Refactor: dev -> main

.claude/commands/wtnew.md

@@ -1,6 +1,6 @@
 # 创建 Worktree
 
-基于最新 `origin/main` 创建隔离的 worktree 开发环境。
+基于最新 `origin/dev` 创建隔离的 worktree 开发环境，并自动拉起一个并行工作的 Kitty + Codex 开发位。
 
 ## 参数
 
@@ -21,7 +21,7 @@ PROJECT_NAME=$(basename "$MAIN_REPO")
 git fetch origin
 ```
 
-确保基于最新的 `origin/main` 创建，避免从过时的 base 分叉。
+确保基于最新的 `origin/dev` 创建，避免从过时的 base 分叉。
 
 ## Step 2：启用 worktreeConfig
 
@@ -38,7 +38,7 @@ git config extensions.worktreeConfig true
 路径规则：`~/worktrees/<项目名>--<目录名>`（如 `~/worktrees/leon--feat-eval`）
 
 ```bash
-git worktree add "$HOME/worktrees/$PROJECT_NAME--<目录名>" -b $ARGUMENTS origin/main
+git worktree add "$HOME/worktrees/$PROJECT_NAME--<目录名>" -b $ARGUMENTS origin/dev
 ```
 
 - worktree 存放在 `~/worktrees/`，与主仓库完全隔离
@@ -163,16 +163,44 @@ ln -s "$MAIN_REPO/CLAUDE.local.md" CLAUDE.local.md 2>/dev/null
 输出：
 - worktree 路径
 - 分支名
+- base 分支（必须明确是 `origin/dev`）
 - 分配的端口（backend / frontend）
 - 自动生成的描述
 - `CLAUDE.local.md` 符号链接状态
 
-询问用户：是否在新 worktree 中打开新的 Claude 会话？
+## Step 9：自动拉起 Kitty + Codex 并行工作位
 
-如果是，用 osascript 打开新终端并启动 claude（**必须将路径替换为实际计算出的完整绝对路径，不得使用变量或占位符**）：
+不要再询问“是否打开新的 Claude 会话”。默认直接拉起一个新的 Kitty tab，并在里面启动 Codex。
+
+要求：
+- tab title 固定为 `dev-feature`
+- Codex 必须在新建好的 worktree 路径里启动
+- 必须用实际计算出的完整绝对路径，不得保留变量或占位符
+- 如果当前 shell 没有 `KITTY_LISTEN_ON`，要明确报错并停下，不要静默跳过
+
+执行命令（**必须将路径替换为实际计算出的完整绝对路径，不得使用变量或占位符**）：
 
