fix LM caption: use poll response data directly, empty body from /job?result=1 was the bug

app.py

@@ -96,11 +96,11 @@ def _get_props():
 
 
 def _poll_job(job_id, timeout=600, progress_cb=None, cancel_check=None):
-    """Poll a job until done/error/timeout/cancelled. Returns (status, elapsed)."""
+    """Poll a job until done/error/timeout/cancelled. Returns (status, elapsed, data)."""
     t0 = time.time()
     while time.time() - t0 < timeout:
         if cancel_check and cancel_check():
-            return "cancelled", time.time() - t0
+            return "cancelled", time.time() - t0, None
         try:
             r = requests.get(f"{ACE_SERVER}/job", params={"id": job_id}, timeout=5)
             data = r.json()
@@ -108,11 +108,11 @@ def _poll_job(job_id, timeout=600, progress_cb=None, cancel_check=None):
             if progress_cb:
                 progress_cb(status, data)
             if status in ("done", "error"):
-                return status, time.time() - t0
+                return status, time.time() - t0, data
         except Exception:
             pass
         time.sleep(1)
-    return "timeout", time.time() - t0
+    return "timeout", time.time() - t0, None
 
 
 def _fetch_result(job_id, timeout=60):
@@ -146,28 +146,42 @@ def _caption_via_understand(audio_path, timeout=600, cancel_check=None):
         logger.warning("[Caption] %s: /understand submit failed: %s", fname, exc)
         return None
 
-    status, elapsed = _poll_job(job_id, timeout=timeout, cancel_check=cancel_check)
+    status, elapsed, poll_data = _poll_job(job_id, timeout=timeout, cancel_check=cancel_check)
     if status != "done":
         logger.warning("[Caption] %s: /understand -> %s (%.0fs)", fname, status, elapsed)
         return None
 
-    try:
-        r = _fetch_result(job_id, timeout=120)
-        if r.status_code != 200:
-            logger.warning("[Caption] %s: result fetch HTTP %d", fname, r.status_code)
-            return None
-        data = r.json()
-        logger.info("[Caption] %s: result keys=%s", fname, list(data.keys()) if isinstance(data, dict) else type(data).__name__)
-        if isinstance(data, dict):
-            if data.get("caption"):
-                return data
-            if data.get("result") and isinstance(data["result"], dict):
-                return data["result"]
-            logger.warning("[Caption] %s: no caption in result: %s", fname, str(data)[:200])
-        return None
-    except Exception as exc:
-        logger.warning("[Caption] %s: result parse failed: %s", fname, exc)
-        return None
+    # Result may be in the poll response itself or need separate fetch
+    result = None
+
+    # Check if poll response contains the result directly
+    if poll_data and isinstance(poll_data, dict):
+        if poll_data.get("caption"):
+            result = poll_data
+        elif poll_data.get("result") and isinstance(poll_data["result"], dict):
+            result = poll_data["result"]
+
+    # Fallback: try fetching result separately
+    if not result:
+        try:
+            r = _fetch_result(job_id, timeout=120)
+            if r.status_code == 200 and r.text.strip():
+                data = r.json()
+                if isinstance(data, dict):
+                    if data.get("caption"):
+                        result = data
+                    elif data.get("result") and isinstance(data["result"], dict):
+                        result = data["result"]
+        except Exception as exc:
+            logger.warning("[Caption] %s: result fetch failed: %s", fname, exc)
+
+    if result:
+        logger.info("[Caption] %s: got caption (%d chars)", fname, len(result.get("caption", "")))
+        return result
+
+    logger.warning("[Caption] %s: no caption in response. poll_keys=%s",
+                   fname, list(poll_data.keys()) if isinstance(poll_data, dict) else "N/A")
+    return None
 
 
 def _run_pipeline(caption, lyrics, bpm, duration, seed, steps, output_format,
@@ -209,7 +223,7 @@ def _run_pipeline(caption, lyrics, bpm, duration, seed, steps, output_format,
 
     if progress_cb:
         progress_cb("lm_poll", {"job_id": lm_job_id})
-    lm_status, lm_elapsed = _poll_job(lm_job_id, timeout=900)
+    lm_status, lm_elapsed, _ = _poll_job(lm_job_id, timeout=900)
     if lm_status != "done":
         raise RuntimeError(f"LM {lm_status} after {lm_elapsed:.0f}s")
 
@@ -234,7 +248,7 @@ def _run_pipeline(caption, lyrics, bpm, duration, seed, steps, output_format,
 
     if progress_cb:
         progress_cb("synth_poll", {"job_id": synth_job_id})
-    synth_status, synth_elapsed = _poll_job(synth_job_id, timeout=600)
+    synth_status, synth_elapsed, _ = _poll_job(synth_job_id, timeout=600)
     if synth_status != "done":
         raise RuntimeError(f"Synth {synth_status} after {synth_elapsed:.0f}s")