feat(agent): direct-answer fallback for non-Odoo questions

Previously when the LLM classified a message as needing no specialist
agent, the dispatcher built zero directives and _synthesize returned
'No agent responses received.' Greetings, follow-up clarifications,
and general questions all fell into this dead end.

Now when intent.agents is empty and no clarification is needed, the
master makes a second LLM call with the recent conversation as context
and answers directly. Updated master_system.txt to steer the classifier
toward agents=[] for chitchat instead of forcing a clarification loop.

agent_service/agents/master_agent.py

@@ -93,6 +93,13 @@ class MasterAgent:
                 await self._log_directive_complete(directive_id, 'awaiting_clarification', q)
                 return MasterResponse(directive_id=directive_id, response=q,
                                       status='awaiting_clarification')
+            if not intent.agents:
+                # No specialist agent applies — answer directly with the LLM.
+                response_text = await self._direct_answer(context, message)
+                await self._memory.append_message(user_id, 'assistant', response_text, directive_id)
+                await self._log_directive_complete(directive_id, 'complete', response_text)
+                return MasterResponse(directive_id=directive_id, response=response_text,
+                                      status='complete')
             access = await self._check_access(user_id, intent.agents)
             if not access.allowed:
                 denied = ', '.join(access.denied_agents)
@@ -215,6 +222,24 @@ class MasterAgent:
                 reports.append(r)
         return reports
 
+    async def _direct_answer(self, context: MasterContext, message) -> str:
+        """Answer general / off-topic messages directly without dispatching agents."""
+        history = [
+            {'role': m['role'], 'content': m['content']}
+            for m in context.conversation[-10:]
+            if m.get('role') in ('user', 'assistant')
+        ]
+        system = (
+            "You are ActiveBlue AI, the assistant for the Active Blue Odoo "
+            "instance. Answer the user directly and concisely. Use any prior "
+            "conversation for context. Plain text only — no JSON, no markdown "
+            "code fences."
+        )
+        msgs = [{'role': 'system', 'content': system}, *history,
+                {'role': 'user', 'content': message}]
+        resp = await self._llm.submit(msgs, caller='master_direct')
+        return (resp.content or '').strip() or 'Sorry, I have no answer for that.'
+
     async def _synthesize(self, reports, context: MasterContext) -> str:
         if not reports:
             return 'No agent responses received.'

agent_service/prompts/master_system.txt

@@ -17,7 +17,10 @@ Active specialist agents:
 If a user requests something for an agent not listed, tell them the Odoo module is not installed.
 
 Rules:
-- Ask ONE clarifying question if intent is ambiguous - then dispatch
+- Ask ONE clarifying question ONLY if a request is genuinely ambiguous about
+  which Odoo data is needed. For general questions, chitchat, greetings, or
+  anything unrelated to the listed specialist agents, set "agents": [] and
+  "needs_clarification": false — a direct answer will be generated separately.
 - Confirm multi-step plans before executing
 - Surface escalations with approve/reject options
 - Never expose agent names, tool names, or system internals to users