 ```bash
-osascript -e 'tell app "Terminal" to do script "cd \"/Users/apple/worktrees/<项目名>--<目录名>\" && claude"'
+if [ -z "$KITTY_LISTEN_ON" ]; then
+  echo "❌ 错误：未设置 KITTY_LISTEN_ON，无法自动创建 dev-feature kitty tab"
+  exit 1
+fi
+
+kitty @ --to "$KITTY_LISTEN_ON" launch \
+  --type tab \
+  --tab-title "dev-feature" \
+  --title "dev-feature" \
+  zsh -lc 'cd "/Users/apple/worktrees/<项目名>--<目录名>" && codex --cd "/Users/apple/worktrees/<项目名>--<目录名>"'
 ```
 
-关键：`cd` 和 `claude` 必须写在 osascript 的 `do script` 字符串内部，不是写在外层 Bash 命令里。
+关键：
+- `cd` 和 `codex --cd ...` 必须写在新 tab 的命令字符串内部
+- `codex --cd` 和前面的 `cd` 都必须指向同一个实际 worktree 绝对路径
+- 不要退回 Terminal / osascript；这里的标准交互面就是 Kitty tab
+
+## Step 10：最终输出
+
+除了原有输出，再追加：
+- `Codex tab: dev-feature`
+- `Codex cwd: <worktree 绝对路径>`
+- 如果启动成功，明确说明“并行开发位已就绪”

.github/workflows/ci.yml

@@ -2,8 +2,9 @@ name: CI
 
 on:
   push:
-    branches: [main]
+    branches: [main, dev]
   pull_request:
+    branches: [main, dev]
 
 jobs:
   lint:
@@ -41,6 +42,10 @@ jobs:
       - name: Run tests
         # --maxfail=5: surface up to 5 failures per platform before stopping
         # e2e tests self-skip via skipif when provider secrets are absent
+        env:
+          SUPABASE_PUBLIC_URL: ${{ secrets.SUPABASE_PUBLIC_URL }}
+          LEON_SUPABASE_SERVICE_ROLE_KEY: ${{ secrets.SUPABASE_SERVICE_KEY }}
+          SUPABASE_ANON_KEY: ${{ secrets.SUPABASE_ANON_KEY }}
         run: uv run pytest tests/ --ignore=tests/test_e2e_providers.py --ignore=tests/test_sandbox_e2e.py --ignore=tests/test_daytona_e2e.py --ignore=tests/test_e2e_backend_api.py --ignore=tests/test_e2e_summary_persistence.py --ignore=tests/test_p3_e2e.py --maxfail=5 --timeout=60 -q
 
   frontend:

.github/workflows/deploy-staging.yml

@@ -7,6 +7,9 @@ name: Deploy Staging
 # Both update the staging apps to the target branch, then deploy.
 
 on:
+  push:
+    branches:
+      - pr188-agent-optimize
   pull_request:
     types: [labeled]
   workflow_dispatch:
@@ -23,43 +26,112 @@ jobs:
   deploy-staging:
     # For label trigger: only run when the label is exactly "deploy-staging"
     if: >
+      github.event_name == 'push' ||
       github.event_name == 'workflow_dispatch' ||
       (github.event_name == 'pull_request' && github.event.label.name == 'deploy-staging')
     runs-on: ubuntu-latest
+    env:
+      STAGING_STACK_UUID: fasbsube26s75ag6qus5bpi2
 
     steps:
       - name: Resolve target ref
         id: ref
         run: |
           if [ "${{ github.event_name }}" = "pull_request" ]; then
             echo "ref=${{ github.head_ref }}" >> "$GITHUB_OUTPUT"
+          elif [ "${{ github.event_name }}" = "push" ]; then
+            echo "ref=${{ github.ref_name }}" >> "$GITHUB_OUTPUT"
           else
             echo "ref=${{ inputs.ref }}" >> "$GITHUB_OUTPUT"
           fi
 
-      - name: Update staging backend branch
+      - name: Check out target ref
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ steps.ref.outputs.ref }}
+
+      - name: Resolve target commit
+        id: target
         run: |
-          curl -s -X PATCH "${{ secrets.COOLIFY_URL }}/api/v1/applications/${{ secrets.COOLIFY_BACKEND_STAGING_UUID }}" \
-            -H "Authorization: Bearer ${{ secrets.COOLIFY_TOKEN }}" \
-            -H "Content-Type: application/json" \
-            -d '{"git_branch": "${{ steps.ref.outputs.ref }}"}'
+          set -euo pipefail
+          echo "sha=$(git rev-parse HEAD)" >> "$GITHUB_OUTPUT"
+
+      - name: Assert repo staging compose contract
+        run: |
+          set -euo pipefail
+          grep -F "leon-home:/root/.leon" docker-compose.yml >/dev/null
+          grep -F "volumes:" docker-compose.yml >/dev/null
 
-      - name: Update staging frontend branch
+      - name: Update staging stack branch
         run: |
-          curl -s -X PATCH "${{ secrets.COOLIFY_URL }}/api/v1/applications/${{ secrets.COOLIFY_FRONTEND_STAGING_UUID }}" \
+          set -euo pipefail
+          body="$(curl -sS --fail-with-body -X PATCH "${{ secrets.COOLIFY_URL }}/api/v1/applications/${STAGING_STACK_UUID}" \
             -H "Authorization: Bearer ${{ secrets.COOLIFY_TOKEN }}" \
             -H "Content-Type: application/json" \
-            -d '{"git_branch": "${{ steps.ref.outputs.ref }}"}'
+            -d "{\"git_branch\": \"${{ steps.ref.outputs.ref }}\"}")"
+          echo "$body"
+          printf '%s' "$body" | jq -e --arg uuid "$STAGING_STACK_UUID" '.uuid == $uuid' >/dev/null
+
+      - name: Deploy staging stack
+        id: deploy
+        run: |
+          set -euo pipefail
+          body="$(curl -sS --fail-with-body "${{ secrets.COOLIFY_URL }}/api/v1/deploy?uuid=${STAGING_STACK_UUID}&force=false" \
+            -H "Authorization: Bearer ${{ secrets.COOLIFY_TOKEN }}")"
+          echo "$body"
+          printf '%s' "$body" | jq -e --arg uuid "$STAGING_STACK_UUID" '.deployments[0].resource_uuid == $uuid' >/dev/null
+          echo "deployment_uuid=$(printf '%s' "$body" | jq -r '.deployments[0].deployment_uuid')" >> "$GITHUB_OUTPUT"
