Feat/wecom channel

pyproject.toml

@@ -14,6 +14,7 @@ dependencies = [
     "textual>=3.0",
     "httpx>=0.27",
     "websockets>=12.0",
+    "wecom-aibot-sdk>=1.0.8",
 ]
 
 [project.scripts]

sample.env

@@ -9,3 +9,8 @@ ENABLE_TOOL_SEARCH=0
 CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1
 CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1
 CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1
+
+# WeCom (企业微信) Bot WebSocket channel
+# Run: uv run agent-box --wecom
+WECOM_BOT_ID=
+WECOM_SECRET=

src/agent_box/agents/claude_code.py

@@ -410,6 +410,12 @@ def _build_options(self) -> ClaudeAgentOptions:
         )
         if self.project.session_id:
             opts.resume = self.project.session_id
+        # Conditionally add wecom_mcp tool when WeCom channel is active
+        from ..tools.wecom_mcp import is_wecom_mcp_enabled
+        if is_wecom_mcp_enabled():
+            from ..tools.wecom_mcp import create_wecom_mcp_server
+            opts.mcp_servers = {"wecom_mcp": create_wecom_mcp_server()}
+            opts.allowed_tools = ["wecom_mcp"]
         return opts
 
     async def _ensure_client(self) -> ClaudeSDKClient:

