fix: run blocking requests/sleep in async routes via to_thread

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

CHANGELOG.md

@@ -9,6 +9,9 @@ Follows [Keep a Changelog](https://keepachangelog.com/en/1.0.0/) · versioned wi
 ### Added
 - Conversation Playground: a Thinking on/off switch next to the Language Model (default off, setting `conversation_thinking`). Off sends `chat_template_kwargs.enable_thinking=false` (plus `think:false` for Ollama) with each reply request and never touches the system prompt; the empty-reply error now says to turn Thinking off.
 
+### Fixed
+- Async routes no longer block the event loop on synchronous HTTP: seed proxy/samples, text refine/persona rewrite, voice matching (its retry back-off is now `asyncio.sleep`), character deep analysis, character image-from-URL, LLM/engine model listing, source audio import/proxy and ElevenLabs shared voices now run `requests` through `asyncio.to_thread` (same timeouts and responses).
+
 ### Changed
 - Conversation Playground: the Language Model base URL is now picked from a dropdown of the language models configured under Engines (Docker LLM containers incl. saved URL overrides, and custom LLM cards; running/stopped marked), with a "Custom…" option that reveals the free-text URL. Selecting an entry saves `conv_llm_url` and fetches the models.
 - Playwright specs now import `test` from `tests/_guard.js`, whose auto fixture aborts and fails any unmocked non-GET request to `/api/**` so a spec can no longer modify live settings. See `architecture/qa.md`.

routes/conversation.py

@@ -720,7 +720,8 @@ async def refine_text(request: Request):
         payload["model"] = model
 
     try:
-        resp = requests.post(
+        resp = await asyncio.to_thread(
+            requests.post,
             f"{llm_url}/chat/completions",
             json=payload,
             headers={"Authorization": f"Bearer {settings.get('llm_api_key') or 'sk-dummy-key'}"},
@@ -781,7 +782,8 @@ async def rewrite_with_persona(request: Request):
         payload["model"] = model
 
     try:
-        resp = requests.post(
+        resp = await asyncio.to_thread(
+            requests.post,
             f"{llm_url}/chat/completions",
             json=payload,
             headers={"Authorization": f"Bearer {settings.get('llm_api_key') or 'sk-dummy-key'}"},
@@ -1015,12 +1017,12 @@ async def match_characters_voices(request: Request):
     }
     if model:
         payload["model"] = model
-    import time as _time
     raw = None
     last_err: Exception | None = None
     for attempt in range(3):
         try:
-            resp = requests.post(
+            resp = await asyncio.to_thread(
+                requests.post,
                 f"{llm_url}/chat/completions", json=payload,
                 headers={"Authorization": f"Bearer {_settings.get('llm_api_key') or 'sk-dummy-key'}"}, timeout=600,
             )
@@ -1032,7 +1034,7 @@ async def match_characters_voices(request: Request):
             # llama-swap (and similar) often drop the first request while swapping/loading
             # the model — wait and retry rather than failing the cast.
             last_err = e
-            _time.sleep(4 + attempt * 3)
+            await asyncio.sleep(4 + attempt * 3)
         except Exception as e:
             last_err = e
             break
@@ -1670,7 +1672,8 @@ async def character_deep_analysis(request: Request):
     if model:
         payload["model"] = model
     try:
-        resp = requests.post(
+        resp = await asyncio.to_thread(
+            requests.post,
             f"{llm_url}/chat/completions", json=payload,
             headers={"Authorization": f"Bearer {_settings.get('llm_api_key') or 'sk-dummy-key'}"}, timeout=600,
         )
@@ -4372,12 +4375,15 @@ async def character_image_from_url(request: Request):
     if not url:
         raise HTTPException(400, "Image URL required")
     try:
-        resp = requests.get(url, timeout=30, headers={"User-Agent": "TTS-Voice-Creator/image-fetch"}, stream=True)
+        def _fetch() -> tuple[str, bytes]:
-        resp.raise_for_status()
+            resp = requests.get(url, timeout=30, headers={"User-Agent": "TTS-Voice-Creator/image-fetch"}, stream=True)
-        ctype = resp.headers.get("Content-Type", "")
+            resp.raise_for_status()
-        if not ctype.startswith("image/"):
+            ctype = resp.headers.get("Content-Type", "")
-            raise HTTPException(400, f"That URL didn't return an image (got {ctype or 'unknown content type'})")
+            if not ctype.startswith("image/"):
-        content = resp.raw.read(15 * 1024 * 1024, decode_content=True)
+                raise HTTPException(400, f"That URL didn't return an image (got {ctype or 'unknown content type'})")
+            return ctype, resp.raw.read(15 * 1024 * 1024, decode_content=True)
+
+        ctype, content = await asyncio.to_thread(_fetch)
         if not content:
             raise HTTPException(502, "Empty response from that URL")
         b64 = base64.b64encode(content).decode("ascii")
@@ -6076,12 +6082,12 @@ async def conversation_llm_models(
     key_to_use = provider_api_key or settings.get("llm_api_key") or "sk-dummy-key"
     try:
         base = _validate_http_url(base, allow_private=True)
-        r = requests.get(f"{base}/models", timeout=5, headers={"Authorization": f"Bearer {key_to_use}"})
+        r = await asyncio.to_thread(requests.get, f"{base}/models", timeout=5, headers={"Authorization": f"Bearer {key_to_use}"})
 