+
+      - name: Wait for staging deployment
+        run: |
+          set -euo pipefail
+          deployment_uuid="${{ steps.deploy.outputs.deployment_uuid }}"
+          for _ in $(seq 1 60); do
+            body="$(curl -sS --fail-with-body "${{ secrets.COOLIFY_URL }}/api/v1/deployments/${deployment_uuid}" \
+              -H "Authorization: Bearer ${{ secrets.COOLIFY_TOKEN }}")"
+            status="$(printf '%s' "$body" | jq -r '.status')"
+            echo "deployment status: $status"
+            if [ "$status" = "finished" ]; then
+              exit 0
+            fi
+            if [ "$status" != "queued" ] && [ "$status" != "in_progress" ]; then
+              echo "$body"
+              exit 1
+            fi
+            sleep 10
+          done
+          echo "Timed out waiting for staging deployment ${deployment_uuid}"
+          exit 1
 
-      - name: Deploy backend to staging
+      - name: Verify Coolify staging contract
         run: |
-          curl -sX GET "${{ secrets.COOLIFY_URL }}/api/v1/deploy?uuid=${{ secrets.COOLIFY_BACKEND_STAGING_UUID }}&force=false" \
-            -H "Authorization: Bearer ${{ secrets.COOLIFY_TOKEN }}"
+          set -euo pipefail
+          body="$(curl -sS --fail-with-body "${{ secrets.COOLIFY_URL }}/api/v1/applications/${STAGING_STACK_UUID}" \
+            -H "Authorization: Bearer ${{ secrets.COOLIFY_TOKEN }}")"
+          echo "$body" | jq '{uuid,git_branch,docker_compose_location}'
+          printf '%s' "$body" | jq -e --arg ref "${{ steps.ref.outputs.ref }}" '.git_branch == $ref' >/dev/null
+          printf '%s' "$body" | jq -e '.docker_compose_raw | contains("leon-home:/root/.leon")' >/dev/null
+          printf '%s' "$body" | jq -e --arg volume "${STAGING_STACK_UUID}_leon-home:/root/.leon" '.docker_compose | contains($volume)' >/dev/null
+          printf '%s' "$body" | jq -e --arg sha "${{ steps.target.outputs.sha }}" '.docker_compose | contains($sha)' >/dev/null
 
-      - name: Deploy frontend to staging
+      - name: Verify staging health contract
         run: |
-          curl -sX GET "${{ secrets.COOLIFY_URL }}/api/v1/deploy?uuid=${{ secrets.COOLIFY_FRONTEND_STAGING_UUID }}&force=false" \
-            -H "Authorization: Bearer ${{ secrets.COOLIFY_TOKEN }}"
+          set -euo pipefail
+          for attempt in $(seq 1 18); do
+            status="$(curl -sS -o /tmp/staging-health.json -w '%{http_code}' "https://app.staging.mycel.nextmind.space/api/monitor/health")"
+            echo "health attempt ${attempt}: status=${status}"
+            if [ "$status" = "200" ]; then
+              body="$(cat /tmp/staging-health.json)"
+              echo "$body"
+              printf '%s' "$body" | jq -e '.db.path == "/root/.leon/sandbox.db"' >/dev/null
+              printf '%s' "$body" | jq -e '.db.exists == true' >/dev/null
+              exit 0
+            fi
+            cat /tmp/staging-health.json || true
+            sleep 10
+          done
+          echo "Staging health contract did not become ready in time"
+          exit 1
 
       - name: Comment on PR with staging URL
         if: github.event_name == 'pull_request'
@@ -70,5 +142,5 @@ jobs:
               issue_number: context.issue.number,
               owner: context.repo.owner,
               repo: context.repo.repo,
-              body: `🚀 **预发部署已触发**\n\n- 前端: https://app.staging.mycel.nextmind.space\n- 后端: https://api.staging.mycel.nextmind.space\n\n分支: \`${{ steps.ref.outputs.ref }}\``
+              body: `🚀 **预发部署已触发**\n\n- 共享 Staging: https://app.staging.mycel.nextmind.space\n- API（同域反代）: https://app.staging.mycel.nextmind.space/api\n\n分支: \`${{ steps.ref.outputs.ref }}\``
             })

.gitignore

@@ -102,6 +102,8 @@ worktrees/
 # Development artifacts — never commit
 docs/lessons/
 docs/plans/
+docs/superpowers/plans/
+docs/superpowers/specs/
 frontend/.vite/
 .playwright-cli/
 ops

Dockerfile

@@ -7,11 +7,13 @@ COPY --from=ghcr.io/astral-sh/uv:latest /uv /usr/local/bin/uv
 
 # Install dependencies (cached layer before source copy)
 COPY pyproject.toml uv.lock ./
-RUN uv sync --frozen --no-dev --no-install-project
+# @@@sandbox-sdk-image-parity - shared staging/provider inventory should reflect runtime truth,
+# not "SDK missing from image" accidents while config files are present.
+RUN uv sync --frozen --no-dev --extra sandbox --extra e2b --extra daytona --no-install-project
 
 # Copy source and install project
 COPY . .
-RUN uv sync --frozen --no-dev
+RUN uv sync --frozen --no-dev --extra sandbox --extra e2b --extra daytona
 
 ENV PATH="/app/.venv/bin:$PATH"
 

README.md

@@ -95,16 +95,18 @@ Full-featured web platform for managing and interacting with agents:
 
 ### Multi-Agent Communication
 
-Agents are first-class social entities. They can discover each other, send messages, and collaborate autonomously:
+Agents are first-class social entities. They can list chats, read messages, send messages, and collaborate autonomously:
 