src/agent_box/channels/wecom.py

@@ -0,0 +1,297 @@
+"""WeCom (企业微信) channel adapter — WebSocket long connection mode.
+
+Uses the official ``wecom-aibot-sdk`` Python SDK which connects via WebSocket
+to wss://openws.work.weixin.qq.com. Only requires two credentials:
+
+Required config:
+  WECOM_BOT_ID   — 机器人 ID (from 企业微信管理后台 → 智能机器人)
+  WECOM_SECRET   — 机器人 Secret
+"""
+
+from __future__ import annotations
+
+import asyncio
+import logging
+import mimetypes
+from pathlib import Path
+from typing import Any
+
+from anyio.abc import ObjectSendStream
+
+from wecom_aibot_sdk import WSClient, generate_req_id
+from wecom_aibot_sdk.types import WsFrame, MessageType as WeComMsgType
+
+from ..config import settings
+from ..models import IncomingMessage, MessageType, OutgoingMessage
+from .base import BaseChannel
+
+log = logging.getLogger(__name__)
+
+_IMAGE_EXTS = frozenset({".jpg", ".jpeg", ".png", ".gif", ".webp", ".bmp"})
+_AUDIO_EXTS = frozenset({".mp3", ".wav", ".ogg", ".flac", ".aac", ".m4a", ".amr"})
+
+
+def _detect_media_type(file_path: str) -> str:
+    ext = Path(file_path).suffix.lower()
+    if ext in _IMAGE_EXTS:
+        return "image"
+    if ext in _AUDIO_EXTS:
+        return "voice"
+    return "file"
+
+
+# ── Message parsing helpers ──────────────────────────────────────────────────
+
+def _parse_message_body(body: dict[str, Any]) -> tuple[str, list[str]]:
+    """Parse WeCom message body into (text, image_urls).
+
+    Returns extracted text and list of image URLs that need to be downloaded.
+    """
+    msgtype = body.get("msgtype", "")
+    text_parts: list[str] = []
+    image_urls: list[str] = []
+
+    if msgtype == "mixed" and body.get("mixed"):
+        # 图文混排消息
+        for item in body["mixed"].get("msg_item", []):
+            if item.get("msgtype") == "text" and item.get("text", {}).get("content"):
+                text_parts.append(item["text"]["content"])
+            elif item.get("msgtype") == "image" and item.get("image", {}).get("url"):
+                image_urls.append(item["image"]["url"])
+    else:
+        # 单条消息
+        if body.get("text", {}).get("content"):
+            text_parts.append(body["text"]["content"])
+        if msgtype == "voice" and body.get("voice", {}).get("content"):
+            # 语音转文字
+            text_parts.append(body["voice"]["content"])
+        if body.get("image", {}).get("url"):
+            image_urls.append(body["image"]["url"])
+
+    # 处理引用消息
+    quote = body.get("quote")
+    if quote:
+        if quote.get("msgtype") == "text" and quote.get("text", {}).get("content"):
+            if not text_parts:
+                text_parts.append(quote["text"]["content"])
+        elif quote.get("msgtype") == "voice" and quote.get("voice", {}).get("content"):
+            if not text_parts:
+                text_parts.append(quote["voice"]["content"])
+        elif quote.get("msgtype") == "image" and quote.get("image", {}).get("url"):
+            image_urls.append(quote["image"]["url"])
+
+    text = "\n".join(text_parts).strip()
+
+    # 对于纯媒体消息（无文本），生成描述
+    if not text and (image_urls or msgtype in ("file", "video")):
+        label = {"image": "图片", "voice": "语音", "video": "视频", "file": "文件"}.get(msgtype, "文件")
+        text = f"[用户发送了{label}]"
+
+    return text, image_urls
+
+
+# ── WecomChannel ─────────────────────────────────────────────────────────────
+
+class WecomChannel(BaseChannel):
+    """WeCom Bot WebSocket channel using official wecom-aibot-sdk."""
+
+    def __init__(self, send_stream: ObjectSendStream[IncomingMessage]) -> None:
+        super().__init__(send_stream)
+        self._client: WSClient | None = None
+        self._download_dir = settings.config_dir / "channels" / "wecom" / "downloads"
+        self._download_dir.mkdir(parents=True, exist_ok=True)
+
+    def _check_config(self) -> bool:
+        missing = [
+            name for name, val in [
+                ("WECOM_BOT_ID", settings.wecom_bot_id),
+                ("WECOM_SECRET", settings.wecom_secret),
+            ] if not val
+        ]
+        if missing:
+            log.error("WeCom channel: missing config: %s", ", ".join(missing))
+            return False
+        return True
+
+    async def start(self) -> None:
+        if not self._check_config():
+            await self.send_stream.aclose()
+            return
+
+        self._client = WSClient(
+            bot_id=settings.wecom_bot_id,
+            secret=settings.wecom_secret,
+            heartbeat_interval=30000,
+            max_reconnect_attempts=10,
+            max_auth_failure_attempts=5,
+        )
+
+        # Register event handlers
+        self._client.on("connected", self._on_connected)
+        self._client.on("authenticated", self._on_authenticated)
+        self._client.on("disconnected", self._on_disconnected)
+        self._client.on("error", self._on_error)
+        self._client.on("message", self._on_message)
+        self._client.on("event", self._on_event)
+
+        try:
+            await self._client.connect()
+            # Keep running until cancelled
+            while True:
+                await asyncio.sleep(1)
+        except asyncio.CancelledError:
+            pass
+        except Exception:
+            log.exception("wecom: unexpected error in start loop")
+        finally:
+            if self._client:
+                await self._client.disconnect()
+            await self.send_stream.aclose()
+
+    # ── Event handlers ────────────────────────────────────────────────
+
+    def _on_connected(self) -> None:
