Council fallback to mock with annotated rationale on real-client failure

If the HF inference endpoint is cold-starting (503), unauthenticated, or
otherwise erroring, the React UI was rendering empty specialist panels.
Now: catch the exception, run MockLLMClient as a deterministic fallback,
and append '[endpoint unavailable: <ErrorType>]' to each rationale so the
failure is visible at a glance — judges always see content, debugging is
trivial.

app/api/council.py

@@ -4,10 +4,17 @@ current env state and returns the resulting CouncilDecision.
 Picks a real OpenAI-compatible client (HF Inference Endpoint, vLLM, etc.)
 when ``API_BASE_URL`` is set, otherwise falls back to ``MockLLMClient``
 so tests and CPU-only laptops still work without API keys / GPU.
+
+If the real endpoint errors (cold-start 503, timeout, unauthenticated),
+we still return a CouncilDecision — generated by the deterministic mock —
+so the React UI always shows specialist content instead of an empty panel.
+The fallback decision's rationale prefixes carry the underlying error so
+judges (and you) can see what went wrong.
 """
 
 import logging
 import os
+import traceback
 
 from fastapi import APIRouter, HTTPException
 
@@ -29,7 +36,8 @@ def _make_client():
     return MockLLMClient()
 
 
-_council = Council(client=_make_client(), use_cache=True)
+_real_council = Council(client=_make_client(), use_cache=True)
+_mock_council = Council(client=MockLLMClient(), use_cache=False)
 
 
 @router.get("/council", response_model=CouncilDecision)
@@ -42,4 +50,18 @@ async def get_council() -> CouncilDecision:
             status_code=409,
             detail="env not initialized or episode complete — call /reset first",
         )
-    return await _council.run_async(env.current_observation())
+
+    obs = env.current_observation()
+    try:
+        return await _real_council.run_async(obs)
+    except Exception as e:
+        # Endpoint cold-start, 503, network — fall back to mock so the UI
+        # always renders. Annotate the rationale so the failure is visible.
+        logger.warning("Real council failed (%s); falling back to mock.", e)
+        logger.debug(traceback.format_exc())
+        decision = await _mock_council.run_async(obs)
+        suffix = f" [endpoint unavailable: {type(e).__name__}]"
+        for v in decision.votes:
+            v.rationale = (v.rationale or "") + suffix
+        decision.rationale = (decision.rationale or "") + suffix
+        return decision