fix(startup): keep the app bootable when config is partially broken

agent/registry.py

@@ -280,6 +280,8 @@ def get_agent_registry() -> AgentRegistry:
     """
 
     global _registry_instance, _registry_signature
+    import os
+
     from config import conf
 
     settings = conf()
@@ -288,7 +290,31 @@ def get_agent_registry() -> AgentRegistry:
         if _registry_pinned and _registry_instance is not None:
             return _registry_instance
         if _registry_instance is None or _registry_signature != signature:
-            _registry_instance = AgentRegistry.from_config(settings)
+            try:
+                _registry_instance = AgentRegistry.from_config(settings)
+            except AgentRegistryError:
+                # An invalid `agents`/`default_agent_id` block would otherwise
+                # bubble all the way up through load_config() and take the whole
+                # process down before the web console can bind — leaving a
+                # desktop user with no UI to fix the very config that is broken.
+                # Fall back to the default single agent so the app still starts;
+                # the console can then edit the bad block. Source deployments
+                # keep failing loudly so the developer sees the error at once.
+                if os.environ.get("COW_DESKTOP") != "1":
+                    raise
+                from common.log import logger
+
+                logger.error(
+                    "[AgentRegistry] invalid 'agents' config; ignoring it and "
+                    "starting with the default agent. Fix it in the console.",
+                    exc_info=True,
+                )
+                fallback = {
+                    k: v
+                    for k, v in dict(settings).items()
+                    if k not in ("agents", "default_agent_id")
+                }
+                _registry_instance = AgentRegistry.from_config(fallback)
             _registry_signature = signature
         return _registry_instance
 

app.py

@@ -82,7 +82,18 @@ class ChannelManager:
         with self._lock:
             channels = []
             for name in channel_names:
-                ch = channel_factory.create_channel(name)
+                # One misconfigured channel (e.g. wechatcom_app without its
+                # corp_id/token/aes_key) must not take the whole process down:
+                # instantiating it can raise while parsing config. The web
+                # console in particular has to come up so the desktop shell can
+                # surface the error and let the user fix the config. Skip the
+                # broken channel and keep the rest.
+                try:
+                    ch = channel_factory.create_channel(name)
+                except Exception as e:
+                    logger.error(f"[ChannelManager] Failed to create channel '{name}', skipping it: {e}")
+                    logger.exception(e)
+                    continue
                 ch.cloud_mode = self.cloud_mode
                 self._channels[name] = ch
                 channels.append((name, ch))

channel/wechatcom/wechatcomapp_channel.py

@@ -40,6 +40,25 @@ class WechatComAppChannel(ChatChannel):
         logger.info(
             "[wechatcom] Initializing WeCom app channel, corp_id: {}, agent_id: {}".format(self.corp_id, self.agent_id)
         )
+        # Fail fast with a readable message when the channel is enabled but its
+        # required credentials are missing. Otherwise WeChatCrypto concatenates
+        # a None aes_key with a string and raises an opaque
+        # "unsupported operand type(s) for +: 'NoneType' and 'str'".
+        missing = [
+            key
+            for key, val in (
+                ("wechatcom_corp_id", self.corp_id),
+                ("wechatcomapp_token", self.token),
+                ("wechatcomapp_aes_key", self.aes_key),
+            )
+            if not val
+        ]
+        if missing:
+            raise RuntimeError(
+                "[wechatcom] WeCom app channel is enabled but missing required config: "
+                + ", ".join(missing)
+                + ". Fill them in config.json or remove 'wechatcom_app' from channel_type."
+            )
         self.crypto = WeChatCrypto(self.token, self.aes_key, self.corp_id)
         self.client = WechatComAppClient(self.corp_id, self.secret)