Merge 2a34226727 into ea15f2a270

docs/docs/providers/datasetio/remote_nvidia.mdx

@@ -14,7 +14,7 @@ NVIDIA's dataset I/O provider for accessing datasets from NVIDIA's data platform
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `api_key` | `str \| None` | No |  | The NVIDIA API key. |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The NVIDIA API key. |
 | `dataset_namespace` | `str \| None` | No | default | The NVIDIA dataset namespace. |
 | `project_id` | `str \| None` | No | test-project | The NVIDIA project ID. |
 | `datasets_url` | `<class 'str'>` | No | http://nemo.test | Base URL for the NeMo Dataset API |

docs/docs/providers/files/remote_s3.mdx

@@ -17,7 +17,7 @@ AWS S3-based file storage provider for scalable cloud file management with metad
 | `bucket_name` | `<class 'str'>` | No |  | S3 bucket name to store files |
 | `region` | `<class 'str'>` | No | us-east-1 | AWS region where the bucket is located |
 | `aws_access_key_id` | `str \| None` | No |  | AWS access key ID (optional if using IAM roles) |
-| `aws_secret_access_key` | `str \| None` | No |  | AWS secret access key (optional if using IAM roles) |
+| `aws_secret_access_key` | `<class 'pydantic.types.SecretStr'>` | No |  | AWS secret access key (optional if using IAM roles) |
 | `endpoint_url` | `str \| None` | No |  | Custom S3 endpoint URL (for MinIO, LocalStack, etc.) |
 | `auto_create_bucket` | `<class 'bool'>` | No | False | Automatically create the S3 bucket if it doesn't exist |
 | `metadata_store` | `utils.sqlstore.sqlstore.SqliteSqlStoreConfig \| utils.sqlstore.sqlstore.PostgresSqlStoreConfig` | No | sqlite | SQL store configuration for file metadata |

docs/docs/providers/inference/remote_anthropic.mdx

@@ -14,7 +14,7 @@ Anthropic inference provider for accessing Claude models and Anthropic's AI serv
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `api_key` | `str \| None` | No |  | API key for Anthropic models |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | API key for Anthropic models |
 
 ## Sample Configuration
 

docs/docs/providers/inference/remote_bedrock.mdx

@@ -15,8 +15,8 @@ AWS Bedrock inference provider for accessing various AI models through AWS's man
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `aws_access_key_id` | `str \| None` | No |  | The AWS access key to use. Default use environment variable: AWS_ACCESS_KEY_ID |
-| `aws_secret_access_key` | `str \| None` | No |  | The AWS secret access key to use. Default use environment variable: AWS_SECRET_ACCESS_KEY |
+| `aws_secret_access_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The AWS secret access key to use. Default use environment variable: AWS_SECRET_ACCESS_KEY |
-| `aws_session_token` | `str \| None` | No |  | The AWS session token to use. Default use environment variable: AWS_SESSION_TOKEN |
+| `aws_session_token` | `<class 'pydantic.types.SecretStr'>` | No |  | The AWS session token to use. Default use environment variable: AWS_SESSION_TOKEN |
 | `region_name` | `str \| None` | No |  | The default AWS Region to use, for example, us-west-1 or us-west-2.Default use environment variable: AWS_DEFAULT_REGION |
 | `profile_name` | `str \| None` | No |  | The profile name that contains credentials to use.Default use environment variable: AWS_PROFILE |
 | `total_max_attempts` | `int \| None` | No |  | An integer representing the maximum number of attempts that will be made for a single request, including the initial attempt. Default use environment variable: AWS_MAX_ATTEMPTS |

docs/docs/providers/inference/remote_fireworks.mdx

@@ -16,7 +16,7 @@ Fireworks AI inference provider for Llama models and other AI models on the Fire
 |-------|------|----------|---------|-------------|
 | `allowed_models` | `list[str \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `url` | `<class 'str'>` | No | https://api.fireworks.ai/inference/v1 | The URL for the Fireworks server |
-| `api_key` | `pydantic.types.SecretStr \| None` | No |  | The Fireworks.ai API Key |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The Fireworks.ai API Key |
 
 ## Sample Configuration
 

docs/docs/providers/inference/remote_gemini.mdx

@@ -14,7 +14,7 @@ Google Gemini inference provider for accessing Gemini models and Google's AI ser
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `api_key` | `str \| None` | No |  | API key for Gemini models |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | API key for Gemini models |
 
 ## Sample Configuration
 

docs/docs/providers/inference/remote_groq.mdx

@@ -14,7 +14,7 @@ Groq inference provider for ultra-fast inference using Groq's LPU technology.
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `api_key` | `str \| None` | No |  | The Groq API key |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The Groq API key |
 | `url` | `<class 'str'>` | No | https://api.groq.com | The URL for the Groq AI server |
 
 ## Sample Configuration

docs/docs/providers/inference/remote_hf_endpoint.mdx

@@ -15,7 +15,7 @@ HuggingFace Inference Endpoints provider for dedicated model serving.
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `endpoint_name` | `<class 'str'>` | No |  | The name of the Hugging Face Inference Endpoint in the format of '&#123;namespace&#125;/&#123;endpoint_name&#125;' (e.g. 'my-cool-org/meta-llama-3-1-8b-instruct-rce'). Namespace is optional and will default to the user account if not provided. |
-| `api_token` | `pydantic.types.SecretStr \| None` | No |  | Your Hugging Face user access token (will default to locally saved token if not provided) |
+| `api_token` | `<class 'pydantic.types.SecretStr'>` | No |  | Your Hugging Face user access token (will default to locally saved token if not provided) |
 
 ## Sample Configuration
 

docs/docs/providers/inference/remote_hf_serverless.mdx

@@ -15,7 +15,7 @@ HuggingFace Inference API serverless provider for on-demand model inference.
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `huggingface_repo` | `<class 'str'>` | No |  | The model ID of the model on the Hugging Face Hub (e.g. 'meta-llama/Meta-Llama-3.1-70B-Instruct') |
-| `api_token` | `pydantic.types.SecretStr \| None` | No |  | Your Hugging Face user access token (will default to locally saved token if not provided) |
+| `api_token` | `<class 'pydantic.types.SecretStr'>` | No |  | Your Hugging Face user access token (will default to locally saved token if not provided) |
 
 ## Sample Configuration
 

docs/docs/providers/inference/remote_llama-openai-compat.mdx

@@ -14,7 +14,7 @@ Llama OpenAI-compatible provider for using Llama models with OpenAI API format.
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `api_key` | `str \| None` | No |  | The Llama API key |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The Llama API key |
 | `openai_compat_api_base` | `<class 'str'>` | No | https://api.llama.com/compat/v1/ | The URL for the Llama API server |
 
 ## Sample Configuration

docs/docs/providers/inference/remote_nvidia.mdx

@@ -15,7 +15,7 @@ NVIDIA inference provider for accessing NVIDIA NIM models and AI services.
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `url` | `<class 'str'>` | No | https://integrate.api.nvidia.com | A base url for accessing the NVIDIA NIM |
-| `api_key` | `pydantic.types.SecretStr \| None` | No |  | The NVIDIA API key, only needed of using the hosted service |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The NVIDIA API key, only needed of using the hosted service |
 | `timeout` | `<class 'int'>` | No | 60 | Timeout for the HTTP requests |
 | `append_api_version` | `<class 'bool'>` | No | True | When set to false, the API version will not be appended to the base_url. By default, it is true. |
 

docs/docs/providers/inference/remote_openai.mdx

@@ -14,7 +14,7 @@ OpenAI inference provider for accessing GPT models and other OpenAI services.
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `api_key` | `str \| None` | No |  | API key for OpenAI models |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | API key for OpenAI models |
 | `base_url` | `<class 'str'>` | No | https://api.openai.com/v1 | Base URL for OpenAI API |
 
 ## Sample Configuration

docs/docs/providers/inference/remote_passthrough.mdx

@@ -15,7 +15,7 @@ Passthrough inference provider for connecting to any external inference service
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `url` | `<class 'str'>` | No |  | The URL for the passthrough endpoint |
-| `api_key` | `pydantic.types.SecretStr \| None` | No |  | API Key for the passthrouth endpoint |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | API Key for the passthrouth endpoint |
 
 ## Sample Configuration
 

docs/docs/providers/inference/remote_runpod.mdx

@@ -15,7 +15,7 @@ RunPod inference provider for running models on RunPod's cloud GPU platform.
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `url` | `str \| None` | No |  | The URL for the Runpod model serving endpoint |
-| `api_token` | `str \| None` | No |  | The API token |
+| `api_token` | `<class 'pydantic.types.SecretStr'>` | No |  | The API token |
 
 ## Sample Configuration
 

docs/docs/providers/inference/remote_sambanova.mdx

@@ -15,7 +15,7 @@ SambaNova inference provider for running models on SambaNova's dataflow architec
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `url` | `<class 'str'>` | No | https://api.sambanova.ai/v1 | The URL for the SambaNova AI server |
-| `api_key` | `pydantic.types.SecretStr \| None` | No |  | The SambaNova cloud API Key |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The SambaNova cloud API Key |
 
 ## Sample Configuration
 

docs/docs/providers/inference/remote_together.mdx

@@ -16,7 +16,7 @@ Together AI inference provider for open-source models and collaborative AI devel
 |-------|------|----------|---------|-------------|
 | `allowed_models` | `list[str \| None` | No |  | List of models that should be registered with the model registry. If None, all models are allowed. |
 | `url` | `<class 'str'>` | No | https://api.together.xyz/v1 | The URL for the Together AI server |
-| `api_key` | `pydantic.types.SecretStr \| None` | No |  | The Together AI API Key |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The Together AI API Key |
 