         # Fallback for gateways like LiteLLM that might prefer /models over /v1/models
         if r.status_code != 200 and base.endswith("/v1"):
             fallback_base = base[:-3]
-            r2 = requests.get(f"{fallback_base}/models", timeout=5, headers={"Authorization": f"Bearer {key_to_use}"})
+            r2 = await asyncio.to_thread(requests.get, f"{fallback_base}/models", timeout=5, headers={"Authorization": f"Bearer {key_to_use}"})
             if r2.status_code == 200:
                 r = r2
 

routes/docker.py

@@ -229,7 +229,7 @@ async def engine_models(url: str, type: str = "", provider_api_key: str = Header
     key_to_use = settings.get("llm_api_key") or provider_api_key or "sk-dummy-key"
     for endpoint in candidates:
         try:
-            r = requests.get(endpoint, timeout=6, headers={"Authorization": f"Bearer {key_to_use}", "User-Agent": "TTS-Voice-Creator/models"})
+            r = await asyncio.to_thread(requests.get, endpoint, timeout=6, headers={"Authorization": f"Bearer {key_to_use}", "User-Agent": "TTS-Voice-Creator/models"})
             if r.status_code >= 400:
                 last_err = f"HTTP {r.status_code} from {endpoint}"
                 continue

routes/sources.py

@@ -554,6 +554,22 @@ async def open_voice_source_search(source: str, q: str = "", lang: str = ""):
         raise HTTPException(e.status, str(e))
 
