Skip to content

Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.

Studio: use the resolved public id in embeddings/completions API monitor rows (#9346)

Sourav Rajviacc0c89fcb1e
diff --git a/studio/backend/routes/inference.py b/studio/backend/routes/inference.py
index 31661520e2a..b988fb80ae7 100644
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@@ -32502,7 +32502,9 @@ async def openai_completions(request: Request, current_subject: str = Depends(ge
target_url = f"{llama_backend.base_url}/v1/completions"
is_stream = body.get("stream", False)
prompt_text = _flatten_monitor_prompt(body.get("prompt", ""))
- monitor_model = str(body.get("model") or _llama_public_model_id(llama_backend) or "default")
+ monitor_model = _monitor_active_model() or str(
+ body.get("model") or _llama_public_model_id(llama_backend) or "default"
+ )
monitor_id = api_monitor.start(
endpoint = request.url.path,
via_api_key = _request_used_api_key(request),
@@ -33283,12 +33285,15 @@ async def openai_embeddings(request: Request, current_subject: str = Depends(get
target_url = f"{llama_backend.base_url}/v1/embeddings"
prompt_text = _flatten_monitor_prompt(body.get("input", ""))
monitor_id = None
+ monitor_model = _monitor_active_model() or str(
+ body.get("model") or _llama_public_model_id(llama_backend) or "default"
+ )
if not getattr(request.state, "skip_api_monitor", False):
monitor_id = api_monitor.start(
endpoint = request.url.path,
via_api_key = _request_used_api_key(request),
method = request.method,
- model = str(body.get("model") or _llama_public_model_id(llama_backend) or "default"),
+ model = monitor_model,
prompt = prompt_text,
context_length = llama_backend.context_length,
subject = current_subject,
@@ -33301,7 +33306,7 @@ async def openai_embeddings(request: Request, current_subject: str = Depends(get
_client = _cancelable_nonstreaming_client()
_tracker = _TrackedCancel(
_cancel_event,
- model = str(body.get("model") or _llama_public_model_id(llama_backend) or "default"),
+ model = monitor_model,
kind = "embeddings",
)
_tracker.__enter__()
diff --git a/studio/backend/tests/test_openai_catalog.py b/studio/backend/tests/test_openai_catalog.py
index 6b0fdcf5334..7fae172c9db 100644
--- a/studio/backend/tests/test_openai_catalog.py
+++ b/studio/backend/tests/test_openai_catalog.py
@@ -491,3 +491,54 @@ def test_an_alias_for_the_resident_weights_is_not_listed_as_unloaded(monkeypatch
monkeypatch.setattr(resolver, "local_servable_model", lambda info: (True, ("Q4_K_M",)))
ids = {m["id"]: m for m in asyncio.run(inf._openai_catalog_objects())}
assert ids["publisher/Qwen3"]["loaded"] is True
+
+
+def test_embeddings_monitor_row_names_the_loaded_model_not_the_client_alias(monkeypatch):
+ from types import SimpleNamespace
+
+ import httpx
+ from core.inference.api_monitor import ApiMonitor
+
+ class Request:
+ state = SimpleNamespace()
+ url = SimpleNamespace(path = "/v1/embeddings")
+ method = "POST"
+
+ async def json(self):
+ return {"input": "hello", "model": "UNSLOTH-bge-m3"}
+
+ async def is_disconnected(self):
+ return False
+
+ class Client:
+ async def aclose(self):
+ return None
+
+ async def post(self, *_args, **_kwargs):
+ return httpx.Response(200, json = {"data": [{"embedding": [0.1]}], "usage": {}})
+
+ llama = SimpleNamespace(
+ is_loaded = True,
+ is_embedding_gguf = True,
+ base_url = "http://llama.test",
+ context_length = 4096,
+ model_identifier = "/cache/models--org--bge-m3/snapshots/abc",
+ hf_variant = "q8_0",
+ _openai_advertised_id = "bge-m3",
+ )
+ monitor = ApiMonitor(max_entries = 3)
+ monkeypatch.setattr(inf, "api_monitor", monitor)
+ monkeypatch.setattr(inf, "_cancelable_nonstreaming_client", Client)
+ monkeypatch.setattr(inf, "get_llama_cpp_backend", lambda: llama)
+ monkeypatch.setattr(inf, "_should_validate_before_switch", lambda: False)
+ monkeypatch.setattr(inf, "_names_studio_embedder", lambda _model: None)
+
+ async def _passthrough(request, _subject, **_kwargs):
+ return await request.json()
+
+ monkeypatch.setattr(inf, "_auto_switch_from_request_body", _passthrough)
+
+ response = asyncio.run(inf.openai_embeddings(Request(), current_subject = "test"))
+ assert response.status_code == 200
+ [entry] = monitor.snapshot()
+ assert entry["model"] == "bge-m3:q8_0"