 ## Sample Configuration
 

docs/docs/providers/inference/remote_vllm.mdx

@@ -16,7 +16,7 @@ Remote vLLM inference provider for connecting to vLLM servers.
 |-------|------|----------|---------|-------------|
 | `url` | `str \| None` | No |  | The URL for the vLLM model serving endpoint |
 | `max_tokens` | `<class 'int'>` | No | 4096 | Maximum number of tokens to generate. |
-| `api_token` | `str \| None` | No | fake | The API token |
+| `api_token` | `<class 'pydantic.types.SecretStr'>` | No | ********** | The API token |
 | `tls_verify` | `bool \| str` | No | True | Whether to verify TLS certificates. Can be a boolean or a path to a CA certificate file. |
 | `refresh_models` | `<class 'bool'>` | No | False | Whether to refresh models periodically |
 

docs/docs/providers/inference/remote_watsonx.mdx

@@ -15,7 +15,7 @@ IBM WatsonX inference provider for accessing AI models on IBM's WatsonX platform
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `url` | `<class 'str'>` | No | https://us-south.ml.cloud.ibm.com | A base url for accessing the watsonx.ai |
-| `api_key` | `pydantic.types.SecretStr \| None` | No |  | The watsonx API key |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The watsonx API key |
 | `project_id` | `str \| None` | No |  | The Project ID key |
 | `timeout` | `<class 'int'>` | No | 60 | Timeout for the HTTP requests |
 

docs/docs/providers/post_training/remote_nvidia.mdx

@@ -14,7 +14,7 @@ NVIDIA's post-training provider for fine-tuning models on NVIDIA's platform.
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `api_key` | `str \| None` | No |  | The NVIDIA API key. |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The NVIDIA API key. |
 | `dataset_namespace` | `str \| None` | No | default | The NVIDIA dataset namespace. |
 | `project_id` | `str \| None` | No | test-example-model@v1 | The NVIDIA project ID. |
 | `customizer_url` | `str \| None` | No |  | Base URL for the NeMo Customizer API |

docs/docs/providers/safety/remote_bedrock.mdx

@@ -15,8 +15,8 @@ AWS Bedrock safety provider for content moderation using AWS's safety services.
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `aws_access_key_id` | `str \| None` | No |  | The AWS access key to use. Default use environment variable: AWS_ACCESS_KEY_ID |
-| `aws_secret_access_key` | `str \| None` | No |  | The AWS secret access key to use. Default use environment variable: AWS_SECRET_ACCESS_KEY |
+| `aws_secret_access_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The AWS secret access key to use. Default use environment variable: AWS_SECRET_ACCESS_KEY |
-| `aws_session_token` | `str \| None` | No |  | The AWS session token to use. Default use environment variable: AWS_SESSION_TOKEN |
+| `aws_session_token` | `<class 'pydantic.types.SecretStr'>` | No |  | The AWS session token to use. Default use environment variable: AWS_SESSION_TOKEN |
 | `region_name` | `str \| None` | No |  | The default AWS Region to use, for example, us-west-1 or us-west-2.Default use environment variable: AWS_DEFAULT_REGION |
 | `profile_name` | `str \| None` | No |  | The profile name that contains credentials to use.Default use environment variable: AWS_PROFILE |
 | `total_max_attempts` | `int \| None` | No |  | An integer representing the maximum number of attempts that will be made for a single request, including the initial attempt. Default use environment variable: AWS_MAX_ATTEMPTS |

docs/docs/providers/safety/remote_sambanova.mdx

@@ -15,7 +15,7 @@ SambaNova's safety provider for content moderation and safety filtering.
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `url` | `<class 'str'>` | No | https://api.sambanova.ai/v1 | The URL for the SambaNova AI server |
-| `api_key` | `pydantic.types.SecretStr \| None` | No |  | The SambaNova cloud API Key |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The SambaNova cloud API Key |
 
 ## Sample Configuration
 

docs/docs/providers/scoring/inline_braintrust.mdx

@@ -14,7 +14,7 @@ Braintrust scoring provider for evaluation and scoring using the Braintrust plat
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `openai_api_key` | `str \| None` | No |  | The OpenAI API Key |
+| `openai_api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The OpenAI API Key |
 
 ## Sample Configuration
 

docs/docs/providers/tool_runtime/remote_bing-search.mdx

@@ -14,7 +14,7 @@ Bing Search tool for web search capabilities using Microsoft's search engine.
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `api_key` | `str \| None` | No |  |  |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The Bing API key |
 | `top_k` | `<class 'int'>` | No | 3 |  |
 
 ## Sample Configuration

docs/docs/providers/tool_runtime/remote_brave-search.mdx

@@ -14,7 +14,7 @@ Brave Search tool for web search capabilities with privacy-focused results.
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `api_key` | `str \| None` | No |  | The Brave Search API Key |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The Brave Search API Key |
 | `max_results` | `<class 'int'>` | No | 3 | The maximum number of results to return |
 
 ## Sample Configuration

docs/docs/providers/tool_runtime/remote_tavily-search.mdx

@@ -14,7 +14,7 @@ Tavily Search tool for AI-optimized web search with structured results.
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `api_key` | `str \| None` | No |  | The Tavily Search API Key |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The Tavily Search API Key |
 | `max_results` | `<class 'int'>` | No | 3 | The maximum number of results to return |
 
 ## Sample Configuration

docs/docs/providers/tool_runtime/remote_wolfram-alpha.mdx

@@ -14,7 +14,7 @@ Wolfram Alpha tool for computational knowledge and mathematical calculations.
 
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
-| `api_key` | `str \| None` | No |  |  |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The WolframAlpha API Key |
 
 ## Sample Configuration
 

docs/docs/providers/vector_io/remote_milvus.mdx

@@ -406,7 +406,7 @@ For more details on TLS configuration, refer to the [TLS setup guide](https://mi
 | Field | Type | Required | Default | Description |
 |-------|------|----------|---------|-------------|
 | `uri` | `<class 'str'>` | No |  | The URI of the Milvus server |
-| `token` | `str \| None` | No |  | The token of the Milvus server |
+| `token` | `<class 'pydantic.types.SecretStr'>` | No |  | The token of the Milvus server |
 | `consistency_level` | `<class 'str'>` | No | Strong | The consistency level of the Milvus server |
 | `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | Config for KV store backend |
 | `config` | `dict` | No | `{}` | This configuration allows additional fields to be passed through to the underlying Milvus client. See the [Milvus](https://milvus.io/docs/install-overview.md) documentation for more details about Milvus in general. |

docs/docs/providers/vector_io/remote_pgvector.mdx

@@ -217,7 +217,7 @@ See [PGVector's documentation](https://github.com/pgvector/pgvector) for more de
 | `port` | `int \| None` | No | 5432 |  |
 | `db` | `str \| None` | No | postgres |  |
 | `user` | `str \| None` | No | postgres |  |
-| `password` | `str \| None` | No | mysecretpassword |  |
+| `password` | `<class 'pydantic.types.SecretStr'>` | No | ********** |  |
 | `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig, annotation=NoneType, required=False, default='sqlite', discriminator='type'` | No |  | Config for KV store backend (SQLite only for now) |
 
 ## Sample Configuration

docs/docs/providers/vector_io/remote_qdrant.mdx

@@ -22,7 +22,7 @@ Please refer to the inline provider documentation.
 | `grpc_port` | `<class 'int'>` | No | 6334 |  |
 | `prefer_grpc` | `<class 'bool'>` | No | False |  |
 | `https` | `bool \| None` | No |  |  |
-| `api_key` | `str \| None` | No |  |  |
+| `api_key` | `<class 'pydantic.types.SecretStr'>` | No |  | The API key for the Qdrant instance |
 | `prefix` | `str \| None` | No |  |  |
 | `timeout` | `int \| None` | No |  |  |
 | `host` | `str \| None` | No |  |  |

llama_stack/cli/stack/_build.py

@@ -216,7 +216,7 @@ def run_stack_build_command(args: argparse.Namespace) -> None:
         with open(args.config) as f:
             try:
                 contents = yaml.safe_load(f)
-                contents = replace_env_vars(contents)
+                contents = replace_env_vars(contents, provider_registry=get_provider_registry())
                 build_config = BuildConfig(**contents)
                 if args.image_type:
                     build_config.image_type = args.image_type

llama_stack/core/configure.py

@@ -165,7 +165,7 @@ def upgrade_from_routing_table(
 def parse_and_maybe_upgrade_config(config_dict: dict[str, Any]) -> StackRunConfig:
     version = config_dict.get("version", None)
     if version == LLAMA_STACK_RUN_CONFIG_VERSION:
-        processed_config_dict = replace_env_vars(config_dict)
+        processed_config_dict = replace_env_vars(config_dict, provider_registry=get_provider_registry())
         return StackRunConfig(**cast_image_name_to_string(processed_config_dict))
 
     if "routing_table" in config_dict:
@@ -177,5 +177,5 @@ def parse_and_maybe_upgrade_config(config_dict: dict[str, Any]) -> StackRunConfi
     if not config_dict.get("external_providers_dir", None):
         config_dict["external_providers_dir"] = EXTERNAL_PROVIDERS_DIR
 
