fix(chat): retry silent rounds with think=True before falling back

Qwen3's tool-call tokens sometimes fail to parse into either content
or tool_calls, burning output tokens and producing empty bubbles.
Detect the signature within a round (empty content, no tool calls,
eval_count > 0) and re-run the same round once with reasoning mode
enabled, which emits more reliable output. The post-loop fallback
remains as the final catch.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/fabledassistant/services/generation_task.py

@@ -410,7 +410,13 @@ async def run_generation(
             buf.append_event("status", {"status": "Generating response..." if _round == 0 else "Composing response..."})
             t_stream = time.monotonic()
 
-            async for chunk in _stream_with_retry(messages, model, tools, think, num_ctx=num_ctx):
+            round_content_start = len(buf.content_so_far)
+            round_tokens_start = timing.get("output_tokens") or 0
+            effective_think = think
+            retried_with_think = False
+
+            while True:
+                async for chunk in _stream_with_retry(messages, model, tools, effective_think, num_ctx=num_ctx):
                     if buf.cancel_event.is_set():
                         cancelled = True
                         break
@@ -508,6 +514,31 @@ async def run_generation(
                             all_tool_calls.append(tool_record)
                             buf.append_event("tool_call", {"tool_call": tool_record})
 
+                if cancelled:
+                    break
+
+                round_content_added = len(buf.content_so_far) - round_content_start
+                round_tokens_added = (timing.get("output_tokens") or 0) - round_tokens_start
+                if (
+                    not round_tool_calls
+                    and round_content_added == 0
+                    and round_tokens_added > 0
+                    and not effective_think
+                    and not retried_with_think
+                ):
+                    # Silent round: qwen3's tool-call tokens sometimes aren't
+                    # parsed into content or tool_calls. Retry once with
+                    # think=True — reasoning mode produces more reliable output.
+                    logger.warning(
+                        "Silent round %d for conv %d (tokens=%d); retrying with think=True",
+                        _round, conv_id, round_tokens_added,
+                    )
+                    buf.append_event("status", {"status": "Retrying with reasoning enabled..."})
+                    effective_think = True
+                    retried_with_think = True
+                    continue
+                break
+
             timing["generation_ms"] = int((time.monotonic() - t_stream) * 1000)
 
             if cancelled: