Merge branch 'meta-llama:main' into dell-distro

llama_stack/apis/common/errors.py

@@ -62,3 +62,13 @@ class SessionNotFoundError(ValueError):
     def __init__(self, session_name: str) -> None:
         message = f"Session '{session_name}' not found or access denied."
         super().__init__(message)
+
+
+class ModelTypeError(TypeError):
+    """raised when a model is present but not the correct type"""
+
+    def __init__(self, model_name: str, model_type: str, expected_model_type: str) -> None:
+        message = (
+            f"Model '{model_name}' is of type '{model_type}' rather than the expected type '{expected_model_type}'"
+        )
+        super().__init__(message)

llama_stack/core/routers/inference.py

@@ -18,7 +18,7 @@ from llama_stack.apis.common.content_types import (
     InterleavedContent,
     InterleavedContentItem,
 )
-from llama_stack.apis.common.errors import ModelNotFoundError
+from llama_stack.apis.common.errors import ModelNotFoundError, ModelTypeError
 from llama_stack.apis.inference import (
     BatchChatCompletionResponse,
     BatchCompletionResponse,
@@ -177,6 +177,15 @@ class InferenceRouter(Inference):
             encoded = self.formatter.encode_content(messages)
         return len(encoded.tokens) if encoded and encoded.tokens else 0
 
+    async def _get_model(self, model_id: str, expected_model_type: str) -> Model:
+        """takes a model id and gets model after ensuring that it is accessible and of the correct type"""
+        model = await self.routing_table.get_model(model_id)
+        if model is None:
+            raise ModelNotFoundError(model_id)
+        if model.model_type != expected_model_type:
+            raise ModelTypeError(model_id, model.model_type, expected_model_type)
+        return model
+
     async def chat_completion(
         self,
         model_id: str,
@@ -195,11 +204,7 @@ class InferenceRouter(Inference):
         )
         if sampling_params is None:
             sampling_params = SamplingParams()
-        model = await self.routing_table.get_model(model_id)
-        if model is None:
-            raise ModelNotFoundError(model_id)
-        if model.model_type == ModelType.embedding:
-            raise ValueError(f"Model '{model_id}' is an embedding model and does not support chat completions")
+        model = await self._get_model(model_id, ModelType.llm)
         if tool_config:
             if tool_choice and tool_choice != tool_config.tool_choice:
                 raise ValueError("tool_choice and tool_config.tool_choice must match")
@@ -301,11 +306,7 @@ class InferenceRouter(Inference):
         logger.debug(
             f"InferenceRouter.completion: {model_id=}, {stream=}, {content=}, {sampling_params=}, {response_format=}",
         )
-        model = await self.routing_table.get_model(model_id)
-        if model is None:
-            raise ModelNotFoundError(model_id)
-        if model.model_type == ModelType.embedding:
-            raise ValueError(f"Model '{model_id}' is an embedding model and does not support chat completions")
+        model = await self._get_model(model_id, ModelType.llm)
         provider = await self.routing_table.get_provider_impl(model_id)
         params = dict(
             model_id=model_id,
@@ -355,11 +356,7 @@ class InferenceRouter(Inference):
         task_type: EmbeddingTaskType | None = None,
     ) -> EmbeddingsResponse:
         logger.debug(f"InferenceRouter.embeddings: {model_id}")
-        model = await self.routing_table.get_model(model_id)
-        if model is None:
-            raise ModelNotFoundError(model_id)
-        if model.model_type == ModelType.llm:
-            raise ValueError(f"Model '{model_id}' is an LLM model and does not support embeddings")
+        await self._get_model(model_id, ModelType.embedding)
         provider = await self.routing_table.get_provider_impl(model_id)
         return await provider.embeddings(
             model_id=model_id,
@@ -395,12 +392,7 @@ class InferenceRouter(Inference):
         logger.debug(
             f"InferenceRouter.openai_completion: {model=}, {stream=}, {prompt=}",
         )
-        model_obj = await self.routing_table.get_model(model)
-        if model_obj is None:
-            raise ModelNotFoundError(model)
-        if model_obj.model_type == ModelType.embedding:
-            raise ValueError(f"Model '{model}' is an embedding model and does not support completions")
-
+        model_obj = await self._get_model(model, ModelType.llm)
         params = dict(
             model=model_obj.identifier,
             prompt=prompt,
@@ -476,11 +468,7 @@ class InferenceRouter(Inference):
         logger.debug(
             f"InferenceRouter.openai_chat_completion: {model=}, {stream=}, {messages=}",
         )
-        model_obj = await self.routing_table.get_model(model)
-        if model_obj is None:
-            raise ModelNotFoundError(model)
-        if model_obj.model_type == ModelType.embedding:
-            raise ValueError(f"Model '{model}' is an embedding model and does not support chat completions")
+        model_obj = await self._get_model(model, ModelType.llm)
 