-    processed_config_dict = replace_env_vars(config_dict)
+    processed_config_dict = replace_env_vars(config_dict, provider_registry=get_provider_registry())
     return StackRunConfig(**cast_image_name_to_string(processed_config_dict))

llama_stack/core/library_client.py

@@ -33,6 +33,7 @@ from termcolor import cprint
 from llama_stack.core.build import print_pip_install_help
 from llama_stack.core.configure import parse_and_maybe_upgrade_config
 from llama_stack.core.datatypes import Api, BuildConfig, BuildProvider, DistributionSpec
+from llama_stack.core.distribution import get_provider_registry
 from llama_stack.core.request_headers import (
     PROVIDER_DATA_VAR,
     request_provider_data_context,
@@ -220,7 +221,9 @@ class AsyncLlamaStackAsLibraryClient(AsyncLlamaStackClient):
             config_path = Path(config_path_or_distro_name)
             if not config_path.exists():
                 raise ValueError(f"Config file {config_path} does not exist")
-            config_dict = replace_env_vars(yaml.safe_load(config_path.read_text()))
+            config_dict = replace_env_vars(
+                yaml.safe_load(config_path.read_text()), provider_registry=get_provider_registry()
+            )
             config = parse_and_maybe_upgrade_config(config_dict)
         else:
             # distribution

llama_stack/core/server/server.py

@@ -43,7 +43,7 @@ from llama_stack.core.datatypes import (
     StackRunConfig,
     process_cors_config,
 )
-from llama_stack.core.distribution import builtin_automatically_routed_apis
+from llama_stack.core.distribution import builtin_automatically_routed_apis, get_provider_registry
 from llama_stack.core.external import load_external_apis
 from llama_stack.core.request_headers import (
     PROVIDER_DATA_VAR,
@@ -371,7 +371,7 @@ def create_app(
                     logger.error(f"Error: {str(e)}")
                     raise ValueError(f"Invalid environment variable format: {env_pair}") from e
 
-        config = replace_env_vars(config_contents)
+        config = replace_env_vars(config_contents, provider_registry=get_provider_registry())
         config = StackRunConfig(**cast_image_name_to_string(config))
 
     _log_run_config(run_config=config)
@@ -524,7 +524,10 @@ def main(args: argparse.Namespace | None = None):
             env_vars=args.env,
         )
     except Exception as e:
+        import traceback
+
         logger.error(f"Error creating app: {str(e)}")
+        logger.error(f"Stack trace:\n{traceback.format_exc()}")
         sys.exit(1)
 
     config_file = resolve_config_or_distro(config_or_distro, Mode.RUN)
@@ -534,7 +537,9 @@ def main(args: argparse.Namespace | None = None):
             logger_config = LoggingConfig(**cfg)
         else:
             logger_config = None
-        config = StackRunConfig(**cast_image_name_to_string(replace_env_vars(config_contents)))
+        config = StackRunConfig(
+            **cast_image_name_to_string(replace_env_vars(config_contents, provider_registry=get_provider_registry()))
+        )
 
     import uvicorn
 

llama_stack/core/stack.py

@@ -141,12 +141,19 @@ class EnvVarError(Exception):
         )
 
 
-def replace_env_vars(config: Any, path: str = "") -> Any:
+def replace_env_vars(
+    config: Any,
+    path: str = "",
+    provider_registry: dict[Api, dict[str, Any]] | None = None,
+    current_provider_context: dict[str, Any] | None = None,
+) -> Any:
     if isinstance(config, dict):
         result = {}
         for k, v in config.items():
             try:
-                result[k] = replace_env_vars(v, f"{path}.{k}" if path else k)
+                result[k] = replace_env_vars(
+                    v, f"{path}.{k}" if path else k, provider_registry, current_provider_context
+                )
             except EnvVarError as e:
                 raise EnvVarError(e.var_name, e.path) from None
         return result
@@ -159,7 +166,9 @@ def replace_env_vars(config: Any, path: str = "") -> Any:
                 # is disabled so that we can skip config env variable expansion and avoid validation errors
                 if isinstance(v, dict) and "provider_id" in v:
                     try:
-                        resolved_provider_id = replace_env_vars(v["provider_id"], f"{path}[{i}].provider_id")
+                        resolved_provider_id = replace_env_vars(
+                            v["provider_id"], f"{path}[{i}].provider_id", provider_registry, current_provider_context
+                        )
                         if resolved_provider_id == "__disabled__":
                             logger.debug(
                                 f"Skipping config env variable expansion for disabled provider: {v.get('provider_id', '')}"
@@ -167,13 +176,19 @@ def replace_env_vars(config: Any, path: str = "") -> Any:
                             # Create a copy with resolved provider_id but original config
                             disabled_provider = v.copy()
                             disabled_provider["provider_id"] = resolved_provider_id
+                            result.append(disabled_provider)
                             continue
                     except EnvVarError:
                         # If we can't resolve the provider_id, continue with normal processing
                         pass
 
+                # Set up provider context for config processing
+                provider_context = current_provider_context
+                if isinstance(v, dict) and "provider_id" in v and "provider_type" in v and provider_registry:
+                    provider_context = _get_provider_context(v, provider_registry)
+
                 # Normal processing for non-disabled providers
-                result.append(replace_env_vars(v, f"{path}[{i}]"))
+                result.append(replace_env_vars(v, f"{path}[{i}]", provider_registry, provider_context))
             except EnvVarError as e:
                 raise EnvVarError(e.var_name, e.path) from None
         return result
@@ -228,7 +243,7 @@ def replace_env_vars(config: Any, path: str = "") -> Any:
             result = re.sub(pattern, get_env_var, config)
             # Only apply type conversion if substitution actually happened
             if result != config:
-                return _convert_string_to_proper_type(result)
+                return _convert_string_to_proper_type_with_config(result, path, current_provider_context)
             return result
         except EnvVarError as e:
             raise EnvVarError(e.var_name, e.path) from None
@@ -236,12 +251,113 @@ def replace_env_vars(config: Any, path: str = "") -> Any:
     return config
 
 
+def _get_provider_context(
+    provider_dict: dict[str, Any], provider_registry: dict[Api, dict[str, Any]]
+) -> dict[str, Any] | None:
+    """Get provider context information including config class for type conversion."""
+    try:
+        provider_type = provider_dict.get("provider_type")
+        if not provider_type:
+            return None
+
+        for api, providers in provider_registry.items():
+            if provider_type in providers:
+                provider_spec = providers[provider_type]
+
+                config_class = instantiate_class_type(provider_spec.config_class)
+
+                return {
+                    "api": api,
+                    "provider_type": provider_type,
+                    "config_class": config_class,
+                    "provider_spec": provider_spec,
+                }
+    except Exception as e:
+        logger.debug(f"Failed to get provider context: {e}")
+    return None
+
+
+def _convert_string_to_proper_type_with_config(value: str, path: str, provider_context: dict[str, Any] | None) -> Any:
+    """Convert string to proper type using provider config class field information."""
+    if not provider_context or not provider_context.get("config_class"):
+        # best effort conversion if we don't have the config class
+        return _convert_string_to_proper_type(value)
+
+    try:
+        # Extract field name from path (e.g., "providers.inference[0].config.api_key" -> "api_key")
+        field_name = path.split(".")[-1] if "." in path else path
+
+        config_class = provider_context["config_class"]
+        # Only instantiate if the class hasn't been instantiated already
+        # This handles the case we entered replace_env_vars() with a dict, which
+        # could happen if we use a sample_run_config() method that returns a dict. Our unit tests do
+        # this on the adhoc config spec creation.
+        if isinstance(config_class, str):
+            config_class = instantiate_class_type(config_class)
+
+        if hasattr(config_class, "model_fields") and field_name in config_class.model_fields:
+            field_info = config_class.model_fields[field_name]
+            field_type = field_info.annotation
+            return _convert_value_by_field_type(value, field_type)
+        else:
+            return _convert_string_to_proper_type(value)
+
+    except Exception as e:
+        logger.debug(f"Failed to convert using config class: {e}")
+        return _convert_string_to_proper_type(value)
+
+
+def _convert_value_by_field_type(value: str, field_type: Any) -> Any:
+    """Convert string value based on Pydantic field type annotation."""
+    import typing
+    from typing import get_args, get_origin
+
+    if value == "":
+        if field_type is None or (hasattr(typing, "get_origin") and get_origin(field_type) is type(None)):
+            return None
+        if hasattr(typing, "get_origin") and get_origin(field_type) is typing.Union:
+            args = get_args(field_type)
+            if type(None) in args:
+                return None
+        return ""
+
+    if field_type is bool or (hasattr(typing, "get_origin") and get_origin(field_type) is bool):
+        lowered = value.lower()
+        if lowered == "true":
+            return True
+        elif lowered == "false":
+            return False
+        else:
+            return value
+
+    if field_type is int or (hasattr(typing, "get_origin") and get_origin(field_type) is int):
+        try:
+            return int(value)
+        except ValueError:
+            return value
+
+    if field_type is float or (hasattr(typing, "get_origin") and get_origin(field_type) is float):
+        try:
+            return float(value)
+        except ValueError:
+            return value
+
+    if hasattr(typing, "get_origin") and get_origin(field_type) is typing.Union:
+        args = get_args(field_type)
+        # Try to convert to the first non-None type
+        for arg in args:
+            if arg is not type(None):
+                try:
+                    return _convert_value_by_field_type(value, arg)
+                except Exception:
+                    continue
+
+    return value
+
+
 def _convert_string_to_proper_type(value: str) -> Any:
-    # This might be tricky depending on what the config type is, if  'str | None' we are
+    # Fallback function for when provider config class is not available
-    # good, if 'str' we need to keep the empty string... 'str | None' is more common and
+    # The main type conversion logic is now in _convert_string_to_proper_type_with_config
-    # providers config should be typed this way.
-    # TODO: we could try to load the config class and see if the config has a field with type 'str | None'
-    # and then convert the empty string to None or not
     if value == "":
         return None
 
@@ -416,7 +532,7 @@ def get_stack_run_config_from_distro(distro: str) -> StackRunConfig:
             raise ValueError(f"Distribution '{distro}' not found at {distro_path}")
         run_config = yaml.safe_load(path.open())
 
-    return StackRunConfig(**replace_env_vars(run_config))
+    return StackRunConfig(**replace_env_vars(run_config, provider_registry=get_provider_registry()))
 
 
 def run_config_from_adhoc_config_spec(
@@ -452,7 +568,11 @@ def run_config_from_adhoc_config_spec(
 
         # call method "sample_run_config" on the provider spec config class
         provider_config_type = instantiate_class_type(provider_spec.config_class)
-        provider_config = replace_env_vars(provider_config_type.sample_run_config(__distro_dir__=distro_dir))
+        provider_config = replace_env_vars(
+            provider_config_type.sample_run_config(__distro_dir__=distro_dir),
+            provider_registry=provider_registry,
+            current_provider_context=provider_spec.model_dump(),
+        )
 
         provider_configs_by_api[api_str] = [
             Provider(

llama_stack/providers/inline/scoring/braintrust/__init__.py

@@ -5,7 +5,7 @@
 # the root directory of this source tree.
 from typing import Any
 
-from pydantic import BaseModel
+from pydantic import BaseModel, SecretStr
 
 from llama_stack.core.datatypes import Api
 
@@ -13,7 +13,7 @@ from .config import BraintrustScoringConfig
 
 
 class BraintrustProviderDataValidator(BaseModel):
-    openai_api_key: str
+    openai_api_key: SecretStr
 
 
 async def get_provider_impl(

llama_stack/providers/inline/scoring/braintrust/braintrust.py

@@ -17,7 +17,7 @@ from autoevals.ragas import (
     ContextRelevancy,
     Faithfulness,
 )
-from pydantic import BaseModel
+from pydantic import BaseModel, SecretStr
 
 from llama_stack.apis.datasetio import DatasetIO
 from llama_stack.apis.datasets import Datasets
@@ -152,9 +152,9 @@ class BraintrustScoringImpl(
                 raise ValueError(
                     'Pass OpenAI API Key in the header X-LlamaStack-Provider-Data as { "openai_api_key": <your api key>}'
                 )
-            self.config.openai_api_key = provider_data.openai_api_key
+            self.config.openai_api_key = SecretStr(provider_data.openai_api_key)
 
-        os.environ["OPENAI_API_KEY"] = self.config.openai_api_key
+        os.environ["OPENAI_API_KEY"] = self.config.openai_api_key.get_secret_value()
 
     async def score_batch(
         self,

llama_stack/providers/inline/scoring/braintrust/config.py

@@ -5,12 +5,11 @@
 # the root directory of this source tree.
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 
 class BraintrustScoringConfig(BaseModel):
-    openai_api_key: str | None = Field(
+    openai_api_key: SecretStr = Field(
-        default=None,
         description="The OpenAI API Key",
     )
 

llama_stack/providers/inline/telemetry/meta_reference/console_span_processor.py

@@ -64,6 +64,7 @@ class ConsoleSpanProcessor(SpanProcessor):
                 for key, value in event.attributes.items():
                     if key.startswith("__") or key in ["message", "severity"]:
                         continue
+
                     logger.info(f"[dim]{key}[/dim]: {value}")
 
     def shutdown(self) -> None:

llama_stack/providers/remote/datasetio/nvidia/config.py

@@ -8,14 +8,14 @@ import os
 import warnings
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 
 class NvidiaDatasetIOConfig(BaseModel):
     """Configuration for NVIDIA DatasetIO implementation."""
 
-    api_key: str | None = Field(
+    api_key: SecretStr = Field(
-        default_factory=lambda: os.getenv("NVIDIA_API_KEY"),
+        default_factory=lambda: SecretStr(os.getenv("NVIDIA_API_KEY", "")),
         description="The NVIDIA API key.",
     )
 

llama_stack/providers/remote/files/s3/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 from llama_stack.providers.utils.sqlstore.sqlstore import SqliteSqlStoreConfig, SqlStoreConfig
 
@@ -17,9 +17,7 @@ class S3FilesImplConfig(BaseModel):
     bucket_name: str = Field(description="S3 bucket name to store files")
     region: str = Field(default="us-east-1", description="AWS region where the bucket is located")
     aws_access_key_id: str | None = Field(default=None, description="AWS access key ID (optional if using IAM roles)")
-    aws_secret_access_key: str | None = Field(
+    aws_secret_access_key: SecretStr = Field(description="AWS secret access key (optional if using IAM roles)")
-        default=None, description="AWS secret access key (optional if using IAM roles)"
-    )
     endpoint_url: str | None = Field(default=None, description="Custom S3 endpoint URL (for MinIO, LocalStack, etc.)")
     auto_create_bucket: bool = Field(
         default=False, description="Automatically create the S3 bucket if it doesn't exist"

llama_stack/providers/remote/files/s3/files.py

@@ -47,7 +47,7 @@ def _create_s3_client(config: S3FilesImplConfig) -> boto3.client:
             s3_config.update(
                 {
                     "aws_access_key_id": config.aws_access_key_id,
-                    "aws_secret_access_key": config.aws_secret_access_key,
+                    "aws_secret_access_key": config.aws_secret_access_key.get_secret_value(),
                 }
             )
 

llama_stack/providers/remote/inference/anthropic/anthropic.py

@@ -4,6 +4,7 @@
 # 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.litellm_openai_mixin import LiteLLMOpenAIMixin
 from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
 

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

@@ -6,22 +6,20 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 from llama_stack.schema_utils import json_schema_type
 
 
 class AnthropicProviderDataValidator(BaseModel):
-    anthropic_api_key: str | None = Field(
+    anthropic_api_key: SecretStr = Field(
-        default=None,
         description="API key for Anthropic models",
     )
 
 
 @json_schema_type
 class AnthropicConfig(BaseModel):
-    api_key: str | None = Field(
+    api_key: SecretStr = Field(
-        default=None,
         description="API key for Anthropic models",
     )
 

llama_stack/providers/remote/inference/azure/azure.py

@@ -21,7 +21,7 @@ class AzureInferenceAdapter(OpenAIMixin, LiteLLMOpenAIMixin):
         LiteLLMOpenAIMixin.__init__(
             self,
             litellm_provider_name="azure",
-            api_key_from_config=config.api_key.get_secret_value(),
+            api_key_from_config=config.api_key,
             provider_data_api_key_field="azure_api_key",
             openai_compat_api_base=str(config.api_base),
         )

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

@@ -18,7 +18,6 @@ class DatabricksImplConfig(BaseModel):
         description="The URL for the Databricks model serving endpoint",
     )
     api_token: SecretStr = Field(
-        default=SecretStr(None),
         description="The Databricks API token",
     )
 

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

@@ -18,8 +18,7 @@ class FireworksImplConfig(RemoteInferenceProviderConfig):
         default="https://api.fireworks.ai/inference/v1",
         description="The URL for the Fireworks server",
     )
-    api_key: SecretStr | None = Field(
+    api_key: SecretStr = Field(
-        default=None,
         description="The Fireworks.ai API Key",
     )
 

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

@@ -6,22 +6,20 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 from llama_stack.schema_utils import json_schema_type
 
 
 class GeminiProviderDataValidator(BaseModel):
-    gemini_api_key: str | None = Field(
+    gemini_api_key: SecretStr = Field(
-        default=None,
         description="API key for Gemini models",
     )
 
 
 @json_schema_type
 class GeminiConfig(BaseModel):
-    api_key: str | None = Field(
+    api_key: SecretStr = Field(
-        default=None,
         description="API key for Gemini models",
     )
 

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

@@ -4,6 +4,7 @@
 # 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.litellm_openai_mixin import LiteLLMOpenAIMixin
 from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
 

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

@@ -6,23 +6,21 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 from llama_stack.schema_utils import json_schema_type
 
 
 class GroqProviderDataValidator(BaseModel):
-    groq_api_key: str | None = Field(
+    groq_api_key: SecretStr = Field(
-        default=None,
         description="API key for Groq models",
     )
 
 
 @json_schema_type
 class GroqConfig(BaseModel):
-    api_key: str | None = Field(
+    api_key: SecretStr = Field(
         # The Groq client library loads the GROQ_API_KEY environment variable by default
-        default=None,
         description="The Groq API key",
     )
 

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

@@ -6,22 +6,20 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 from llama_stack.schema_utils import json_schema_type
 
 
 class LlamaProviderDataValidator(BaseModel):
-    llama_api_key: str | None = Field(
+    llama_api_key: SecretStr = Field(
-        default=None,
         description="API key for api.llama models",
     )
 
 
 @json_schema_type
 class LlamaCompatConfig(BaseModel):
-    api_key: str | None = Field(
+    api_key: SecretStr = Field(
-        default=None,
         description="The Llama API key",
     )
 

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

@@ -39,8 +39,8 @@ class NVIDIAConfig(BaseModel):
         default_factory=lambda: os.getenv("NVIDIA_BASE_URL", "https://integrate.api.nvidia.com"),
         description="A base url for accessing the NVIDIA NIM",
     )
-    api_key: SecretStr | None = Field(
+    api_key: SecretStr = Field(
-        default_factory=lambda: SecretStr(os.getenv("NVIDIA_API_KEY")),
+        default_factory=lambda: SecretStr(os.getenv("NVIDIA_API_KEY", "")),
         description="The NVIDIA API key, only needed of using the hosted service",
     )
     timeout: int = Field(

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

@@ -6,22 +6,20 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 from llama_stack.schema_utils import json_schema_type
 
 
 class OpenAIProviderDataValidator(BaseModel):
-    openai_api_key: str | None = Field(
+    openai_api_key: SecretStr = Field(
-        default=None,
         description="API key for OpenAI models",
     )
 
 
 @json_schema_type
 class OpenAIConfig(BaseModel):
-    api_key: str | None = Field(
+    api_key: SecretStr = Field(
-        default=None,
         description="API key for OpenAI models",
     )
     base_url: str = Field(

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

@@ -18,8 +18,7 @@ class PassthroughImplConfig(BaseModel):
         description="The URL for the passthrough endpoint",
     )
 
-    api_key: SecretStr | None = Field(
+    api_key: SecretStr = Field(
-        default=None,
         description="API Key for the passthrouth endpoint",
     )
 

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

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 from llama_stack.schema_utils import json_schema_type
 
@@ -17,8 +17,7 @@ class RunpodImplConfig(BaseModel):
         default=None,
         description="The URL for the Runpod model serving endpoint",
     )
-    api_token: str | None = Field(
+    api_token: SecretStr = Field(
-        default=None,
         description="The API token",
     )
 

llama_stack/providers/remote/inference/runpod/runpod.py

@@ -103,7 +103,10 @@ class RunpodInferenceAdapter(
             tool_config=tool_config,
         )
 
-        client = OpenAI(base_url=self.config.url, api_key=self.config.api_token)
+        client = OpenAI(
+            base_url=self.config.url,
+            api_key=self.config.api_token.get_secret_value() if self.config.api_token else None,
+        )
         if stream:
             return self._stream_chat_completion(request, client)
         else:

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

@@ -12,8 +12,7 @@ from llama_stack.schema_utils import json_schema_type
 
 
 class SambaNovaProviderDataValidator(BaseModel):
-    sambanova_api_key: str | None = Field(
+    sambanova_api_key: SecretStr = Field(
-        default=None,
         description="Sambanova Cloud API key",
     )
 
@@ -24,8 +23,7 @@ class SambaNovaImplConfig(BaseModel):
         default="https://api.sambanova.ai/v1",
         description="The URL for the SambaNova AI server",
     )
-    api_key: SecretStr | None = Field(
+    api_key: SecretStr = Field(
-        default=None,
         description="The SambaNova cloud API Key",
     )
 

llama_stack/providers/remote/inference/sambanova/sambanova.py

@@ -29,7 +29,7 @@ class SambaNovaInferenceAdapter(OpenAIMixin, LiteLLMOpenAIMixin):
         LiteLLMOpenAIMixin.__init__(
             self,
             litellm_provider_name="sambanova",
-            api_key_from_config=self.config.api_key.get_secret_value() if self.config.api_key else None,
+            api_key_from_config=self.config.api_key,
             provider_data_api_key_field="sambanova_api_key",
             openai_compat_api_base=self.config.url,
             download_images=True,  # SambaNova requires base64 image encoding

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

@@ -32,8 +32,7 @@ class InferenceEndpointImplConfig(BaseModel):
     endpoint_name: str = Field(
         description="The name of the Hugging Face Inference Endpoint in the format of '{namespace}/{endpoint_name}' (e.g. 'my-cool-org/meta-llama-3-1-8b-instruct-rce'). Namespace is optional and will default to the user account if not provided.",
     )
-    api_token: SecretStr | None = Field(
+    api_token: SecretStr = Field(
-        default=None,
         description="Your Hugging Face user access token (will default to locally saved token if not provided)",
     )
 
@@ -55,8 +54,7 @@ class InferenceAPIImplConfig(BaseModel):
     huggingface_repo: str = Field(
         description="The model ID of the model on the Hugging Face Hub (e.g. 'meta-llama/Meta-Llama-3.1-70B-Instruct')",
     )
-    api_token: SecretStr | None = Field(
+    api_token: SecretStr = Field(
-        default=None,
         description="Your Hugging Face user access token (will default to locally saved token if not provided)",
     )
 

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

@@ -18,8 +18,7 @@ class TogetherImplConfig(RemoteInferenceProviderConfig):
         default="https://api.together.xyz/v1",
         description="The URL for the Together AI server",
     )
-    api_key: SecretStr | None = Field(
+    api_key: SecretStr = Field(
-        default=None,
         description="The Together AI API Key",
     )
 

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

@@ -8,6 +8,7 @@ from typing import Any
 
 import google.auth.transport.requests
 from google.auth import default
+from pydantic import SecretStr
 
 from llama_stack.apis.inference import ChatCompletionRequest
 from llama_stack.providers.utils.inference.litellm_openai_mixin import (
@@ -23,12 +24,12 @@ class VertexAIInferenceAdapter(OpenAIMixin, LiteLLMOpenAIMixin):
         LiteLLMOpenAIMixin.__init__(
             self,
             litellm_provider_name="vertex_ai",
-            api_key_from_config=None,  # Vertex AI uses ADC, not API keys
+            api_key_from_config=SecretStr(""),  # 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:
+    def get_api_key(self) -> SecretStr:
         """
         Get an access token for Vertex AI using Application Default Credentials.
 
@@ -39,11 +40,11 @@ class VertexAIInferenceAdapter(OpenAIMixin, LiteLLMOpenAIMixin):
             # Get default credentials - will read from GOOGLE_APPLICATION_CREDENTIALS
             credentials, _ = default(scopes=["https://www.googleapis.com/auth/cloud-platform"])
             credentials.refresh(google.auth.transport.requests.Request())
-            return str(credentials.token)
+            return SecretStr(credentials.token)
         except Exception:
             # If we can't get credentials, return empty string to let LiteLLM handle it
             # This allows the LiteLLM mixin to work with ADC directly
-            return ""
+            return SecretStr("")
 
     def get_base_url(self) -> str:
         """

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

@@ -4,13 +4,15 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from pydantic import BaseModel
+from pydantic import BaseModel, Field, SecretStr
 
 from .config import VLLMInferenceAdapterConfig
 
 
 class VLLMProviderDataValidator(BaseModel):
-    vllm_api_token: str | None = None
+    vllm_api_token: SecretStr = Field(
+        description="API token for vLLM models",
+    )
 
 
 async def get_adapter_impl(config: VLLMInferenceAdapterConfig, _deps):

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

@@ -6,7 +6,7 @@
 
 from pathlib import Path
 
-from pydantic import BaseModel, Field, field_validator
+from pydantic import BaseModel, Field, SecretStr, field_validator
 
 from llama_stack.schema_utils import json_schema_type
 
@@ -21,8 +21,8 @@ class VLLMInferenceAdapterConfig(BaseModel):
         default=4096,
         description="Maximum number of tokens to generate.",
     )
-    api_token: str | None = Field(
+    api_token: SecretStr = Field(
-        default="fake",
+        default=SecretStr("fake"),
         description="The API token",
     )
     tls_verify: bool | str = Field(

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

@@ -24,8 +24,8 @@ class WatsonXConfig(BaseModel):
         default_factory=lambda: os.getenv("WATSONX_BASE_URL", "https://us-south.ml.cloud.ibm.com"),
         description="A base url for accessing the watsonx.ai",
     )
-    api_key: SecretStr | None = Field(
+    api_key: SecretStr = Field(
-        default_factory=lambda: os.getenv("WATSONX_API_KEY"),
+        default_factory=lambda: SecretStr(os.getenv("WATSONX_API_KEY", "")),
         description="The watsonx API key",
     )
     project_id: str | None = Field(

llama_stack/providers/remote/post_training/nvidia/config.py

@@ -7,7 +7,7 @@
 import os
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 # TODO: add default values for all fields
 
@@ -15,8 +15,8 @@ from pydantic import BaseModel, Field
 class NvidiaPostTrainingConfig(BaseModel):
     """Configuration for NVIDIA Post Training implementation."""
 
-    api_key: str | None = Field(
+    api_key: SecretStr = Field(
-        default_factory=lambda: os.getenv("NVIDIA_API_KEY"),
+        default_factory=lambda: SecretStr(os.getenv("NVIDIA_API_KEY", "")),
         description="The NVIDIA API key.",
     )
 

llama_stack/providers/remote/safety/sambanova/config.py

@@ -12,8 +12,7 @@ from llama_stack.schema_utils import json_schema_type
 
 
 class SambaNovaProviderDataValidator(BaseModel):
-    sambanova_api_key: str | None = Field(
+    sambanova_api_key: SecretStr = Field(
-        default=None,
         description="Sambanova Cloud API key",
     )
 
@@ -24,8 +23,7 @@ class SambaNovaSafetyConfig(BaseModel):
         default="https://api.sambanova.ai/v1",
         description="The URL for the SambaNova AI server",
     )
-    api_key: SecretStr | None = Field(
+    api_key: SecretStr = Field(
-        default=None,
         description="The SambaNova cloud API Key",
     )
 

llama_stack/providers/remote/tool_runtime/bing_search/bing_search.py

@@ -40,7 +40,7 @@ class BingSearchToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, NeedsReq
 
     def _get_api_key(self) -> str:
         if self.config.api_key:
-            return self.config.api_key
+            return self.config.api_key.get_secret_value()
 
         provider_data = self.get_request_provider_data()
         if provider_data is None or not provider_data.bing_search_api_key:

llama_stack/providers/remote/tool_runtime/bing_search/config.py

@@ -6,13 +6,15 @@
 
 from typing import Any
 
-from pydantic import BaseModel
+from pydantic import BaseModel, Field, SecretStr
 
 
 class BingSearchToolConfig(BaseModel):
     """Configuration for Bing Search Tool Runtime"""
 
-    api_key: str | None = None
+    api_key: SecretStr = Field(
+        description="The Bing API key",
+    )
     top_k: int = 3
 
     @classmethod

llama_stack/providers/remote/tool_runtime/brave_search/brave_search.py

@@ -39,7 +39,7 @@ class BraveSearchToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, NeedsRe
 
     def _get_api_key(self) -> str:
         if self.config.api_key:
-            return self.config.api_key
+            return self.config.api_key.get_secret_value()
 
         provider_data = self.get_request_provider_data()
         if provider_data is None or not provider_data.brave_search_api_key:

llama_stack/providers/remote/tool_runtime/brave_search/config.py

@@ -6,12 +6,11 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 
 class BraveSearchToolConfig(BaseModel):
-    api_key: str | None = Field(
+    api_key: SecretStr = Field(
-        default=None,
         description="The Brave Search API Key",
     )
     max_results: int = Field(

llama_stack/providers/remote/tool_runtime/tavily_search/config.py

@@ -6,12 +6,11 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 
 class TavilySearchToolConfig(BaseModel):
-    api_key: str | None = Field(
+    api_key: SecretStr = Field(
-        default=None,
         description="The Tavily Search API Key",
     )
     max_results: int = Field(

llama_stack/providers/remote/tool_runtime/tavily_search/tavily_search.py

@@ -39,7 +39,7 @@ class TavilySearchToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, NeedsR
 
     def _get_api_key(self) -> str:
         if self.config.api_key:
-            return self.config.api_key
+            return self.config.api_key.get_secret_value()
 
         provider_data = self.get_request_provider_data()
         if provider_data is None or not provider_data.tavily_search_api_key:

llama_stack/providers/remote/tool_runtime/wolfram_alpha/config.py

@@ -6,13 +6,15 @@
 
 from typing import Any
 
-from pydantic import BaseModel
+from pydantic import BaseModel, Field, SecretStr
 
 
 class WolframAlphaToolConfig(BaseModel):
     """Configuration for WolframAlpha Tool Runtime"""
 
-    api_key: str | None = None
+    api_key: SecretStr = Field(
+        description="The WolframAlpha API Key",
+    )
 
     @classmethod
     def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> dict[str, Any]:

llama_stack/providers/remote/tool_runtime/wolfram_alpha/wolfram_alpha.py

@@ -40,7 +40,7 @@ class WolframAlphaToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, NeedsR
 
     def _get_api_key(self) -> str:
         if self.config.api_key:
-            return self.config.api_key
+            return self.config.api_key.get_secret_value()
 
         provider_data = self.get_request_provider_data()
         if provider_data is None or not provider_data.wolfram_alpha_api_key:

llama_stack/providers/remote/vector_io/milvus/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import BaseModel, ConfigDict, Field
+from pydantic import BaseModel, ConfigDict, Field, SecretStr
 
 from llama_stack.providers.utils.kvstore.config import KVStoreConfig, SqliteKVStoreConfig
 from llama_stack.schema_utils import json_schema_type
@@ -15,7 +15,7 @@ from llama_stack.schema_utils import json_schema_type
 @json_schema_type
 class MilvusVectorIOConfig(BaseModel):
     uri: str = Field(description="The URI of the Milvus server")
-    token: str | None = Field(description="The token of the Milvus server")
+    token: SecretStr = Field(description="The token of the Milvus server")
     consistency_level: str = Field(description="The consistency level of the Milvus server", default="Strong")
     kvstore: KVStoreConfig = Field(description="Config for KV store backend")
 

llama_stack/providers/remote/vector_io/pgvector/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 from llama_stack.providers.utils.kvstore.config import (
     KVStoreConfig,
@@ -21,7 +21,7 @@ class PGVectorVectorIOConfig(BaseModel):
     port: int | None = Field(default=5432)
     db: str | None = Field(default="postgres")
     user: str | None = Field(default="postgres")
-    password: str | None = Field(default="mysecretpassword")
+    password: SecretStr = Field(default=SecretStr("mysecretpassword"))
     kvstore: KVStoreConfig | None = Field(description="Config for KV store backend (SQLite only for now)", default=None)
 
     @classmethod

llama_stack/providers/remote/vector_io/pgvector/pgvector.py

@@ -366,7 +366,7 @@ class PGVectorVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtoco
                 port=self.config.port,
                 database=self.config.db,
                 user=self.config.user,
-                password=self.config.password,
+                password=self.config.password.get_secret_value(),
             )
             self.conn.autocommit = True
             with self.conn.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur:

llama_stack/providers/remote/vector_io/qdrant/config.py

@@ -6,7 +6,7 @@
 
 from typing import Any
 
-from pydantic import BaseModel
+from pydantic import BaseModel, Field, SecretStr
 
 from llama_stack.providers.utils.kvstore.config import (
     KVStoreConfig,
@@ -23,7 +23,9 @@ class QdrantVectorIOConfig(BaseModel):
     grpc_port: int = 6334
     prefer_grpc: bool = False
     https: bool | None = None
-    api_key: str | None = None
+    api_key: SecretStr = Field(
+        description="The API key for the Qdrant instance",
+    )
     prefix: str | None = None
     timeout: int | None = None
     host: str | None = None

llama_stack/providers/remote/vector_io/qdrant/qdrant.py

@@ -173,7 +173,7 @@ class QdrantVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtocolP
         self._qdrant_lock = asyncio.Lock()
 
     async def initialize(self) -> None:
-        client_config = self.config.model_dump(exclude_none=True, exclude={"kvstore"})
+        client_config = self.config.model_dump(exclude_none=True, exclude={"kvstore"}, mode="json")
         self.client = AsyncQdrantClient(**client_config)
         self.kvstore = await kvstore_impl(self.config.kvstore)
 

llama_stack/providers/utils/bedrock/client.py

@@ -50,8 +50,8 @@ def create_bedrock_client(config: BedrockBaseConfig, service_name: str = "bedroc
 
         session_args = {
             "aws_access_key_id": config.aws_access_key_id,
-            "aws_secret_access_key": config.aws_secret_access_key,
+            "aws_secret_access_key": config.aws_secret_access_key.get_secret_value(),
-            "aws_session_token": config.aws_session_token,
+            "aws_session_token": config.aws_session_token.get_secret_value(),
             "region_name": config.region_name,
             "profile_name": config.profile_name,
             "session_ttl": config.session_ttl,

llama_stack/providers/utils/bedrock/config.py

@@ -6,7 +6,7 @@
 
 import os
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 
 class BedrockBaseConfig(BaseModel):
@@ -14,12 +14,12 @@ class BedrockBaseConfig(BaseModel):
         default_factory=lambda: os.getenv("AWS_ACCESS_KEY_ID"),
         description="The AWS access key to use. Default use environment variable: AWS_ACCESS_KEY_ID",
     )
-    aws_secret_access_key: str | None = Field(
+    aws_secret_access_key: SecretStr = Field(
-        default_factory=lambda: os.getenv("AWS_SECRET_ACCESS_KEY"),
+        default_factory=lambda: SecretStr(os.getenv("AWS_SECRET_ACCESS_KEY", "")),
         description="The AWS secret access key to use. Default use environment variable: AWS_SECRET_ACCESS_KEY",
     )
-    aws_session_token: str | None = Field(
+    aws_session_token: SecretStr = Field(
-        default_factory=lambda: os.getenv("AWS_SESSION_TOKEN"),
+        default_factory=lambda: SecretStr(os.getenv("AWS_SESSION_TOKEN", "")),
         description="The AWS session token to use. Default use environment variable: AWS_SESSION_TOKEN",
     )
     region_name: str | None = Field(

llama_stack/providers/utils/inference/litellm_openai_mixin.py

@@ -8,6 +8,7 @@ from collections.abc import AsyncGenerator, AsyncIterator
 from typing import Any
 
 import litellm
+from pydantic import SecretStr
 
 from llama_stack.apis.common.content_types import (
     InterleavedContent,
@@ -61,7 +62,7 @@ class LiteLLMOpenAIMixin(
     def __init__(
         self,
         litellm_provider_name: str,
-        api_key_from_config: str | None,
+        api_key_from_config: SecretStr,
         provider_data_api_key_field: str,
         model_entries: list[ProviderModelEntry] | None = None,
         openai_compat_api_base: str | None = None,
@@ -240,14 +241,14 @@ class LiteLLMOpenAIMixin(
 
         return {
             "model": request.model,
-            "api_key": self.get_api_key(),
+            "api_key": self.get_api_key().get_secret_value(),
             "api_base": self.api_base,
             **input_dict,
             "stream": request.stream,
             **get_sampling_options(request.sampling_params),
         }
 
-    def get_api_key(self) -> str:
+    def get_api_key(self) -> SecretStr:
         provider_data = self.get_request_provider_data()
         key_field = self.provider_data_api_key_field
         if provider_data and getattr(provider_data, key_field, None):
@@ -280,7 +281,7 @@ class LiteLLMOpenAIMixin(
         response = litellm.embedding(
             model=self.get_litellm_model_name(model_obj.provider_resource_id),
             input=input_list,
-            api_key=self.get_api_key(),
+            api_key=self.get_api_key().get_secret_value(),
             api_base=self.api_base,
             dimensions=dimensions,
         )
@@ -343,7 +344,7 @@ class LiteLLMOpenAIMixin(
             user=user,
             guided_choice=guided_choice,
             prompt_logprobs=prompt_logprobs,
-            api_key=self.get_api_key(),
+            api_key=self.get_api_key().get_secret_value(),
             api_base=self.api_base,
         )
         return await litellm.atext_completion(**params)
@@ -407,7 +408,7 @@ class LiteLLMOpenAIMixin(
             top_logprobs=top_logprobs,
             top_p=top_p,
             user=user,
-            api_key=self.get_api_key(),
+            api_key=self.get_api_key().get_secret_value(),
             api_base=self.api_base,
         )
         return await litellm.acompletion(**params)

llama_stack/providers/utils/inference/openai_mixin.py

@@ -11,6 +11,7 @@ from collections.abc import AsyncIterator
 from typing import Any
 
 from openai import NOT_GIVEN, AsyncOpenAI
+from pydantic import SecretStr
 
 from llama_stack.apis.inference import (
     Model,
@@ -70,14 +71,14 @@ class OpenAIMixin(ModelRegistryHelper, ABC):
     allowed_models: list[str] = []
 
     @abstractmethod
-    def get_api_key(self) -> str:
+    def get_api_key(self) -> SecretStr:
         """
         Get the API key.
 
