}
+
+func TestTagKeepsPreviewStyle(t *testing.T) {
+	if got := Tag("preview-dev.abcd123"); got != "preview-dev.abcd123" {
+		t.Fatalf("expected preview tag unchanged, got %q", got)
+	}
+}
+
+func TestVersionFromVercelEnv(t *testing.T) {
+	t.Setenv("VERCEL_GIT_COMMIT_TAG", "")
+	t.Setenv("VERCEL_GIT_COMMIT_REF", "dev")
+	t.Setenv("VERCEL_GIT_COMMIT_SHA", "abcdef123456")
+	if got := versionFromVercelEnv(); got != "preview-dev.abcdef1" {
+		t.Fatalf("unexpected vercel preview version: %q", got)
+	}
+}

plans/refactor-baseline.md

@@ -1,32 +0,0 @@
-# DS2API Refactor Baseline (Historical Snapshot)
-
-- Snapshot time: `2026-02-22T08:53:54Z`
-- Snapshot branch: `dev`
-- Snapshot HEAD: `5d3989a`
-- Scope: backend + node api + webui large-file decoupling (no behavior change)
-
-## Gate Commands
-
-1. `./tests/scripts/run-unit-all.sh`
-   - Result: PASS
-   - Includes:
-     - `go test ./...`
-     - `node --test api/helpers/stream-tool-sieve.test.js api/chat-stream.test.js api/compat/js_compat_test.js`
-2. `npm --prefix webui run build`
-   - Result: PASS
-3. `./tests/scripts/check-refactor-line-gate.sh`
-   - Result: PASS (`checked=131 missing=0 over_limit=0`)
-4. Stage gates (1-5) replay:
-   - `go test ./internal/config ./internal/admin ./internal/account ./internal/deepseek ./internal/format/openai` -> PASS
-   - `go test ./internal/adapter/openai ./internal/util ./internal/sse ./internal/compat` -> PASS
-   - `go test ./internal/adapter/claude ./internal/adapter/gemini ./internal/config` -> PASS
-   - `go test ./internal/testsuite ./cmd/ds2api-tests` -> PASS
-   - `node --test api/helpers/stream-tool-sieve.test.js api/chat-stream.test.js api/compat/js_compat_test.js` -> PASS
-5. Final full regression:
-   - `go test ./... -count=1` -> PASS
-
-## Notes
-
-- This file records a historical baseline for refactor process tracking.
-- It is not intended to represent the current repository HEAD.
-- Frontend manual smoke for phase 6 still requires human execution and sign-off.

plans/refactor-line-gate-targets.txt

@@ -1,6 +1,8 @@
 # Line gate targets for large-file decoupling refactor.
-# Default limit: 300 lines
+# 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

plans/refactor-line-gate.md

@@ -1,21 +0,0 @@
-# Refactor Line Gate
-
-## Rules
-
-1. Production file default upper bound: `<= 300` lines.
-2. Entry/facade files upper bound: `<= 120` lines.
-3. Scope is limited to target files in `plans/refactor-line-gate-targets.txt`.
-4. Test files are out of scope for this gate.
-
-## Command
-
-```bash
-./tests/scripts/check-refactor-line-gate.sh
-```
-
-## Naming Note
-
-- Original split plan used `internal/admin/handler_accounts_test.go` for account probing logic.
-- In Go, `*_test.go` files are test-only compilation units and cannot host production handlers.
-- The production file is implemented as `internal/admin/handler_accounts_testing.go`.
-

tests/compat/expected/toolcalls_allowlist_empty.json

@@ -1,3 +1,8 @@
 {
-  "calls": []
-}
+  "calls": [],
+  "sawToolCallSyntax": true,
+  "rejectedByPolicy": true,
+  "rejectedToolNames": [
+    "unknown_tool"
+  ]
+}

tests/compat/expected/toolcalls_case_insensitive_canonical.json

@@ -6,5 +6,8 @@
         "path": "README.MD"
       }
     }
-  ]
-}
+  ],
+  "sawToolCallSyntax": true,
+  "rejectedByPolicy": false,
+  "rejectedToolNames": []
+}

tests/compat/expected/toolcalls_fenced_json.json

@@ -1,3 +1,13 @@
 {
-  "calls": []
+  "calls": [
+    {
+      "name": "read_file",
+      "input": {
+        "path": "README.MD"
+      }
+    }
+  ],
+  "sawToolCallSyntax": true,
+  "rejectedByPolicy": false,
+  "rejectedToolNames": []
 }

tests/compat/expected/toolcalls_function_call_tag.json

@@ -0,0 +1,13 @@
+{
+  "calls": [
+    {
+      "name": "read_file",
+      "input": {
+        "path": "README.MD"
+      }
+    }
+  ],
+  "sawToolCallSyntax": true,
+  "rejectedByPolicy": false,
+  "rejectedToolNames": []
+}

tests/compat/expected/toolcalls_invoke_attr.json

@@ -0,0 +1,13 @@
+{
+  "calls": [
+    {
+      "name": "read_file",
+      "input": {
+        "path": "README.MD"
+      }
+    }
+  ],
+  "sawToolCallSyntax": true,
+  "rejectedByPolicy": false,
+  "rejectedToolNames": []
+}

tests/compat/expected/toolcalls_loose_normalize.json

@@ -0,0 +1,13 @@
+{
+  "calls": [
+    {
+      "name": "read_file",
+      "input": {
+        "path": "README.MD"
+      }
+    }
+  ],
+  "sawToolCallSyntax": true,
+  "rejectedByPolicy": false,
+  "rejectedToolNames": []
+}

tests/compat/expected/toolcalls_namespace_tail_normalize.json

@@ -0,0 +1,13 @@
+{
+  "calls": [
+    {
+      "name": "read_file",
+      "input": {
+        "path": "README.MD"
+      }
+    }
+  ],
+  "sawToolCallSyntax": true,
+  "rejectedByPolicy": false,
+  "rejectedToolNames": []
+}

tests/compat/expected/toolcalls_standalone_fenced_example.json

@@ -1,3 +1,13 @@
 {
-  "calls": []
+  "calls": [
+    {
+      "name": "read_file",
+      "input": {
+        "path": "README.MD"
+      }
+    }
+  ],
+  "sawToolCallSyntax": true,
+  "rejectedByPolicy": false,
+  "rejectedToolNames": []
 }

tests/compat/expected/toolcalls_standalone_mixed_prose.json

@@ -1,3 +1,13 @@
 {
-  "calls": []
+  "calls": [
+    {
+      "name": "read_file",
+      "input": {
+        "path": "README.MD"
+      }
+    }
+  ],
+  "sawToolCallSyntax": true,
+  "rejectedByPolicy": false,
+  "rejectedToolNames": []
 }

tests/compat/expected/toolcalls_standalone_pure.json

@@ -6,5 +6,8 @@
         "path": "README.MD"
       }
     }
-  ]
-}
+  ],
+  "sawToolCallSyntax": true,
+  "rejectedByPolicy": false,
+  "rejectedToolNames": []
+}