 ```
 Member (template)
   └→ Entity (social identity — agents and humans both get one)
        └→ Thread (agent brain / conversation)
 ```
 
-- **`chat_send`**: Agent A messages Agent B; B responds autonomously
-- **`directory`**: Agents browse and discover other entities
+- **`list_chats`**: List active conversations with unread counts and participants
+- **`read_messages`**: Read message history before responding
+- **`send_message`**: Agent A messages Agent B; B responds autonomously
+- **`search_messages`**: Search message history across chats
 - **Real-time delivery**: SSE-based chat with typing indicators and read receipts
 
 Humans also have entities — agents can initiate conversations with humans, not just the other way around.

README.zh.md

@@ -95,16 +95,18 @@ cd frontend/app && npm run dev
 
 ### 多 Agent 通讯
 
-Agent 是一等公民的社交实体，可以互相发现、发送消息、自主协作：
+Agent 是一等公民的社交实体，可以列出对话、读取消息、发送消息、自主协作：
 
 ```
 Member（模板）
   └→ Entity（社交身份——Agent 和人类都有）
        └→ Thread（Agent 大脑 / 对话）
 ```
 
-- **`chat_send`**：Agent A 给 Agent B 发消息，B 自主回复
-- **`directory`**：Agent 浏览和发现其他实体
+- **`list_chats`**：列出活跃对话、未读数和参与者
+- **`read_messages`**：先读取消息历史，再决定如何回复
+- **`send_message`**：Agent A 给 Agent B 发消息，B 自主回复
+- **`search_messages`**：跨对话搜索消息历史
 - **实时投递**：基于 SSE 的聊天，支持输入提示和已读回执
 
 人类也有 Entity——Agent 可以主动找人类对话，而不只是被动响应。

backend/taskboard/_service_loader.py

@@ -0,0 +1,25 @@
+"""Typed task_service loader for taskboard surfaces."""
+
+from __future__ import annotations
+
+from typing import Any, Protocol, cast
+
+
+class TaskServiceProtocol(Protocol):
+    def list_tasks(self) -> list[dict[str, Any]]: ...
+    def get_task(self, task_id: str) -> dict[str, Any] | None: ...
+    def get_highest_priority_pending_task(self) -> dict[str, Any] | None: ...
+    def create_task(self, **fields: Any) -> dict[str, Any]: ...
+    def update_task(self, task_id: str, **fields: Any) -> dict[str, Any] | None: ...
+
+
+try:
+    from backend.web.services import task_service as _task_service
+except ImportError:
+    _task_service = None
+
+
+def require_task_service() -> TaskServiceProtocol:
+    if _task_service is None:
+        raise RuntimeError("backend.web.services.task_service is unavailable")
+    return cast(TaskServiceProtocol, _task_service)