         This method must be implemented by child classes to provide the API key
         for authenticating with the OpenAI API or compatible endpoints.
 
-        :return: The API key as a string
+        :return: The API key as a SecretStr
         """
         pass
 
@@ -113,7 +114,7 @@ class OpenAIMixin(ModelRegistryHelper, ABC):
         implemented by child classes.
         """
         return AsyncOpenAI(
-            api_key=self.get_api_key(),
+            api_key=self.get_api_key().get_secret_value(),
             base_url=self.get_base_url(),
             **self.get_extra_client_params(),
         )

llama_stack/providers/utils/kvstore/config.py

@@ -8,7 +8,7 @@ import re
 from enum import Enum
 from typing import Annotated, Literal
 
-from pydantic import BaseModel, Field, field_validator
+from pydantic import BaseModel, Field, SecretStr, field_validator
 
 from llama_stack.core.utils.config_dirs import RUNTIME_BASE_DIR
 
@@ -74,7 +74,7 @@ class PostgresKVStoreConfig(CommonConfig):
     port: int = 5432
     db: str = "llamastack"
     user: str
-    password: str | None = None
+    password: SecretStr = SecretStr("")
     ssl_mode: str | None = None
     ca_cert_path: str | None = None
     table_name: str = "llamastack_kvstore"
@@ -118,7 +118,7 @@ class MongoDBKVStoreConfig(CommonConfig):
     port: int = 27017
     db: str = "llamastack"
     user: str | None = None
-    password: str | None = None
+    password: SecretStr = SecretStr("")
     collection_name: str = "llamastack_kvstore"
 
     @classmethod

llama_stack/providers/utils/kvstore/mongodb/mongodb.py

@@ -34,7 +34,7 @@ class MongoDBKVStoreImpl(KVStore):
                 "host": self.config.host,
                 "port": self.config.port,
                 "username": self.config.user,
-                "password": self.config.password,
+                "password": self.config.password.get_secret_value(),
             }
             conn_creds = {k: v for k, v in conn_creds.items() if v is not None}
             self.conn = AsyncMongoClient(**conn_creds)

llama_stack/providers/utils/kvstore/postgres/postgres.py

@@ -30,7 +30,7 @@ class PostgresKVStoreImpl(KVStore):
                 port=self.config.port,
                 database=self.config.db,
                 user=self.config.user,
-                password=self.config.password,
+                password=self.config.password.get_secret_value(),
                 sslmode=self.config.ssl_mode,
                 sslrootcert=self.config.ca_cert_path,
             )

llama_stack/providers/utils/sqlstore/sqlstore.py

@@ -9,7 +9,7 @@ from enum import StrEnum
 from pathlib import Path
 from typing import Annotated, Literal
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 from llama_stack.core.utils.config_dirs import RUNTIME_BASE_DIR
 
@@ -63,11 +63,11 @@ class PostgresSqlStoreConfig(SqlAlchemySqlStoreConfig):
     port: int = 5432
     db: str = "llamastack"
     user: str
-    password: str | None = None
+    password: SecretStr = SecretStr("")
 
     @property
     def engine_str(self) -> str:
-        return f"postgresql+asyncpg://{self.user}:{self.password}@{self.host}:{self.port}/{self.db}"
+        return f"postgresql+asyncpg://{self.user}:{self.password.get_secret_value()}@{self.host}:{self.port}/{self.db}"
 
     @classmethod
     def pip_packages(cls) -> list[str]:

tests/unit/providers/files/conftest.py

@@ -7,6 +7,7 @@
 import boto3
 import pytest
 from moto import mock_aws
+from pydantic import SecretStr
 
 from llama_stack.providers.remote.files.s3 import S3FilesImplConfig, get_adapter_impl
 from llama_stack.providers.utils.sqlstore.sqlstore import SqliteSqlStoreConfig
@@ -43,6 +44,7 @@ def s3_config(tmp_path):
         region="not-a-region",
         auto_create_bucket=True,
         metadata_store=SqliteSqlStoreConfig(db_path=db_path.as_posix()),
+        aws_secret_access_key=SecretStr("fake"),
     )
 
 

tests/unit/providers/inference/bedrock/test_config.py

@@ -17,7 +17,7 @@ class TestBedrockBaseConfig:
 
             # Basic creds should be None
             assert config.aws_access_key_id is None
-            assert config.aws_secret_access_key is None
+            assert not config.aws_secret_access_key
             assert config.region_name is None
 
             # Timeouts get defaults
@@ -39,7 +39,7 @@ class TestBedrockBaseConfig:
             config = BedrockBaseConfig()
 
             assert config.aws_access_key_id == "AKIATEST123"
-            assert config.aws_secret_access_key == "secret123"
+            assert config.aws_secret_access_key.get_secret_value() == "secret123"
             assert config.region_name == "us-west-2"
             assert config.total_max_attempts == 5
             assert config.retry_mode == "adaptive"

tests/unit/providers/inference/test_inference_client_caching.py

@@ -7,6 +7,8 @@
 import json
 from unittest.mock import MagicMock
 
+from pydantic import SecretStr
+
 from llama_stack.core.request_headers import request_provider_data_context
 from llama_stack.providers.remote.inference.groq.config import GroqConfig
 from llama_stack.providers.remote.inference.groq.groq import GroqInferenceAdapter
@@ -21,7 +23,7 @@ from llama_stack.providers.remote.inference.together.together import TogetherInf
 def test_groq_provider_openai_client_caching():
     """Ensure the Groq provider does not cache api keys across client requests"""
 
-    config = GroqConfig()
+    config = GroqConfig(api_key=SecretStr(""))
     inference_adapter = GroqInferenceAdapter(config)
 
     inference_adapter.__provider_spec__ = MagicMock()
@@ -39,7 +41,7 @@ def test_groq_provider_openai_client_caching():
 def test_openai_provider_openai_client_caching():
     """Ensure the OpenAI provider does not cache api keys across client requests"""
 
-    config = OpenAIConfig()
+    config = OpenAIConfig(api_key=SecretStr(""))
     inference_adapter = OpenAIInferenceAdapter(config)
 
     inference_adapter.__provider_spec__ = MagicMock()
@@ -58,7 +60,7 @@ def test_openai_provider_openai_client_caching():
 def test_together_provider_openai_client_caching():
     """Ensure the Together provider does not cache api keys across client requests"""
 
-    config = TogetherImplConfig()
+    config = TogetherImplConfig(api_key=SecretStr(""))
     inference_adapter = TogetherInferenceAdapter(config)
 
     inference_adapter.__provider_spec__ = MagicMock()
@@ -76,7 +78,7 @@ def test_together_provider_openai_client_caching():
 
 def test_llama_compat_provider_openai_client_caching():
     """Ensure the LlamaCompat provider does not cache api keys across client requests"""
-    config = LlamaCompatConfig()
+    config = LlamaCompatConfig(api_key=SecretStr(""))
     inference_adapter = LlamaCompatInferenceAdapter(config)
 
     inference_adapter.__provider_spec__ = MagicMock()

tests/unit/providers/inference/test_litellm_openai_mixin.py

@@ -8,7 +8,7 @@ import json
 from unittest.mock import MagicMock
 
 import pytest
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, SecretStr
 
 from llama_stack.core.request_headers import request_provider_data_context
 from llama_stack.providers.utils.inference.litellm_openai_mixin import LiteLLMOpenAIMixin
@@ -16,11 +16,11 @@ from llama_stack.providers.utils.inference.litellm_openai_mixin import LiteLLMOp
 
 # Test fixtures and helper classes
 class TestConfig(BaseModel):
-    api_key: str | None = Field(default=None)
+    api_key: SecretStr | None = Field(default=None)
 
 
 class TestProviderDataValidator(BaseModel):
-    test_api_key: str | None = Field(default=None)
+    test_api_key: SecretStr | None = Field(default=None)
 
 
 class TestLiteLLMAdapter(LiteLLMOpenAIMixin):
@@ -36,7 +36,7 @@ class TestLiteLLMAdapter(LiteLLMOpenAIMixin):
 @pytest.fixture
 def adapter_with_config_key():
     """Fixture to create adapter with API key in config"""
-    config = TestConfig(api_key="config-api-key")
+    config = TestConfig(api_key=SecretStr("config-api-key"))
     adapter = TestLiteLLMAdapter(config)
     adapter.__provider_spec__ = MagicMock()
     adapter.__provider_spec__.provider_data_validator = (
@@ -59,7 +59,7 @@ def adapter_without_config_key():
 
 def test_api_key_from_config_when_no_provider_data(adapter_with_config_key):
     """Test that adapter uses config API key when no provider data is available"""
-    api_key = adapter_with_config_key.get_api_key()
+    api_key = adapter_with_config_key.get_api_key().get_secret_value()
     assert api_key == "config-api-key"
 