         # Use the OpenAI client for a bit of extra input validation without
         # exposing the OpenAI client itself as part of our API surface
@@ -567,12 +555,7 @@ class InferenceRouter(Inference):
         logger.debug(
             f"InferenceRouter.openai_embeddings: {model=}, input_type={type(input)}, {encoding_format=}, {dimensions=}",
         )
-        model_obj = await self.routing_table.get_model(model)
-        if model_obj is None:
-            raise ModelNotFoundError(model)
-        if model_obj.model_type != ModelType.embedding:
-            raise ValueError(f"Model '{model}' is not an embedding model")
-
+        model_obj = await self._get_model(model, ModelType.embedding)
         params = dict(
             model=model_obj.identifier,
             input=input,

llama_stack/core/routing_tables/toolgroups.py

@@ -124,10 +124,7 @@ class ToolGroupsRoutingTable(CommonRoutingTableImpl, ToolGroups):
         return toolgroup
 
     async def unregister_toolgroup(self, toolgroup_id: str) -> None:
-        tool_group = await self.get_tool_group(toolgroup_id)
-        if tool_group is None:
-            raise ToolGroupNotFoundError(toolgroup_id)
-        await self.unregister_object(tool_group)
+        await self.unregister_object(await self.get_tool_group(toolgroup_id))
 
     async def shutdown(self) -> None:
         pass

llama_stack/core/routing_tables/vector_dbs.py

@@ -8,7 +8,7 @@ from typing import Any
 
 from pydantic import TypeAdapter
 
-from llama_stack.apis.common.errors import ModelNotFoundError, VectorStoreNotFoundError
+from llama_stack.apis.common.errors import ModelNotFoundError, ModelTypeError, VectorStoreNotFoundError
 from llama_stack.apis.models import ModelType
 from llama_stack.apis.resource import ResourceType
 from llama_stack.apis.vector_dbs import ListVectorDBsResponse, VectorDB, VectorDBs
@@ -66,7 +66,7 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
         if model is None:
             raise ModelNotFoundError(embedding_model)
         if model.model_type != ModelType.embedding:
-            raise ValueError(f"Model {embedding_model} is not an embedding model")
+            raise ModelTypeError(embedding_model, model.model_type, ModelType.embedding)
         if "embedding_dimension" not in model.metadata:
             raise ValueError(f"Model {embedding_model} does not have an embedding dimension")
         vector_db_data = {

llama_stack/distributions/ci-tests/build.yaml

@@ -14,6 +14,7 @@ distribution_spec:
     - provider_type: remote::openai
     - provider_type: remote::anthropic
     - provider_type: remote::gemini
+    - provider_type: remote::vertexai
     - provider_type: remote::groq
     - provider_type: remote::sambanova
     - provider_type: inline::sentence-transformers

llama_stack/distributions/ci-tests/run.yaml

@@ -65,6 +65,11 @@ providers:
     provider_type: remote::gemini
     config:
       api_key: ${env.GEMINI_API_KEY:=}
+  - provider_id: ${env.VERTEX_AI_PROJECT:+vertexai}
+    provider_type: remote::vertexai
+    config:
+      project: ${env.VERTEX_AI_PROJECT:=}
+      location: ${env.VERTEX_AI_LOCATION:=us-central1}
   - provider_id: groq
     provider_type: remote::groq
     config:

llama_stack/distributions/starter/build.yaml

@@ -14,6 +14,7 @@ distribution_spec:
     - provider_type: remote::openai
     - provider_type: remote::anthropic
     - provider_type: remote::gemini
+    - provider_type: remote::vertexai
     - provider_type: remote::groq
     - provider_type: remote::sambanova
     - provider_type: inline::sentence-transformers

llama_stack/distributions/starter/run.yaml

@@ -65,6 +65,11 @@ providers:
     provider_type: remote::gemini
     config:
       api_key: ${env.GEMINI_API_KEY:=}
+  - provider_id: ${env.VERTEX_AI_PROJECT:+vertexai}
+    provider_type: remote::vertexai
+    config:
+      project: ${env.VERTEX_AI_PROJECT:=}
+      location: ${env.VERTEX_AI_LOCATION:=us-central1}
   - provider_id: groq
     provider_type: remote::groq
     config:

llama_stack/distributions/starter/starter.py

@@ -56,6 +56,7 @@ ENABLED_INFERENCE_PROVIDERS = [
     "fireworks",
     "together",
     "gemini",
+    "vertexai",
     "groq",
     "sambanova",
     "anthropic",
@@ -71,6 +72,7 @@ INFERENCE_PROVIDER_IDS = {
     "tgi": "${env.TGI_URL:+tgi}",
     "cerebras": "${env.CEREBRAS_API_KEY:+cerebras}",
     "nvidia": "${env.NVIDIA_API_KEY:+nvidia}",
+    "vertexai": "${env.VERTEX_AI_PROJECT:+vertexai}",
 }
 
 
@@ -246,6 +248,14 @@ def get_distribution_template() -> DistributionTemplate:
                 "",
                 "Gemini API Key",
             ),
+            "VERTEX_AI_PROJECT": (
+                "",
+                "Google Cloud Project ID for Vertex AI",
+            ),
+            "VERTEX_AI_LOCATION": (
+                "us-central1",
+                "Google Cloud Location for Vertex AI",
+            ),
             "SAMBANOVA_API_KEY": (
                 "",
                 "SambaNova API Key",

llama_stack/log.py

@@ -99,7 +99,8 @@ def parse_environment_config(env_config: str) -> dict[str, int]:
         Dict[str, int]: A dictionary mapping categories to their log levels.
     """
     category_levels = {}
-    for pair in env_config.split(";"):
+    delimiter = ","
+    for pair in env_config.split(delimiter):
         if not pair.strip():
             continue
 

llama_stack/providers/inline/safety/prompt_guard/prompt_guard.py

@@ -15,6 +15,7 @@ from llama_stack.apis.safety import (
     RunShieldResponse,
     Safety,
     SafetyViolation,
+    ShieldStore,
     ViolationLevel,
 )
 from llama_stack.apis.shields import Shield
@@ -32,6 +33,8 @@ PROMPT_GUARD_MODEL = "Prompt-Guard-86M"
 
 
 class PromptGuardSafetyImpl(Safety, ShieldsProtocolPrivate):
+    shield_store: ShieldStore
+
     def __init__(self, config: PromptGuardConfig, _deps) -> None:
         self.config = config
 
@@ -53,7 +56,7 @@ class PromptGuardSafetyImpl(Safety, ShieldsProtocolPrivate):
         self,
         shield_id: str,
         messages: list[Message],
-        params: dict[str, Any] = None,
+        params: dict[str, Any],
     ) -> RunShieldResponse:
         shield = await self.shield_store.get_shield(shield_id)
         if not shield:
@@ -61,6 +64,9 @@ class PromptGuardSafetyImpl(Safety, ShieldsProtocolPrivate):
 
         return await self.shield.run(messages)
 
+    async def run_moderation(self, input: str | list[str], model: str):
+        raise NotImplementedError("run_moderation not implemented for PromptGuard")
+
 
 class PromptGuardShield:
     def __init__(
@@ -117,8 +123,10 @@ class PromptGuardShield:
         elif self.config.guard_type == PromptGuardType.jailbreak.value and score_malicious > self.threshold:
             violation = SafetyViolation(
                 violation_level=ViolationLevel.ERROR,
-                violation_type=f"prompt_injection:malicious={score_malicious}",
-                violation_return_message="Sorry, I cannot do this.",
+                user_message="Sorry, I cannot do this.",
+                metadata={
+                    "violation_type": f"prompt_injection:malicious={score_malicious}",
+                },
             )
 
         return RunShieldResponse(violation=violation)

llama_stack/providers/inline/vector_io/faiss/faiss.py

@@ -174,7 +174,9 @@ class FaissIndex(EmbeddingIndex):
         k: int,
         score_threshold: float,
     ) -> QueryChunksResponse:
-        raise NotImplementedError("Keyword search is not supported in FAISS")
+        raise NotImplementedError(
+            "Keyword search is not supported - underlying DB FAISS does not support this search mode"
+        )
 
     async def query_hybrid(
         self,
@@ -185,7 +187,9 @@ class FaissIndex(EmbeddingIndex):
         reranker_type: str,
         reranker_params: dict[str, Any] | None = None,
     ) -> QueryChunksResponse:
-        raise NotImplementedError("Hybrid search is not supported in FAISS")
+        raise NotImplementedError(
+            "Hybrid search is not supported - underlying DB FAISS does not support this search mode"
+        )
 
 
 class FaissVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtocolPrivate):

llama_stack/providers/registry/inference.py

@@ -213,6 +213,36 @@ def available_providers() -> list[ProviderSpec]:
                 description="Google Gemini inference provider for accessing Gemini models and Google's AI services.",
             ),
         ),
+        remote_provider_spec(
+            api=Api.inference,
+            adapter=AdapterSpec(
+                adapter_type="vertexai",
+                pip_packages=["litellm", "google-cloud-aiplatform"],
+                module="llama_stack.providers.remote.inference.vertexai",
+                config_class="llama_stack.providers.remote.inference.vertexai.VertexAIConfig",
+                provider_data_validator="llama_stack.providers.remote.inference.vertexai.config.VertexAIProviderDataValidator",
+                description="""Google Vertex AI inference provider enables you to use Google's Gemini models through Google Cloud's Vertex AI platform, providing several advantages:
+
+• Enterprise-grade security: Uses Google Cloud's security controls and IAM
+• Better integration: Seamless integration with other Google Cloud services
+• Advanced features: Access to additional Vertex AI features like model tuning and monitoring
+• Authentication: Uses Google Cloud Application Default Credentials (ADC) instead of API keys
+
+Configuration:
+- Set VERTEX_AI_PROJECT environment variable (required)
+- Set VERTEX_AI_LOCATION environment variable (optional, defaults to us-central1)
+- Use Google Cloud Application Default Credentials or service account key
+
+Authentication Setup:
+Option 1 (Recommended): gcloud auth application-default login
+Option 2: Set GOOGLE_APPLICATION_CREDENTIALS to service account key path
+
+Available Models:
+- vertex_ai/gemini-2.0-flash
+- vertex_ai/gemini-2.5-flash
+- vertex_ai/gemini-2.5-pro""",
+            ),
+        ),
         remote_provider_spec(
             api=Api.inference,
             adapter=AdapterSpec(

llama_stack/providers/registry/vector_io.py

@@ -45,6 +45,18 @@ That means you'll get fast and efficient vector retrieval.
 - Lightweight and easy to use
 - Fully integrated with Llama Stack
 - GPU support
+- **Vector search** - FAISS supports pure vector similarity search using embeddings
+
+## Search Modes
+
+**Supported:**
+- **Vector Search** (`mode="vector"`): Performs vector similarity search using embeddings
+
+**Not Supported:**
+- **Keyword Search** (`mode="keyword"`): Not supported by FAISS
+- **Hybrid Search** (`mode="hybrid"`): Not supported by FAISS
+
+> **Note**: FAISS is designed as a pure vector similarity search library. See the [FAISS GitHub repository](https://github.com/facebookresearch/faiss) for more details about FAISS's core functionality.
 
 ## Usage
 
@@ -535,6 +547,7 @@ That means you're not limited to storing vectors in memory or in a separate serv
 
 - Easy to use
 - Fully integrated with Llama Stack
+- Supports all search modes: vector, keyword, and hybrid search (both inline and remote configurations)
 
 ## Usage
 
@@ -625,6 +638,92 @@ vector_io:
 - **`client_pem_path`**: Path to the **client certificate** file (required for mTLS).
 - **`client_key_path`**: Path to the **client private key** file (required for mTLS).
 
+## Search Modes
+
+Milvus supports three different search modes for both inline and remote configurations:
+
+### Vector Search
+Vector search uses semantic similarity to find the most relevant chunks based on embedding vectors. This is the default search mode and works well for finding conceptually similar content.
+
+```python
+# Vector search example
+search_response = client.vector_stores.search(
+    vector_store_id=vector_store.id,
+    query="What is machine learning?",
+    search_mode="vector",
+    max_num_results=5,
+)
+```
+
+### Keyword Search
+Keyword search uses traditional text-based matching to find chunks containing specific terms or phrases. This is useful when you need exact term matches.
+
+```python
+# Keyword search example
+search_response = client.vector_stores.search(
+    vector_store_id=vector_store.id,
+    query="Python programming language",
+    search_mode="keyword",
+    max_num_results=5,
+)
+```
+
+### Hybrid Search
+Hybrid search combines both vector and keyword search methods to provide more comprehensive results. It leverages the strengths of both semantic similarity and exact term matching.
+
+#### Basic Hybrid Search
+```python
+# Basic hybrid search example (uses RRF ranker with default impact_factor=60.0)
+search_response = client.vector_stores.search(
+    vector_store_id=vector_store.id,
+    query="neural networks in Python",
+    search_mode="hybrid",
+    max_num_results=5,
+)
+```
+
+**Note**: The default `impact_factor` value of 60.0 was empirically determined to be optimal in the original RRF research paper: ["Reciprocal Rank Fusion outperforms Condorcet and individual Rank Learning Methods"](https://plg.uwaterloo.ca/~gvcormac/cormacksigir09-rrf.pdf) (Cormack et al., 2009).
+
+#### Hybrid Search with RRF (Reciprocal Rank Fusion) Ranker
+RRF combines rankings from vector and keyword search by using reciprocal ranks. The impact factor controls how much weight is given to higher-ranked results.
+
+```python
+# Hybrid search with custom RRF parameters
+search_response = client.vector_stores.search(
+    vector_store_id=vector_store.id,
+    query="neural networks in Python",
+    search_mode="hybrid",
+    max_num_results=5,
+    ranking_options={
+        "ranker": {
+            "type": "rrf",
+            "impact_factor": 100.0,  # Higher values give more weight to top-ranked results
+        }
+    },
+)
+```
+
+#### Hybrid Search with Weighted Ranker
+Weighted ranker linearly combines normalized scores from vector and keyword search. The alpha parameter controls the balance between the two search methods.
+
+```python
+# Hybrid search with weighted ranker
+search_response = client.vector_stores.search(
+    vector_store_id=vector_store.id,
+    query="neural networks in Python",
+    search_mode="hybrid",
+    max_num_results=5,
+    ranking_options={
+        "ranker": {
+            "type": "weighted",
+            "alpha": 0.7,  # 70% vector search, 30% keyword search
+        }
+    },
+)
+```
+
+For detailed documentation on RRF and Weighted rankers, please refer to the [Milvus Reranking Guide](https://milvus.io/docs/reranking.md).
+
 ## Documentation
 See the [Milvus documentation](https://milvus.io/docs/install-overview.md) for more details about Milvus in general.
 

llama_stack/providers/remote/inference/gemini/models.py

@@ -13,7 +13,9 @@ LLM_MODEL_IDS = [
     "gemini-1.5-flash",
     "gemini-1.5-pro",
     "gemini-2.0-flash",
+    "gemini-2.0-flash-lite",
     "gemini-2.5-flash",
+    "gemini-2.5-flash-lite",
     "gemini-2.5-pro",
 ]
 

llama_stack/providers/remote/inference/ollama/ollama.py

@@ -457,9 +457,6 @@ class OllamaInferenceAdapter(
         user: str | None = None,
     ) -> OpenAIEmbeddingsResponse:
         model_obj = await self._get_model(model)
-        if model_obj.model_type != ModelType.embedding:
-            raise ValueError(f"Model {model} is not an embedding model")
-
         if model_obj.provider_resource_id is None:
             raise ValueError(f"Model {model} has no provider_resource_id set")
 

llama_stack/providers/remote/inference/vertexai/__init__.py

@@ -0,0 +1,15 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from .config import VertexAIConfig
+
+
+async def get_adapter_impl(config: VertexAIConfig, _deps):
+    from .vertexai import VertexAIInferenceAdapter
+
+    impl = VertexAIInferenceAdapter(config)
+    await impl.initialize()
+    return impl

llama_stack/providers/remote/inference/vertexai/config.py

@@ -0,0 +1,45 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from typing import Any
+
+from pydantic import BaseModel, Field
+
+from llama_stack.schema_utils import json_schema_type
+
+
+class VertexAIProviderDataValidator(BaseModel):
+    vertex_project: str | None = Field(
+        default=None,
+        description="Google Cloud project ID for Vertex AI",
+    )
+    vertex_location: str | None = Field(
+        default=None,
+        description="Google Cloud location for Vertex AI (e.g., us-central1)",
+    )
+
+
+@json_schema_type
+class VertexAIConfig(BaseModel):
+    project: str = Field(
+        description="Google Cloud project ID for Vertex AI",
+    )
+    location: str = Field(
+        default="us-central1",
+        description="Google Cloud location for Vertex AI",
+    )
+
+    @classmethod
+    def sample_run_config(
+        cls,
+        project: str = "${env.VERTEX_AI_PROJECT:=}",
+        location: str = "${env.VERTEX_AI_LOCATION:=us-central1}",
+        **kwargs,
+    ) -> dict[str, Any]:
+        return {
+            "project": project,
+            "location": location,
+        }

llama_stack/providers/remote/inference/vertexai/models.py

@@ -0,0 +1,20 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from llama_stack.providers.utils.inference.model_registry import (
+    ProviderModelEntry,
+)
+
+# Vertex AI model IDs with vertex_ai/ prefix as required by litellm
+LLM_MODEL_IDS = [
+    "vertex_ai/gemini-2.0-flash",
+    "vertex_ai/gemini-2.5-flash",
+    "vertex_ai/gemini-2.5-pro",
+]
+
+SAFETY_MODELS_ENTRIES = list[ProviderModelEntry]()
+
+MODEL_ENTRIES = [ProviderModelEntry(provider_model_id=m) for m in LLM_MODEL_IDS] + SAFETY_MODELS_ENTRIES

llama_stack/providers/remote/inference/vertexai/vertexai.py

@@ -0,0 +1,52 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from typing import Any
+
+from llama_stack.apis.inference import ChatCompletionRequest
+from llama_stack.providers.utils.inference.litellm_openai_mixin import (
+    LiteLLMOpenAIMixin,
+)
+
+from .config import VertexAIConfig
+from .models import MODEL_ENTRIES
+
+
+class VertexAIInferenceAdapter(LiteLLMOpenAIMixin):
+    def __init__(self, config: VertexAIConfig) -> None:
+        LiteLLMOpenAIMixin.__init__(
+            self,
+            MODEL_ENTRIES,
+            litellm_provider_name="vertex_ai",
+            api_key_from_config=None,  # Vertex AI uses ADC, not API keys
+            provider_data_api_key_field="vertex_project",  # Use project for validation
+        )
+        self.config = config
+
+    def get_api_key(self) -> str:
+        # Vertex AI doesn't use API keys, it uses Application Default Credentials
+        # Return empty string to let litellm handle authentication via ADC
+        return ""
+
+    async def _get_params(self, request: ChatCompletionRequest) -> dict[str, Any]:
+        # Get base parameters from parent
+        params = await super()._get_params(request)
+
+        # Add Vertex AI specific parameters
+        provider_data = self.get_request_provider_data()
+        if provider_data:
+            if getattr(provider_data, "vertex_project", None):
+                params["vertex_project"] = provider_data.vertex_project
+            if getattr(provider_data, "vertex_location", None):
+                params["vertex_location"] = provider_data.vertex_location
+        else:
+            params["vertex_project"] = self.config.project
+            params["vertex_location"] = self.config.location
+
+        # Remove api_key since Vertex AI uses ADC
+        params.pop("api_key", None)
+
+        return params

llama_stack/providers/utils/inference/openai_compat.py

@@ -70,7 +70,7 @@ from openai.types.chat.chat_completion_chunk import (
 from openai.types.chat.chat_completion_content_part_image_param import (
     ImageURL as OpenAIImageURL,
 )
-from openai.types.chat.chat_completion_message_tool_call_param import (
+from openai.types.chat.chat_completion_message_tool_call import (
     Function as OpenAIFunction,
 )
 from pydantic import BaseModel

llama_stack/providers/utils/telemetry/tracing.py

@@ -9,7 +9,9 @@ import contextvars
 import logging
 import queue
 import random
+import sys
 import threading
+import time
 from collections.abc import Callable
 from datetime import UTC, datetime
 from functools import wraps
@@ -30,6 +32,16 @@ from llama_stack.providers.utils.telemetry.trace_protocol import serialize_value
 
 logger = get_logger(__name__, category="core")
 
+# Fallback logger that does NOT propagate to TelemetryHandler to avoid recursion
+_fallback_logger = logging.getLogger("llama_stack.telemetry.background")
+if not _fallback_logger.handlers:
+    _fallback_logger.propagate = False
+    _fallback_logger.setLevel(logging.ERROR)
+    _fallback_handler = logging.StreamHandler(sys.stderr)
+    _fallback_handler.setLevel(logging.ERROR)
+    _fallback_handler.setFormatter(logging.Formatter("%(asctime)s [%(levelname)s] %(name)s: %(message)s"))
+    _fallback_logger.addHandler(_fallback_handler)
+
 