+        log.info("wecom: WebSocket connected")
+
+    def _on_authenticated(self) -> None:
+        log.info("wecom: authenticated successfully")
+        # Expose WSClient to the wecom_mcp tool
+        from ..tools.wecom_mcp import set_ws_client
+        set_ws_client(self._client)
+
+    def _on_disconnected(self, reason: str) -> None:
+        log.warning("wecom: disconnected: %s", reason)
+        from ..tools.wecom_mcp import set_ws_client
+        set_ws_client(None)
+
+    def _on_error(self, error: Exception) -> None:
+        log.error("wecom: error: %s", error)
+
+    async def _on_message(self, frame: WsFrame) -> None:
+        """Handle incoming message callback."""
+        body = frame.get("body") or {}
+        msgtype = body.get("msgtype", "")
+        from_user = body.get("from", {}).get("userid", "")
+        chat_id = body.get("chatid") or from_user
+
+        if not from_user:
+            return
+
+        text, image_urls = _parse_message_body(body)
+
+        # Download images to local files
+        if image_urls:
+            for url in image_urls:
+                aes_key = body.get("image", {}).get("aeskey")
+                local_path = await self._download_media(url, aes_key)
+                if local_path:
+                    text += f"\n文件路径: {local_path}"
+
+        if not text:
+            return
+
+        log.debug("wecom: inbound from %s: %s", from_user, text[:80])
+        await self.send_stream.send(
+            IncomingMessage(
+                text=text,
+                user_id=from_user,
+                channel="wecom",
+                raw={
+                    "frame": frame,
+                    "chat_id": chat_id,
+                    "chat_type": body.get("chattype", "single"),
+                },
+            )
+        )
+
+    async def _on_event(self, frame: WsFrame) -> None:
+        """Handle event callbacks (template card clicks, etc.)."""
+        body = frame.get("body") or {}
+        event = body.get("event", {})
+        event_type = event.get("eventtype", "")
+
+        # For now, log events but don't route them
+        log.debug("wecom: event received: type=%s", event_type)
+
+    # ── Media download ────────────────────────────────────────────────
+
+    async def _download_media(self, url: str, aes_key: str | None = None) -> str | None:
+        """Download and optionally decrypt a media file."""
+        if not self._client:
+            return None
+        try:
+            result = await self._client.download_file(url, aes_key)
+            buffer: bytes = result["buffer"]
+            filename: str | None = result.get("filename")
+
+            if not filename:
+                # Guess extension from content
+                import magic  # noqa: F401 - optional
+                ext = ".bin"
+                filename = f"media_{generate_req_id('dl')}{ext}"
+
+            dest = self._download_dir / filename
+            dest.write_bytes(buffer)
+            return str(dest)
+        except ImportError:
+            # No python-magic, use a generic extension
+            pass
+        except Exception:
+            log.exception("wecom: failed to download media from %s", url)
+        return None
+
+    # ── Outbound reply ────────────────────────────────────────────────
+
+    async def send_reply(self, msg: OutgoingMessage) -> None:
+        if not self._client or not self._client.is_connected:
+            log.warning("wecom: cannot send reply, not connected")
+            return
+
+        if msg.type != MessageType.text:
+            return
+
+        raw = msg.data or {}
+        chat_id = raw.get("chat_id") or msg.user_id
+        file_path: str | None = raw.get("file_path") or raw.get("image_path")
+
+        if file_path:
+            await self._send_file(chat_id, file_path)
+            return
+
+        # Send text as markdown via proactive send
+        try:
+            await self._client.send_message(chat_id, {
+                "msgtype": "markdown",
+                "markdown": {"content": msg.text},
+            })
+        except Exception:
+            log.exception("wecom: failed to send text to %s", chat_id)
+
+    async def _send_file(self, chat_id: str, file_path: str) -> None:
+        """Upload and send a file/image."""
+        if not self._client:
+            return
+
+        try:
+            p = Path(file_path)
+            if not p.exists():
+                log.error("wecom: file not found: %s", file_path)
+                return
+
+            media_type = _detect_media_type(file_path)
+            file_data = p.read_bytes()
+
+            upload_result = await self._client.upload_media(
+                file_data,
+                type=media_type,  # type: ignore[arg-type]
+                filename=p.name,
+            )
+            media_id = upload_result["media_id"]
+
+            await self._client.send_media_message(
+                chat_id,
+                media_type=media_type,  # type: ignore[arg-type]
+                media_id=media_id,
+            )
+        except Exception:
+            log.exception("wecom: failed to send file %s to %s", file_path, chat_id)

src/agent_box/config.py

@@ -20,6 +20,10 @@ class Settings(BaseSettings):
     qqbot_app_id: str = ""
     qqbot_client_secret: str = ""
 
+    # WeCom (企业微信) Bot WebSocket channel (long connection mode)
+    wecom_bot_id: str = ""
+    wecom_secret: str = ""
+
     # GLM (ZhipuAI) ASR — voice-to-text for audio attachments. Empty skips transcription.
     glm_api_key: str = ""
     glm_asr_model: str = "glm-asr-2512"