 
+def _download_audio_wav(audio_url: str, dest: Path) -> Path:
+    """Blocking: stream audio_url to dest (size-capped) and convert to 24k wav. Run via to_thread."""
+    with requests.get(audio_url, headers=_VOICE_SOURCE_HEADERS, timeout=30, stream=True) as r:
+        r.raise_for_status()
+        total = 0
+        with dest.open("wb") as f:
+            for chunk in r.iter_content(1024 * 1024):
+                if not chunk:
+                    continue
+                total += len(chunk)
+                if total > _MAX_UPLOAD_BYTES:
+                    raise HTTPException(413, "Downloaded audio is too large")
+                f.write(chunk)
+    return _to_wav_24k(dest)
+
+
 @router.post("/api/import-source-audio")
 async def import_source_audio(request: Request):
     data = await request.json()
@@ -564,18 +580,7 @@ async def import_source_audio(request: Request):
         suffix = ".bin"
     dest = TEMP_DIR / f"{uuid.uuid4().hex}{suffix}"
     try:
-        with requests.get(audio_url, headers=_VOICE_SOURCE_HEADERS, timeout=30, stream=True) as r:
+        wav = await asyncio.to_thread(_download_audio_wav, audio_url, dest)
-            r.raise_for_status()
-            total = 0
-            with dest.open("wb") as f:
-                for chunk in r.iter_content(1024 * 1024):
-                    if not chunk:
-                        continue
-                    total += len(chunk)
-                    if total > _MAX_UPLOAD_BYTES:
-                        raise HTTPException(413, "Downloaded audio is too large")
-                    f.write(chunk)
-        wav = _to_wav_24k(dest)
     except HTTPException:
         raise
     except Exception as e:
@@ -594,7 +599,7 @@ async def proxy_audio(url: str):
     if not any(domain == d or domain.endswith("." + d) for d in _PROXY_AUDIO_DOMAINS):
         raise HTTPException(403, f"Domain not in audio proxy allowlist: {parsed.netloc}")
     try:
-        resp = requests.get(url, headers=_VOICE_SOURCE_HEADERS, timeout=30, stream=True)
+        resp = await asyncio.to_thread(requests.get, url, headers=_VOICE_SOURCE_HEADERS, timeout=30, stream=True)
         resp.raise_for_status()
     except Exception as e:
         raise HTTPException(502, f"Proxy fetch failed: {e}")
@@ -638,18 +643,7 @@ async def quick_import_voice(request: Request):
         suffix = ".bin"
     dest = TEMP_DIR / f"{uuid.uuid4().hex}{suffix}"
     try:
-        with requests.get(audio_url, headers=_VOICE_SOURCE_HEADERS, timeout=30, stream=True) as r:
+        wav = await asyncio.to_thread(_download_audio_wav, audio_url, dest)
-            r.raise_for_status()
-            total = 0
-            with dest.open("wb") as f:
-                for chunk in r.iter_content(1024 * 1024):
-                    if not chunk:
-                        continue
-                    total += len(chunk)
-                    if total > _MAX_UPLOAD_BYTES:
-                        raise HTTPException(413, "Downloaded audio is too large")
-                    f.write(chunk)
-        wav = _to_wav_24k(dest)
         if source_kind:  # sample clip: trim to [start, start+length] (length 3..120 s)
             from pydub import AudioSegment
             try:
@@ -807,7 +801,8 @@ async def elevenlabs_shared_voices(request: Request):
     params.setdefault("page_size", "24")
     headers = {"xi-api-key": api_key} if api_key else {}
     try:
-        r = requests.get(
+        r = await asyncio.to_thread(
+            requests.get,
             "https://api.elevenlabs.io/v1/shared-voices",
             params=params,
             headers=headers,

routes/tts.py

@@ -1306,7 +1306,8 @@ async def tts_voice_seed(request: Request):
     settings = _load_settings()
     tts_base = _preview_backend_base_url(settings, "voice_clone").rstrip("/")
     try:
-        resp = requests.post(
+        resp = await asyncio.to_thread(
+            requests.post,
             f"{tts_base}/voice-seed",
             json=data,
             timeout=10,
@@ -1431,7 +1432,7 @@ async def list_seed_samples(voice_name: str):
     settings = _load_settings()
     tts_base = _preview_backend_base_url(settings, "voice_clone").rstrip("/")
     try:
-        resp = requests.get(f"{tts_base}/seed-samples/{voice_name}", timeout=5)
+        resp = await asyncio.to_thread(requests.get, f"{tts_base}/seed-samples/{voice_name}", timeout=5)
         resp.raise_for_status()
         return resp.json()
     except requests.exceptions.ConnectionError:
@@ -1446,7 +1447,7 @@ async def get_seed_sample(voice_name: str, seed: int):
     settings = _load_settings()
     tts_base = _preview_backend_base_url(settings, "voice_clone").rstrip("/")
     try:
-        resp = requests.get(f"{tts_base}/seed-sample/{voice_name}/{seed}", timeout=30, stream=True)
+        resp = await asyncio.to_thread(requests.get, f"{tts_base}/seed-sample/{voice_name}/{seed}", timeout=30)  # body is read inside the worker thread
         resp.raise_for_status()
         return Response(content=resp.content, media_type="audio/wav")
     except requests.exceptions.ConnectionError: