update starter

docs/source/providers/inference/remote_anthropic.md

@@ -13,7 +13,7 @@ Anthropic inference provider for accessing Claude models and Anthropic's AI serv
 ## Sample Configuration
 
 ```yaml
-api_key: ${env.ANTHROPIC_API_KEY}
+api_key: ${env.ANTHROPIC_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_cerebras-openai-compat.md

@@ -15,7 +15,7 @@ Cerebras OpenAI-compatible provider for using Cerebras models with OpenAI API fo
 
 ```yaml
 openai_compat_api_base: https://api.cerebras.ai/v1
-api_key: ${env.CEREBRAS_API_KEY}
+api_key: ${env.CEREBRAS_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_cerebras.md

@@ -15,7 +15,7 @@ Cerebras inference provider for running models on Cerebras Cloud platform.
 
 ```yaml
 base_url: https://api.cerebras.ai
-api_key: ${env.CEREBRAS_API_KEY}
+api_key: ${env.CEREBRAS_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_databricks.md

@@ -14,8 +14,8 @@ Databricks inference provider for running models on Databricks' unified analytic
 ## Sample Configuration
 
 ```yaml
-url: ${env.DATABRICKS_URL}
-api_token: ${env.DATABRICKS_API_TOKEN}
+url: ${env.DATABRICKS_URL:=}
+api_token: ${env.DATABRICKS_API_TOKEN:=}
 
 ```
 

docs/source/providers/inference/remote_fireworks-openai-compat.md

@@ -15,7 +15,7 @@ Fireworks AI OpenAI-compatible provider for using Fireworks models with OpenAI A
 
 ```yaml
 openai_compat_api_base: https://api.fireworks.ai/inference/v1
-api_key: ${env.FIREWORKS_API_KEY}
+api_key: ${env.FIREWORKS_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_fireworks.md

@@ -15,7 +15,7 @@ Fireworks AI inference provider for Llama models and other AI models on the Fire
 
 ```yaml
 url: https://api.fireworks.ai/inference/v1
-api_key: ${env.FIREWORKS_API_KEY}
+api_key: ${env.FIREWORKS_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_gemini.md

@@ -13,7 +13,7 @@ Google Gemini inference provider for accessing Gemini models and Google's AI ser
 ## Sample Configuration
 
 ```yaml
-api_key: ${env.GEMINI_API_KEY}
+api_key: ${env.GEMINI_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_groq-openai-compat.md

@@ -15,7 +15,7 @@ Groq OpenAI-compatible provider for using Groq models with OpenAI API format.
 
 ```yaml
 openai_compat_api_base: https://api.groq.com/openai/v1
-api_key: ${env.GROQ_API_KEY}
+api_key: ${env.GROQ_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_groq.md

@@ -15,7 +15,7 @@ Groq inference provider for ultra-fast inference using Groq's LPU technology.
 
 ```yaml
 url: https://api.groq.com
-api_key: ${env.GROQ_API_KEY}
+api_key: ${env.GROQ_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_llama-openai-compat.md

@@ -15,7 +15,7 @@ Llama OpenAI-compatible provider for using Llama models with OpenAI API format.
 
 ```yaml
 openai_compat_api_base: https://api.llama.com/compat/v1/
-api_key: ${env.LLAMA_API_KEY}
+api_key: ${env.LLAMA_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_openai.md

@@ -13,7 +13,7 @@ OpenAI inference provider for accessing GPT models and other OpenAI services.
 ## Sample Configuration
 
 ```yaml
-api_key: ${env.OPENAI_API_KEY}
+api_key: ${env.OPENAI_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_passthrough.md

@@ -15,7 +15,7 @@ Passthrough inference provider for connecting to any external inference service
 
 ```yaml
 url: ${env.PASSTHROUGH_URL}
-api_key: ${env.PASSTHROUGH_API_KEY}
+api_key: ${env.PASSTHROUGH_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_sambanova-openai-compat.md

@@ -15,7 +15,7 @@ SambaNova OpenAI-compatible provider for using SambaNova models with OpenAI API
 
 ```yaml
 openai_compat_api_base: https://api.sambanova.ai/v1
-api_key: ${env.SAMBANOVA_API_KEY}
+api_key: ${env.SAMBANOVA_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_sambanova.md

@@ -15,7 +15,7 @@ SambaNova inference provider for running models on SambaNova's dataflow architec
 
 ```yaml
 url: https://api.sambanova.ai/v1
-api_key: ${env.SAMBANOVA_API_KEY}
+api_key: ${env.SAMBANOVA_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_together-openai-compat.md

@@ -15,7 +15,7 @@ Together AI OpenAI-compatible provider for using Together models with OpenAI API
 
 ```yaml
 openai_compat_api_base: https://api.together.xyz/v1
-api_key: ${env.TOGETHER_API_KEY}
+api_key: ${env.TOGETHER_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_together.md

@@ -15,7 +15,7 @@ Together AI inference provider for open-source models and collaborative AI devel
 
 ```yaml
 url: https://api.together.xyz/v1
-api_key: ${env.TOGETHER_API_KEY}
+api_key: ${env.TOGETHER_API_KEY:=}
 
 ```
 

docs/source/providers/inference/remote_vllm.md

@@ -16,7 +16,7 @@ Remote vLLM inference provider for connecting to vLLM servers.
 ## Sample Configuration
 
 ```yaml
-url: ${env.VLLM_URL}
+url: ${env.VLLM_URL:=http://localhost:8000/v1}
 max_tokens: ${env.VLLM_MAX_TOKENS:=4096}
 api_token: ${env.VLLM_API_TOKEN:=fake}
 tls_verify: ${env.VLLM_TLS_VERIFY:=true}

docs/source/providers/safety/remote_sambanova.md

@@ -15,7 +15,7 @@ SambaNova's safety provider for content moderation and safety filtering.
 
 ```yaml
 url: https://api.sambanova.ai/v1
-api_key: ${env.SAMBANOVA_API_KEY}
+api_key: ${env.SAMBANOVA_API_KEY:=}
 
 ```
 

docs/source/providers/vector_io/remote_qdrant.md

@@ -24,7 +24,7 @@ Please refer to the inline provider documentation.
 ## Sample Configuration
 
 ```yaml
-api_key: ${env.QDRANT_API_KEY}
+api_key: ${env.QDRANT_API_KEY:=}
 
 ```
 

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

@@ -26,7 +26,7 @@ class AnthropicConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.ANTHROPIC_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.ANTHROPIC_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "api_key": api_key,
         }

llama_stack/providers/remote/inference/cerebras/cerebras.py

@@ -65,9 +65,10 @@ class CerebrasInferenceAdapter(
         )
         self.config = config
 
+        api_key = self.config.api_key.get_secret_value() if self.config.api_key else ""
         self.client = AsyncCerebras(
             base_url=self.config.base_url,
-            api_key=self.config.api_key.get_secret_value(),
+            api_key=api_key,
         )
 
     async def initialize(self) -> None:

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

@@ -26,7 +26,7 @@ class CerebrasImplConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.CEREBRAS_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.CEREBRAS_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "base_url": DEFAULT_BASE_URL,
             "api_key": api_key,

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

@@ -31,7 +31,7 @@ class CerebrasCompatConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.CEREBRAS_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.CEREBRAS_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "openai_compat_api_base": "https://api.cerebras.ai/v1",
             "api_key": api_key,

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

@@ -25,8 +25,8 @@ class DatabricksImplConfig(BaseModel):
     @classmethod
     def sample_run_config(
         cls,
-        url: str = "${env.DATABRICKS_URL}",
-        api_token: str = "${env.DATABRICKS_API_TOKEN}",
+        url: str = "${env.DATABRICKS_URL:=}",
+        api_token: str = "${env.DATABRICKS_API_TOKEN:=}",
         **kwargs: Any,
     ) -> dict[str, Any]:
         return {

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

@@ -23,7 +23,7 @@ class FireworksImplConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.FIREWORKS_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.FIREWORKS_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "url": "https://api.fireworks.ai/inference/v1",
             "api_key": api_key,

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

@@ -96,6 +96,7 @@ class FireworksInferenceAdapter(ModelRegistryHelper, Inference, NeedsRequestProv
 
     def _get_client(self) -> Fireworks:
         fireworks_api_key = self._get_api_key()
+        print(f">>>>>> fireworks_api_key: {fireworks_api_key} <<<<<")
         return Fireworks(api_key=fireworks_api_key)
 
     def _get_openai_client(self) -> AsyncOpenAI:

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

@@ -31,7 +31,7 @@ class FireworksCompatConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.FIREWORKS_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.FIREWORKS_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "openai_compat_api_base": "https://api.fireworks.ai/inference/v1",
             "api_key": api_key,

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

@@ -26,7 +26,7 @@ class GeminiConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.GEMINI_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.GEMINI_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "api_key": api_key,
         }

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

@@ -32,7 +32,7 @@ class GroqConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.GROQ_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.GROQ_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "url": "https://api.groq.com",
             "api_key": api_key,

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

@@ -31,7 +31,7 @@ class GroqCompatConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.GROQ_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.GROQ_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "openai_compat_api_base": "https://api.groq.com/openai/v1",
             "api_key": api_key,

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

@@ -31,7 +31,7 @@ class LlamaCompatConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.LLAMA_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.LLAMA_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "openai_compat_api_base": "https://api.llama.com/compat/v1/",
             "api_key": api_key,

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

@@ -26,7 +26,7 @@ class OpenAIConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.OPENAI_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.OPENAI_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "api_key": api_key,
         }

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

@@ -25,7 +25,7 @@ class PassthroughImplConfig(BaseModel):
 
     @classmethod
     def sample_run_config(
-        cls, url: str = "${env.PASSTHROUGH_URL}", api_key: str = "${env.PASSTHROUGH_API_KEY}", **kwargs
+        cls, url: str = "${env.PASSTHROUGH_URL}", api_key: str = "${env.PASSTHROUGH_API_KEY:=}", **kwargs
     ) -> dict[str, Any]:
         return {
             "url": url,

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

@@ -30,7 +30,7 @@ class SambaNovaImplConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.SAMBANOVA_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.SAMBANOVA_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "url": "https://api.sambanova.ai/v1",
             "api_key": api_key,

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

@@ -31,7 +31,7 @@ class SambaNovaCompatConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.SAMBANOVA_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.SAMBANOVA_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "openai_compat_api_base": "https://api.sambanova.ai/v1",
             "api_key": api_key,

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

@@ -26,5 +26,5 @@ class TogetherImplConfig(BaseModel):
     def sample_run_config(cls, **kwargs) -> dict[str, Any]:
         return {
             "url": "https://api.together.xyz/v1",
-            "api_key": "${env.TOGETHER_API_KEY}",
+            "api_key": "${env.TOGETHER_API_KEY:=}",
         }

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

@@ -31,7 +31,7 @@ class TogetherCompatConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.TOGETHER_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.TOGETHER_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "openai_compat_api_base": "https://api.together.xyz/v1",
             "api_key": api_key,

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

@@ -46,7 +46,7 @@ class VLLMInferenceAdapterConfig(BaseModel):
     @classmethod
     def sample_run_config(
         cls,
-        url: str = "${env.VLLM_URL}",
+        url: str = "${env.VLLM_URL:=http://localhost:8000/v1}",
         **kwargs,
     ):
         return {

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

@@ -30,7 +30,7 @@ class SambaNovaSafetyConfig(BaseModel):
     )
 
     @classmethod
-    def sample_run_config(cls, api_key: str = "${env.SAMBANOVA_API_KEY}", **kwargs) -> dict[str, Any]:
+    def sample_run_config(cls, api_key: str = "${env.SAMBANOVA_API_KEY:=}", **kwargs) -> dict[str, Any]:
         return {
             "url": "https://api.sambanova.ai/v1",
             "api_key": api_key,

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

@@ -27,5 +27,5 @@ class QdrantVectorIOConfig(BaseModel):
     @classmethod
     def sample_run_config(cls, **kwargs: Any) -> dict[str, Any]:
         return {
-            "api_key": "${env.QDRANT_API_KEY}",
+            "api_key": "${env.QDRANT_API_KEY:=}",
         }

llama_stack/templates/open-benchmark/run.yaml

@@ -33,7 +33,7 @@ providers:
     provider_type: remote::together
     config:
       url: https://api.together.xyz/v1
-      api_key: ${env.TOGETHER_API_KEY}
+      api_key: ${env.TOGETHER_API_KEY:=}
   vector_io:
   - provider_id: sqlite-vec
     provider_type: inline::sqlite-vec

llama_stack/templates/starter/build.yaml

@@ -6,27 +6,13 @@ distribution_spec:
     - remote::cerebras
     - remote::ollama
     - remote::vllm
-    - remote::tgi
-    - remote::hf::serverless
-    - remote::hf::endpoint
     - remote::fireworks
     - remote::together
-    - remote::bedrock
-    - remote::databricks
-    - remote::nvidia
-    - remote::runpod
     - remote::openai
     - remote::anthropic
     - remote::gemini
     - remote::groq
-    - remote::fireworks-openai-compat
-    - remote::llama-openai-compat
-    - remote::together-openai-compat
-    - remote::groq-openai-compat
-    - remote::sambanova-openai-compat
-    - remote::cerebras-openai-compat
     - remote::sambanova
-    - remote::passthrough
     - inline::sentence-transformers
     vector_io:
     - inline::faiss

llama_stack/templates/starter/starter.py

@@ -113,6 +113,19 @@ from llama_stack.templates.template import (
     get_model_registry,
 )
 
+REMOTE_INFERENCE_PROVIDERS_FOR_STARTER = {
+    "anthropic",
+    "cerebras",
+    "fireworks",
+    "gemini",
+    "groq",
+    "ollama",
+    "openai",
+    "sambanova",
+    "together",
+    "vllm",
+}
+
 
 def _get_model_entries_for_provider(provider_type: str) -> list[ProviderModelEntry]:
     """Get model entries for a specific provider type."""
@@ -207,41 +220,27 @@ def get_remote_inference_providers() -> tuple[list[Provider], dict[str, list[Pro
     remote_providers = [
         provider
         for provider in all_providers
-        # TODO: re-add once the Python 3.13 issue is fixed
-        # discussion: https://github.com/meta-llama/llama-stack/pull/2327#discussion_r2156883828
-        if hasattr(provider, "adapter") and provider.adapter.adapter_type != "watsonx"
+        if hasattr(provider, "adapter") and provider.adapter.adapter_type in REMOTE_INFERENCE_PROVIDERS_FOR_STARTER
     ]
 
-    providers = []
+    inference_providers = []
     available_models = {}
 
     for provider_spec in remote_providers:
         provider_type = provider_spec.adapter.adapter_type
 
-        # Build the environment variable name for enabling this provider
-        env_var = f"ENABLE_{provider_type.upper().replace('-', '_').replace('::', '_')}"
         model_entries = _get_model_entries_for_provider(provider_type)
         config = _get_config_for_provider(provider_spec)
-        providers.append(
-            (
-                f"${{env.{env_var}:=__disabled__}}",
-                provider_type,
-                model_entries,
-                config,
-            )
-        )
-        available_models[f"${{env.{env_var}:=__disabled__}}"] = model_entries
 
-    inference_providers = []
-    for provider_id, provider_type, model_entries, config in providers:
         inference_providers.append(
             Provider(
-                provider_id=provider_id,
+                provider_id=provider_type,
                 provider_type=f"remote::{provider_type}",
                 config=config,
             )
         )
-        available_models[provider_id] = model_entries
+        available_models[provider_type] = model_entries
+
     return inference_providers, available_models