 
@@ -68,28 +68,28 @@ def test_provider_data_takes_priority_over_config(adapter_with_config_key):
     with request_provider_data_context(
         {"x-llamastack-provider-data": json.dumps({"test_api_key": "provider-data-key"})}
     ):
-        api_key = adapter_with_config_key.get_api_key()
+        api_key = adapter_with_config_key.get_api_key().get_secret_value()
         assert api_key == "provider-data-key"
 
 
 def test_fallback_to_config_when_provider_data_missing_key(adapter_with_config_key):
     """Test fallback to config when provider data doesn't have the required key"""
     with request_provider_data_context({"x-llamastack-provider-data": json.dumps({"wrong_key": "some-value"})}):
-        api_key = adapter_with_config_key.get_api_key()
+        api_key = adapter_with_config_key.get_api_key().get_secret_value()
         assert api_key == "config-api-key"
 
 
 def test_error_when_no_api_key_available(adapter_without_config_key):
     """Test that ValueError is raised when neither config nor provider data have API key"""
     with pytest.raises(ValueError, match="API key is not set"):
-        adapter_without_config_key.get_api_key()
+        adapter_without_config_key.get_api_key().get_secret_value()
 
 
 def test_error_when_provider_data_has_wrong_key(adapter_without_config_key):
     """Test that ValueError is raised when provider data exists but doesn't have required key"""
     with request_provider_data_context({"x-llamastack-provider-data": json.dumps({"wrong_key": "some-value"})}):
         with pytest.raises(ValueError, match="API key is not set"):
-            adapter_without_config_key.get_api_key()
+            adapter_without_config_key.get_api_key().get_secret_value()
 
 
 def test_provider_data_works_when_config_is_none(adapter_without_config_key):
@@ -97,14 +97,14 @@ def test_provider_data_works_when_config_is_none(adapter_without_config_key):
     with request_provider_data_context(
         {"x-llamastack-provider-data": json.dumps({"test_api_key": "provider-only-key"})}
     ):
-        api_key = adapter_without_config_key.get_api_key()
+        api_key = adapter_without_config_key.get_api_key().get_secret_value()
         assert api_key == "provider-only-key"
 
 
 def test_error_message_includes_correct_field_names(adapter_without_config_key):
     """Test that error message includes correct field name and header information"""
     try:
-        adapter_without_config_key.get_api_key()
+        adapter_without_config_key.get_api_key().get_secret_value()
         raise AssertionError("Should have raised ValueError")
     except ValueError as e:
         assert "test_api_key" in str(e)  # Should mention the correct field name

tests/unit/providers/inference/test_openai_base_url_config.py

@@ -7,6 +7,8 @@
 import os
 from unittest.mock import MagicMock, patch
 
+from pydantic import SecretStr
+
 from llama_stack.core.stack import replace_env_vars
 from llama_stack.providers.remote.inference.openai.config import OpenAIConfig
 from llama_stack.providers.remote.inference.openai.openai import OpenAIInferenceAdapter
@@ -59,14 +61,14 @@ class TestOpenAIBaseURLConfig:
         adapter = OpenAIInferenceAdapter(config)
 
         # Mock the get_api_key method since it's delegated to LiteLLMOpenAIMixin
-        adapter.get_api_key = MagicMock(return_value="test-key")
+        adapter.get_api_key = MagicMock(return_value=SecretStr("test-key"))
 
         # Access the client property to trigger AsyncOpenAI initialization
         _ = adapter.client
 
         # Verify AsyncOpenAI was called with the correct base_url
         mock_openai_class.assert_called_once_with(
-            api_key="test-key",
+            api_key=SecretStr("test-key").get_secret_value(),
             base_url=custom_url,
         )
 
@@ -78,7 +80,7 @@ class TestOpenAIBaseURLConfig:
         adapter = OpenAIInferenceAdapter(config)
 
         # Mock the get_api_key method
-        adapter.get_api_key = MagicMock(return_value="test-key")
+        adapter.get_api_key = MagicMock(return_value=SecretStr("test-key"))
 
         # Mock a model object that will be returned by models.list()
         mock_model = MagicMock()
@@ -101,7 +103,7 @@ class TestOpenAIBaseURLConfig:
 
         # Verify the client was created with the custom URL
         mock_openai_class.assert_called_with(
-            api_key="test-key",
+            api_key=SecretStr("test-key").get_secret_value(),
             base_url=custom_url,
         )
 
@@ -119,7 +121,7 @@ class TestOpenAIBaseURLConfig:
         adapter = OpenAIInferenceAdapter(config)
 
         # Mock the get_api_key method
-        adapter.get_api_key = MagicMock(return_value="test-key")
+        adapter.get_api_key = MagicMock(return_value=SecretStr("test-key"))
 
         # Mock a model object that will be returned by models.list()
         mock_model = MagicMock()
@@ -142,6 +144,6 @@ class TestOpenAIBaseURLConfig:
 
         # Verify the client was created with the environment variable URL
         mock_openai_class.assert_called_with(
-            api_key="test-key",
+            api_key=SecretStr("test-key").get_secret_value(),
             base_url="https://proxy.openai.com/v1",
         )

tests/unit/providers/inference/test_remote_vllm.py

@@ -26,6 +26,7 @@ from openai.types.chat.chat_completion_chunk import (
     ChoiceDeltaToolCallFunction as OpenAIChoiceDeltaToolCallFunction,
 )
 from openai.types.model import Model as OpenAIModel
+from pydantic import SecretStr
 
 from llama_stack.apis.inference import (
     ChatCompletionRequest,
@@ -688,31 +689,35 @@ async def test_should_refresh_models():
     """
 
     # Test case 1: refresh_models is True, api_token is None
-    config1 = VLLMInferenceAdapterConfig(url="http://test.localhost", api_token=None, refresh_models=True)
+    config1 = VLLMInferenceAdapterConfig(url="http://test.localhost", api_token=SecretStr(""), refresh_models=True)
     adapter1 = VLLMInferenceAdapter(config1)
     result1 = await adapter1.should_refresh_models()
     assert result1 is True, "should_refresh_models should return True when refresh_models is True"
 
     # Test case 2: refresh_models is True, api_token is empty string
-    config2 = VLLMInferenceAdapterConfig(url="http://test.localhost", api_token="", refresh_models=True)
+    config2 = VLLMInferenceAdapterConfig(url="http://test.localhost", api_token=SecretStr(""), refresh_models=True)
     adapter2 = VLLMInferenceAdapter(config2)
     result2 = await adapter2.should_refresh_models()
     assert result2 is True, "should_refresh_models should return True when refresh_models is True"
 
     # Test case 3: refresh_models is True, api_token is "fake" (default)
-    config3 = VLLMInferenceAdapterConfig(url="http://test.localhost", api_token="fake", refresh_models=True)
+    config3 = VLLMInferenceAdapterConfig(url="http://test.localhost", api_token=SecretStr("fake"), refresh_models=True)
     adapter3 = VLLMInferenceAdapter(config3)
     result3 = await adapter3.should_refresh_models()
     assert result3 is True, "should_refresh_models should return True when refresh_models is True"
 
     # Test case 4: refresh_models is True, api_token is real token
-    config4 = VLLMInferenceAdapterConfig(url="http://test.localhost", api_token="real-token-123", refresh_models=True)
+    config4 = VLLMInferenceAdapterConfig(
+        url="http://test.localhost", api_token=SecretStr("real-token-123"), refresh_models=True
+    )
     adapter4 = VLLMInferenceAdapter(config4)
     result4 = await adapter4.should_refresh_models()
     assert result4 is True, "should_refresh_models should return True when refresh_models is True"
 
     # Test case 5: refresh_models is False, api_token is real token
-    config5 = VLLMInferenceAdapterConfig(url="http://test.localhost", api_token="real-token-456", refresh_models=False)
+    config5 = VLLMInferenceAdapterConfig(
+        url="http://test.localhost", api_token=SecretStr("real-token-456"), refresh_models=False
+    )
     adapter5 = VLLMInferenceAdapter(config5)
     result5 = await adapter5.should_refresh_models()
     assert result5 is False, "should_refresh_models should return False when refresh_models is False"
@@ -735,7 +740,7 @@ async def test_provider_data_var_context_propagation(vllm_inference_adapter):
 
         # Mock provider data to return test data
         mock_provider_data = MagicMock()
-        mock_provider_data.vllm_api_token = "test-token-123"
+        mock_provider_data.vllm_api_token = SecretStr("test-token-123")
         mock_provider_data.vllm_url = "http://test-server:8000/v1"
         mock_get_provider_data.return_value = mock_provider_data
 

tests/unit/providers/nvidia/test_parameters.py

@@ -9,6 +9,7 @@ import warnings
 from unittest.mock import patch
 
 import pytest
+from pydantic import SecretStr
 
 from llama_stack.apis.post_training.post_training import (
     DataConfig,
@@ -32,7 +33,7 @@ class TestNvidiaParameters:
         """Setup and teardown for each test method."""
         os.environ["NVIDIA_CUSTOMIZER_URL"] = "http://nemo.test"
 
-        config = NvidiaPostTrainingConfig(customizer_url=os.environ["NVIDIA_CUSTOMIZER_URL"], api_key=None)
+        config = NvidiaPostTrainingConfig(customizer_url=os.environ["NVIDIA_CUSTOMIZER_URL"], api_key=SecretStr(""))
         self.adapter = NvidiaPostTrainingAdapter(config)
 
         self.make_request_patcher = patch(

tests/unit/providers/nvidia/test_supervised_fine_tuning.py

@@ -9,6 +9,7 @@ import warnings
 from unittest.mock import patch
 
 import pytest
+from pydantic import SecretStr
 
 from llama_stack.apis.post_training.post_training import (
     DataConfig,
@@ -34,7 +35,7 @@ def nvidia_post_training_adapter():
     """Fixture to create and configure the NVIDIA post training adapter."""
     os.environ["NVIDIA_CUSTOMIZER_URL"] = "http://nemo.test"  # needed for nemo customizer
 
-    config = NvidiaPostTrainingConfig(customizer_url=os.environ["NVIDIA_CUSTOMIZER_URL"], api_key=None)
+    config = NvidiaPostTrainingConfig(customizer_url=os.environ["NVIDIA_CUSTOMIZER_URL"], api_key=SecretStr(""))
     adapter = NvidiaPostTrainingAdapter(config)
 
     with patch.object(adapter, "_make_request") as mock_make_request: