Fix retry syntax: async generator can't use return, use break+try/finally

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

scripts/cache_aware_proxy.py

@@ -446,7 +446,7 @@ async def _handle_combined(api, req_data, token_ids, input_length, session_id, h
 
         async def generate():
             prefill_done = False
-            last_err = None
+            try:
                 for attempt in range(MAX_STREAM_RETRIES):
                     try:
                         async with inst.client.stream("POST", api, json=req_data, headers=headers) as resp:
@@ -459,15 +459,11 @@ async def _handle_combined(api, req_data, token_ids, input_length, session_id, h
                                     prefill_done = True
                                 yield chunk
                         inst.record_prefix(token_ids)
-                    return
-                except (httpx.ConnectError, httpx.RemoteProtocolError) as e:
-                    last_err = e
-                    if prefill_done:
+                        break
+                    except (httpx.ConnectError, httpx.RemoteProtocolError):
+                        if prefill_done or attempt >= MAX_STREAM_RETRIES - 1:
                             raise
-                    if attempt < MAX_STREAM_RETRIES - 1:
                         await asyncio.sleep(RETRY_DELAY_S)
-            if last_err:
-                raise last_err
             finally:
                 if not prefill_done:
                     inst.pending_prefill_tokens -= estimated_new