fix: update tests for OpenAI-style models endpoint (#4053)

The llama-stack-client now uses /`v1/openai/v1/models` which returns
OpenAI-compatible model objects with 'id' and 'custom_metadata' fields
instead of the Resource-style 'identifier' field. Updated api_recorder
to handle the new endpoint and modified tests to access model metadata
appropriately. Deleted stale model recordings for re-recording.

**NOTE: CI will be red on this one since it is dependent on
https://github.com/llamastack/llama-stack-client-python/pull/291/files
landing. I verified locally that it is green.**

src/llama_stack/core/ui/page/distribution/models.py

@@ -12,7 +12,7 @@ from llama_stack.core.ui.modules.api import llama_stack_api
 def models():
     # Models Section
     st.header("Models")
-    models_info = {m.identifier: m.to_dict() for m in llama_stack_api.client.models.list()}
+    models_info = {m.id: m.model_dump() for m in llama_stack_api.client.models.list()}
 
     selected_model = st.selectbox("Select a model", list(models_info.keys()))
     st.json(models_info[selected_model])

src/llama_stack/core/ui/page/playground/chat.py

@@ -12,7 +12,11 @@ from llama_stack.core.ui.modules.api import llama_stack_api
 with st.sidebar:
     st.header("Configuration")
     available_models = llama_stack_api.client.models.list()
-    available_models = [model.identifier for model in available_models if model.model_type == "llm"]
+    available_models = [
+        model.id
+        for model in available_models
+        if model.custom_metadata and model.custom_metadata.get("model_type") == "llm"
+    ]
     selected_model = st.selectbox(
         "Choose a model",
         available_models,

src/llama_stack/testing/api_recorder.py

@@ -156,7 +156,7 @@ def normalize_inference_request(method: str, url: str, headers: dict[str, Any],
     }
 
     # Include test_id for isolation, except for shared infrastructure endpoints
-    if parsed.path not in ("/api/tags", "/v1/models"):
+    if parsed.path not in ("/api/tags", "/v1/models", "/v1/openai/v1/models"):
         normalized["test_id"] = test_id
 
     normalized_json = json.dumps(normalized, sort_keys=True)
@@ -430,7 +430,7 @@ class ResponseStorage:
 
         # For model-list endpoints, include digest in filename to distinguish different model sets
         endpoint = request.get("endpoint")
-        if endpoint in ("/api/tags", "/v1/models"):
+        if endpoint in ("/api/tags", "/v1/models", "/v1/openai/v1/models"):
             digest = _model_identifiers_digest(endpoint, response)
             response_file = f"models-{request_hash}-{digest}.json"
 
@@ -554,13 +554,14 @@ def _model_identifiers_digest(endpoint: str, response: dict[str, Any]) -> str:
         Supported endpoints:
         - '/api/tags' (Ollama): response body has 'models': [ { name/model/digest/id/... }, ... ]
         - '/v1/models' (OpenAI): response body is: [ { id: ... }, ... ]
+        - '/v1/openai/v1/models' (OpenAI): response body is: [ { id: ... }, ... ]
         Returns a list of unique identifiers or None if structure doesn't match.
         """
         if "models" in response["body"]:
             # ollama
             items = response["body"]["models"]
         else:
-            # openai
+            # openai or openai-style endpoints
             items = response["body"]
         idents = [m.model if endpoint == "/api/tags" else m.id for m in items]
         return sorted(set(idents))
@@ -581,7 +582,7 @@ def _combine_model_list_responses(endpoint: str, records: list[dict[str, Any]])
     seen: dict[str, dict[str, Any]] = {}
     for rec in records:
         body = rec["response"]["body"]
-        if endpoint == "/v1/models":
+        if endpoint in ("/v1/models", "/v1/openai/v1/models"):
             for m in body:
                 key = m.id
                 seen[key] = m
@@ -665,7 +666,7 @@ async def _patched_inference_method(original_method, self, client_type, endpoint
         logger.info(f"  Test context: {get_test_context()}")
 
     if mode == APIRecordingMode.LIVE or storage is None:
-        if endpoint == "/v1/models":
+        if endpoint in ("/v1/models", "/v1/openai/v1/models"):
             return original_method(self, *args, **kwargs)
         else:
             return await original_method(self, *args, **kwargs)
@@ -699,7 +700,7 @@ async def _patched_inference_method(original_method, self, client_type, endpoint
     recording = None
     if mode == APIRecordingMode.REPLAY or mode == APIRecordingMode.RECORD_IF_MISSING:
         # Special handling for model-list endpoints: merge all recordings with this hash
-        if endpoint in ("/api/tags", "/v1/models"):
+        if endpoint in ("/api/tags", "/v1/models", "/v1/openai/v1/models"):
             records = storage._model_list_responses(request_hash)
             recording = _combine_model_list_responses(endpoint, records)
         else:
@@ -739,13 +740,13 @@ async def _patched_inference_method(original_method, self, client_type, endpoint
             )
 
     if mode == APIRecordingMode.RECORD or (mode == APIRecordingMode.RECORD_IF_MISSING and not recording):
-        if endpoint == "/v1/models":
+        if endpoint in ("/v1/models", "/v1/openai/v1/models"):
             response = original_method(self, *args, **kwargs)
         else:
             response = await original_method(self, *args, **kwargs)
 
         # we want to store the result of the iterator, not the iterator itself
-        if endpoint == "/v1/models":
+        if endpoint in ("/v1/models", "/v1/openai/v1/models"):
             response = [m async for m in response]
 
         request_data = {