 INVALID_SPAN_ID = 0x0000000000000000
 INVALID_TRACE_ID = 0x00000000000000000000000000000000
@@ -79,19 +91,32 @@ def generate_trace_id() -> str:
 CURRENT_TRACE_CONTEXT = contextvars.ContextVar("trace_context", default=None)
 BACKGROUND_LOGGER = None
 
+LOG_QUEUE_FULL_LOG_INTERVAL_SECONDS = 60.0
+
 
 class BackgroundLogger:
     def __init__(self, api: Telemetry, capacity: int = 100000):
         self.api = api
-        self.log_queue = queue.Queue(maxsize=capacity)
+        self.log_queue: queue.Queue[Any] = queue.Queue(maxsize=capacity)
         self.worker_thread = threading.Thread(target=self._process_logs, daemon=True)
         self.worker_thread.start()
+        self._last_queue_full_log_time: float = 0.0
+        self._dropped_since_last_notice: int = 0
 
     def log_event(self, event):
         try:
             self.log_queue.put_nowait(event)
         except queue.Full:
-            logger.error("Log queue is full, dropping event")
+            # Aggregate drops and emit at most once per interval via fallback logger
+            self._dropped_since_last_notice += 1
+            current_time = time.time()
+            if current_time - self._last_queue_full_log_time >= LOG_QUEUE_FULL_LOG_INTERVAL_SECONDS:
+                _fallback_logger.error(
+                    "Log queue is full; dropped %d events since last notice",
+                    self._dropped_since_last_notice,
+                )
+                self._last_queue_full_log_time = current_time
+                self._dropped_since_last_notice = 0
 
     def _process_logs(self):
         while True:

llama_stack/ui/app/chat-playground/page.tsx

@@ -175,7 +175,7 @@ const handleSubmitWithContent = async (content: string) => {
   return (
     <div className="flex flex-col h-full max-w-4xl mx-auto">
       <div className="mb-4 flex justify-between items-center">
-        <h1 className="text-2xl font-bold">Chat Playground</h1>
+        <h1 className="text-2xl font-bold">Chat Playground (Completions)</h1>
         <div className="flex gap-2">
           <Select value={selectedModel} onValueChange={setSelectedModel} disabled={isModelsLoading || isGenerating}>
             <SelectTrigger className="w-[180px]">

llama_stack/ui/components/layout/app-sidebar.tsx

@@ -6,6 +6,8 @@ import {
   MoveUpRight,
   Database,
   MessageCircle,
+  Settings2,
+  Compass,
 } from "lucide-react";
 import Link from "next/link";
 import { usePathname } from "next/navigation";
@@ -22,15 +24,16 @@ import {
   SidebarMenuItem,
   SidebarHeader,
 } from "@/components/ui/sidebar";
-// Extracted Chat Playground item
-const chatPlaygroundItem = {
-  title: "Chat Playground",
-  url: "/chat-playground",
-  icon: MessageCircle,
-};
 
-// Removed Chat Playground from log items
-const logItems = [
+const createItems = [
+  {
+    title: "Chat Playground",
+    url: "/chat-playground",
+    icon: MessageCircle,
+  },
+];
+
+const manageItems = [
   {
     title: "Chat Completions",
     url: "/logs/chat-completions",
@@ -53,77 +56,96 @@ const logItems = [
   },
 ];
 
+const optimizeItems: { title: string; url: string; icon: React.ElementType }[] = [
+    {
+        title: "Evaluations",
+        url: "",
+        icon: Compass,
+    },
+    {
+        title: "Fine-tuning",
+        url: "",
+        icon: Settings2,
+    },
+];
+
+interface SidebarItem {
+  title: string;
+  url: string;
+  icon: React.ElementType;
+}
+
 export function AppSidebar() {
   const pathname = usePathname();
 
-  return (
-    <Sidebar>
-      <SidebarHeader>
-        <Link href="/">Llama Stack</Link>
-      </SidebarHeader>
-      <SidebarContent>
-        {/* Chat Playground as its own section */}
-        <SidebarGroup>
-          <SidebarGroupContent>
-            <SidebarMenu>
-              <SidebarMenuItem>
+  const renderSidebarItems = (items: SidebarItem[]) => {
+    return items.map((item) => {
+      const isActive = pathname.startsWith(item.url);
+      return (
+        <SidebarMenuItem key={item.title}>
+          <SidebarMenuButton
+            asChild
+            className={cn(
+              "justify-start",
+              isActive &&
+                "bg-gray-200 dark:bg-gray-700 hover:bg-gray-200 dark:hover:bg-gray-700 text-gray-900 dark:text-gray-100",
+            )}
+          >
+            <Link href={item.url}>
+              <item.icon
+                className={cn(
+                  isActive && "text-gray-900 dark:text-gray-100",
+                  "mr-2 h-4 w-4",
+                )}
+              />
+              <span>{item.title}</span>
+            </Link>
+          </SidebarMenuButton>
+        </SidebarMenuItem>
+      );
+    });
+  };
+
+return (
+  <Sidebar>
+    <SidebarHeader>
+      <Link href="/">Llama Stack</Link>
+    </SidebarHeader>
+    <SidebarContent>
+      <SidebarGroup>
+        <SidebarGroupLabel>Create</SidebarGroupLabel>
+        <SidebarGroupContent>
+          <SidebarMenu>{renderSidebarItems(createItems)}</SidebarMenu>
+        </SidebarGroupContent>
+      </SidebarGroup>
+
+      <SidebarGroup>
+        <SidebarGroupLabel>Manage</SidebarGroupLabel>
+        <SidebarGroupContent>
+          <SidebarMenu>{renderSidebarItems(manageItems)}</SidebarMenu>
+        </SidebarGroupContent>
+      </SidebarGroup>
+
+      <SidebarGroup>
+        <SidebarGroupLabel>Optimize</SidebarGroupLabel>
+        <SidebarGroupContent>
+          <SidebarMenu>
+            {optimizeItems.map((item) => (
+              <SidebarMenuItem key={item.title}>
                 <SidebarMenuButton
-                  asChild
-                  className={cn(
-                    "justify-start",
-                    pathname.startsWith(chatPlaygroundItem.url) &&
-                      "bg-gray-200 dark:bg-gray-700 hover:bg-gray-200 dark:hover:bg-gray-700 text-gray-900 dark:text-gray-100",
-                  )}
+                  disabled
+                  className="justify-start opacity-60 cursor-not-allowed"
                 >
-                  <Link href={chatPlaygroundItem.url}>
-                    <chatPlaygroundItem.icon
-                      className={cn(
-                        pathname.startsWith(chatPlaygroundItem.url) && "text-gray-900 dark:text-gray-100",
-                        "mr-2 h-4 w-4",
-                      )}
-                    />
-                    <span>{chatPlaygroundItem.title}</span>
-                  </Link>
+                  <item.icon className="mr-2 h-4 w-4" />
+                  <span>{item.title}</span>
+                  <span className="ml-2 text-xs text-gray-500">(Coming Soon)</span>
                 </SidebarMenuButton>
               </SidebarMenuItem>
-            </SidebarMenu>
-          </SidebarGroupContent>
-        </SidebarGroup>
-
-        {/* Logs section */}
-        <SidebarGroup>
-          <SidebarGroupLabel>Logs</SidebarGroupLabel>
-          <SidebarGroupContent>
-            <SidebarMenu>
-              {logItems.map((item) => {
-                const isActive = pathname.startsWith(item.url);
-                return (
-                  <SidebarMenuItem key={item.title}>
-                    <SidebarMenuButton
-                      asChild
-                      className={cn(
-                        "justify-start",
-                        isActive &&
-                          "bg-gray-200 dark:bg-gray-700 hover:bg-gray-200 dark:hover:bg-gray-700 text-gray-900 dark:text-gray-100",
-                      )}
-                    >
-                      <Link href={item.url}>
-                        <item.icon
-                          className={cn(
-                            isActive && "text-gray-900 dark:text-gray-100",
-                            "mr-2 h-4 w-4",
-                          )}
-                        />
-                        <span>{item.title}</span>
-                      </Link>
-                    </SidebarMenuButton>
-                  </SidebarMenuItem>
-                );
-              })}
-            </SidebarMenu>
-          </SidebarGroupContent>
-        </SidebarGroup>
-      </SidebarContent>
-    </Sidebar>
+            ))}
+          </SidebarMenu>
+        </SidebarGroupContent>
+      </SidebarGroup>
+    </SidebarContent>
+  </Sidebar>
   );
 }