feat: add (openai, anthropic, gemini) providers via litellm (#1267)

# What does this PR do?

This PR introduces more non-llama model support to llama stack.
Providers introduced: openai, anthropic and gemini. All of these
providers use essentially the same piece of code -- the implementation
works via the `litellm` library.

We will expose only specific models for providers we enable making sure
they all work well and pass tests. This setup (instead of automatically
enabling _all_ providers and models allowed by LiteLLM) ensures we can
also perform any needed prompt tuning on a per-model basis as needed
(just like we do it for llama models.)

## Test Plan

```bash
#!/bin/bash

args=("$@")
for model in openai/gpt-4o anthropic/claude-3-5-sonnet-latest gemini/gemini-1.5-flash; do
    LLAMA_STACK_CONFIG=dev pytest -s -v tests/client-sdk/inference/test_text_inference.py \
        --embedding-model=all-MiniLM-L6-v2 \
        --vision-inference-model="" \
        --inference-model=$model "${args[@]}"
done
```

llama_stack/templates/ci-tests/ci_tests.py

@@ -57,17 +57,6 @@ def get_distribution_template() -> DistributionTemplate:
         config=SentenceTransformersInferenceConfig.sample_run_config(),
     )
 
-    core_model_to_hf_repo = {m.descriptor(): m.huggingface_repo for m in all_registered_models()}
-    default_models = [
-        ModelInput(
-            model_id=core_model_to_hf_repo[m.llama_model] if m.llama_model else m.provider_model_id,
-            provider_model_id=m.provider_model_id,
-            provider_id="fireworks",
-            metadata=m.metadata,
-            model_type=m.model_type,
-        )
-        for m in MODEL_ENTRIES
-    ]
     default_tool_groups = [
         ToolGroupInput(
             toolgroup_id="builtin::websearch",
@@ -82,6 +71,16 @@ def get_distribution_template() -> DistributionTemplate:
             provider_id="code-interpreter",
         ),
     ]
+    core_model_to_hf_repo = {m.descriptor(): m.huggingface_repo for m in all_registered_models()}
+    default_models = [
+        ModelInput(
+            model_id=core_model_to_hf_repo[m.llama_model] if m.llama_model else m.provider_model_id,
+            provider_id="fireworks",
+            model_type=m.model_type,
+            metadata=m.metadata,
+        )
+        for m in MODEL_ENTRIES
+    ]
     embedding_model = ModelInput(
         model_id="all-MiniLM-L6-v2",
         provider_id="sentence-transformers",
@@ -98,7 +97,7 @@ def get_distribution_template() -> DistributionTemplate:
         container_image=None,
         template_path=None,
         providers=providers,
-        default_models=default_models,
+        default_models=default_models + [embedding_model],
         run_configs={
             "run.yaml": RunConfigSettings(
                 provider_overrides={

llama_stack/templates/ci-tests/run.yaml

@@ -93,59 +93,48 @@ models:
 - metadata: {}
   model_id: meta-llama/Llama-3.1-8B-Instruct
   provider_id: fireworks
-  provider_model_id: accounts/fireworks/models/llama-v3p1-8b-instruct
   model_type: llm
 - metadata: {}
   model_id: meta-llama/Llama-3.1-70B-Instruct
   provider_id: fireworks
-  provider_model_id: accounts/fireworks/models/llama-v3p1-70b-instruct
   model_type: llm
 - metadata: {}
   model_id: meta-llama/Llama-3.1-405B-Instruct-FP8
   provider_id: fireworks
-  provider_model_id: accounts/fireworks/models/llama-v3p1-405b-instruct
   model_type: llm
 - metadata: {}
   model_id: meta-llama/Llama-3.2-1B-Instruct
   provider_id: fireworks
-  provider_model_id: accounts/fireworks/models/llama-v3p2-1b-instruct
   model_type: llm
 - metadata: {}
   model_id: meta-llama/Llama-3.2-3B-Instruct
   provider_id: fireworks
-  provider_model_id: accounts/fireworks/models/llama-v3p2-3b-instruct
   model_type: llm
 - metadata: {}
   model_id: meta-llama/Llama-3.2-11B-Vision-Instruct
   provider_id: fireworks
-  provider_model_id: accounts/fireworks/models/llama-v3p2-11b-vision-instruct
   model_type: llm
 - metadata: {}
   model_id: meta-llama/Llama-3.2-90B-Vision-Instruct
   provider_id: fireworks
-  provider_model_id: accounts/fireworks/models/llama-v3p2-90b-vision-instruct
   model_type: llm
 - metadata: {}
   model_id: meta-llama/Llama-3.3-70B-Instruct
   provider_id: fireworks
-  provider_model_id: accounts/fireworks/models/llama-v3p3-70b-instruct
   model_type: llm
 - metadata: {}
   model_id: meta-llama/Llama-Guard-3-8B
   provider_id: fireworks
-  provider_model_id: accounts/fireworks/models/llama-guard-3-8b
   model_type: llm
 - metadata: {}
   model_id: meta-llama/Llama-Guard-3-11B-Vision
   provider_id: fireworks
-  provider_model_id: accounts/fireworks/models/llama-guard-3-11b-vision
   model_type: llm
 - metadata:
     embedding_dimension: 768
     context_length: 8192
   model_id: nomic-ai/nomic-embed-text-v1.5
   provider_id: fireworks
-  provider_model_id: nomic-ai/nomic-embed-text-v1.5
   model_type: embedding
 - metadata:
     embedding_dimension: 384

llama_stack/templates/dev/__init__.py

@@ -0,0 +1,7 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from .dev import get_distribution_template  # noqa: F401

llama_stack/templates/dev/build.yaml

@@ -0,0 +1,36 @@
+version: '2'
+distribution_spec:
+  description: Distribution for running e2e tests in CI
+  providers:
+    inference:
+    - remote::openai
+    - remote::fireworks
+    - remote::anthropic
+    - remote::gemini
+    - inline::sentence-transformers
+    vector_io:
+    - inline::sqlite-vec
+    - remote::chromadb
+    - remote::pgvector
+    safety:
+    - inline::llama-guard
+    agents:
+    - inline::meta-reference
+    telemetry:
+    - inline::meta-reference
+    eval:
+    - inline::meta-reference
+    datasetio:
+    - remote::huggingface
+    - inline::localfs
+    scoring:
+    - inline::basic
+    - inline::llm-as-judge
+    - inline::braintrust
+    tool_runtime:
+    - remote::brave-search
+    - remote::tavily-search
+    - inline::code-interpreter
+    - inline::rag-runtime
+    - remote::model-context-protocol
+image_type: conda

llama_stack/templates/dev/dev.py

@@ -0,0 +1,174 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from typing import List, Tuple
+
+from llama_stack.apis.models.models import ModelType
+from llama_stack.distribution.datatypes import (
+    ModelInput,
+    Provider,
+    ShieldInput,
+    ToolGroupInput,
+)
+from llama_stack.models.llama.sku_list import all_registered_models
+from llama_stack.providers.inline.inference.sentence_transformers import (
+    SentenceTransformersInferenceConfig,
+)
+from llama_stack.providers.inline.vector_io.sqlite_vec.config import SQLiteVectorIOConfig
+from llama_stack.providers.remote.inference.anthropic.config import AnthropicConfig
+from llama_stack.providers.remote.inference.anthropic.models import MODEL_ENTRIES as ANTHROPIC_MODEL_ENTRIES
+from llama_stack.providers.remote.inference.fireworks.config import FireworksImplConfig
+from llama_stack.providers.remote.inference.fireworks.models import MODEL_ENTRIES as FIREWORKS_MODEL_ENTRIES
+from llama_stack.providers.remote.inference.gemini.config import GeminiConfig
+from llama_stack.providers.remote.inference.gemini.models import MODEL_ENTRIES as GEMINI_MODEL_ENTRIES
+from llama_stack.providers.remote.inference.openai.config import OpenAIConfig
+from llama_stack.providers.remote.inference.openai.models import MODEL_ENTRIES as OPENAI_MODEL_ENTRIES
+from llama_stack.templates.template import DistributionTemplate, RunConfigSettings
+
+
+def get_inference_providers() -> Tuple[List[Provider], List[ModelInput]]:
+    # in this template, we allow each API key to be optional
+    providers = [
+        (
+            "openai",
+            OPENAI_MODEL_ENTRIES,
+            OpenAIConfig.sample_run_config(api_key="${env.OPENAI_API_KEY:}"),
+        ),
+        (
+            "fireworks",
+            FIREWORKS_MODEL_ENTRIES,
+            FireworksImplConfig.sample_run_config(api_key="${env.FIREWORKS_API_KEY:}"),
+        ),
+        (
+            "anthropic",
+            ANTHROPIC_MODEL_ENTRIES,
+            AnthropicConfig.sample_run_config(api_key="${env.ANTHROPIC_API_KEY:}"),
+        ),
+        (
+            "gemini",
+            GEMINI_MODEL_ENTRIES,
+            GeminiConfig.sample_run_config(api_key="${env.GEMINI_API_KEY:}"),
+        ),
+    ]
+    inference_providers = []
+    default_models = []
+    core_model_to_hf_repo = {m.descriptor(): m.huggingface_repo for m in all_registered_models()}
+    for provider_id, model_entries, config in providers:
+        inference_providers.append(
+            Provider(
+                provider_id=provider_id,
+                provider_type=f"remote::{provider_id}",
+                config=config,
+            )
+        )
+        default_models.extend(
+            ModelInput(
+                model_id=core_model_to_hf_repo[m.llama_model] if m.llama_model else m.provider_model_id,
+                provider_model_id=m.provider_model_id,
+                provider_id=provider_id,
+                model_type=m.model_type,
+                metadata=m.metadata,
+            )
+            for m in model_entries
+        )
+    return inference_providers, default_models
+
+
+def get_distribution_template() -> DistributionTemplate:
+    providers = {
+        "inference": [
+            "remote::openai",
+            "remote::fireworks",
+            "remote::anthropic",
+            "remote::gemini",
+            "inline::sentence-transformers",
+        ],
+        "vector_io": ["inline::sqlite-vec", "remote::chromadb", "remote::pgvector"],
+        "safety": ["inline::llama-guard"],
+        "agents": ["inline::meta-reference"],
+        "telemetry": ["inline::meta-reference"],
+        "eval": ["inline::meta-reference"],
+        "datasetio": ["remote::huggingface", "inline::localfs"],
+        "scoring": ["inline::basic", "inline::llm-as-judge", "inline::braintrust"],
+        "tool_runtime": [
+            "remote::brave-search",
+            "remote::tavily-search",
+            "inline::code-interpreter",
+            "inline::rag-runtime",
+            "remote::model-context-protocol",
+        ],
+    }
+    name = "dev"
+
+    vector_io_provider = Provider(
+        provider_id="sqlite-vec",
+        provider_type="inline::sqlite-vec",
+        config=SQLiteVectorIOConfig.sample_run_config(f"distributions/{name}"),
+    )
+    embedding_provider = Provider(
+        provider_id="sentence-transformers",
+        provider_type="inline::sentence-transformers",
+        config=SentenceTransformersInferenceConfig.sample_run_config(),
+    )
+
+    default_tool_groups = [
+        ToolGroupInput(
+            toolgroup_id="builtin::websearch",
+            provider_id="tavily-search",
+        ),
+        ToolGroupInput(
+            toolgroup_id="builtin::rag",
+            provider_id="rag-runtime",
+        ),
+        ToolGroupInput(
+            toolgroup_id="builtin::code_interpreter",
+            provider_id="code-interpreter",
+        ),
+    ]
+    embedding_model = ModelInput(
+        model_id="all-MiniLM-L6-v2",
+        provider_id=embedding_provider.provider_id,
+        model_type=ModelType.embedding,
+        metadata={
+            "embedding_dimension": 384,
+        },
+    )
+    inference_providers, default_models = get_inference_providers()
+
+    return DistributionTemplate(
+        name=name,
+        distro_type="self_hosted",
+        description="Distribution for running e2e tests in CI",
+        container_image=None,
+        template_path=None,
+        providers=providers,
+        default_models=[],
+        run_configs={
+            "run.yaml": RunConfigSettings(
+                provider_overrides={
+                    "inference": inference_providers + [embedding_provider],
+                    "vector_io": [vector_io_provider],
+                },
+                default_models=default_models + [embedding_model],
+                default_tool_groups=default_tool_groups,
+                default_shields=[ShieldInput(shield_id="meta-llama/Llama-Guard-3-8B")],
+            ),
+        },
+        run_config_env_vars={
+            "LLAMA_STACK_PORT": (
+                "5001",
+                "Port for the Llama Stack distribution server",
+            ),
+            "FIREWORKS_API_KEY": (
+                "",
+                "Fireworks API Key",
+            ),
+            "OPENAI_API_KEY": (
+                "",
+                "OpenAI API Key",
+            ),
+        },
+    )

llama_stack/templates/dev/run.yaml

@@ -0,0 +1,261 @@
+version: '2'
+image_name: dev
+apis:
+- agents
+- datasetio
+- eval
+- inference
+- safety
+- scoring
+- telemetry
+- tool_runtime
+- vector_io
+providers:
+  inference:
+  - provider_id: openai
+    provider_type: remote::openai
+    config:
+      api_key: ${env.OPENAI_API_KEY:}
+  - provider_id: fireworks
+    provider_type: remote::fireworks
+    config:
+      url: https://api.fireworks.ai/inference/v1
+      api_key: ${env.FIREWORKS_API_KEY:}
+  - provider_id: anthropic
+    provider_type: remote::anthropic
+    config:
+      api_key: ${env.ANTHROPIC_API_KEY:}
+  - provider_id: gemini
+    provider_type: remote::gemini
+    config:
+      api_key: ${env.GEMINI_API_KEY:}
+  - provider_id: sentence-transformers
+    provider_type: inline::sentence-transformers
+    config: {}
+  vector_io:
+  - provider_id: sqlite-vec
+    provider_type: inline::sqlite-vec
+    config:
+      db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/dev}/sqlite_vec.db
+  safety:
+  - provider_id: llama-guard
+    provider_type: inline::llama-guard
+    config: {}
+  agents:
+  - provider_id: meta-reference
+    provider_type: inline::meta-reference
+    config:
+      persistence_store:
+        type: sqlite
+        namespace: null
+        db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/dev}/agents_store.db
+  telemetry:
+  - provider_id: meta-reference
+    provider_type: inline::meta-reference
+    config:
+      service_name: ${env.OTEL_SERVICE_NAME:llama-stack}
+      sinks: ${env.TELEMETRY_SINKS:console,sqlite}
+      sqlite_db_path: ${env.SQLITE_DB_PATH:~/.llama/distributions/dev/trace_store.db}
+  eval:
+  - provider_id: meta-reference
+    provider_type: inline::meta-reference
+    config: {}
+  datasetio:
+  - provider_id: huggingface
+    provider_type: remote::huggingface
+    config: {}
+  - provider_id: localfs
+    provider_type: inline::localfs
+    config: {}
+  scoring:
+  - provider_id: basic
+    provider_type: inline::basic
+    config: {}
+  - provider_id: llm-as-judge
+    provider_type: inline::llm-as-judge
+    config: {}
+  - provider_id: braintrust
+    provider_type: inline::braintrust
+    config:
+      openai_api_key: ${env.OPENAI_API_KEY:}
+  tool_runtime:
+  - provider_id: brave-search
+    provider_type: remote::brave-search
+    config:
+      api_key: ${env.BRAVE_SEARCH_API_KEY:}
+      max_results: 3
+  - provider_id: tavily-search
+    provider_type: remote::tavily-search
+    config:
+      api_key: ${env.TAVILY_SEARCH_API_KEY:}
+      max_results: 3
+  - provider_id: code-interpreter
+    provider_type: inline::code-interpreter
+    config: {}
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
+    config: {}
+  - provider_id: model-context-protocol
+    provider_type: remote::model-context-protocol
+    config: {}
+metadata_store:
+  type: sqlite
+  db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/dev}/registry.db
+models:
+- metadata: {}
+  model_id: openai/gpt-4o
+  provider_id: openai
+  provider_model_id: openai/gpt-4o
+  model_type: llm
+- metadata: {}
+  model_id: openai/gpt-4o-mini
+  provider_id: openai
+  provider_model_id: openai/gpt-4o-mini
+  model_type: llm
+- metadata: {}
+  model_id: openai/chatgpt-4o-latest
+  provider_id: openai
+  provider_model_id: openai/chatgpt-4o-latest
+  model_type: llm
+- metadata:
+    embedding_dimension: 1536
+  model_id: openai/text-embedding-3-small
+  provider_id: openai
+  provider_model_id: openai/text-embedding-3-small
+  model_type: embedding
+- metadata:
+    embedding_dimension: 3072
+  model_id: openai/text-embedding-3-large
+  provider_id: openai
+  provider_model_id: openai/text-embedding-3-large
+  model_type: embedding
+- metadata: {}
+  model_id: meta-llama/Llama-3.1-8B-Instruct
+  provider_id: fireworks
+  provider_model_id: accounts/fireworks/models/llama-v3p1-8b-instruct
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-3.1-70B-Instruct
+  provider_id: fireworks
+  provider_model_id: accounts/fireworks/models/llama-v3p1-70b-instruct
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-3.1-405B-Instruct-FP8
+  provider_id: fireworks
+  provider_model_id: accounts/fireworks/models/llama-v3p1-405b-instruct
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-3.2-1B-Instruct
+  provider_id: fireworks
+  provider_model_id: accounts/fireworks/models/llama-v3p2-1b-instruct
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-3.2-3B-Instruct
+  provider_id: fireworks
+  provider_model_id: accounts/fireworks/models/llama-v3p2-3b-instruct
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-3.2-11B-Vision-Instruct
+  provider_id: fireworks
+  provider_model_id: accounts/fireworks/models/llama-v3p2-11b-vision-instruct
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-3.2-90B-Vision-Instruct
+  provider_id: fireworks
+  provider_model_id: accounts/fireworks/models/llama-v3p2-90b-vision-instruct
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-3.3-70B-Instruct
+  provider_id: fireworks
+  provider_model_id: accounts/fireworks/models/llama-v3p3-70b-instruct
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-Guard-3-8B
+  provider_id: fireworks
+  provider_model_id: accounts/fireworks/models/llama-guard-3-8b
+  model_type: llm
+- metadata: {}
+  model_id: meta-llama/Llama-Guard-3-11B-Vision
+  provider_id: fireworks
+  provider_model_id: accounts/fireworks/models/llama-guard-3-11b-vision
+  model_type: llm
+- metadata:
+    embedding_dimension: 768
+    context_length: 8192
+  model_id: nomic-ai/nomic-embed-text-v1.5
+  provider_id: fireworks
+  provider_model_id: nomic-ai/nomic-embed-text-v1.5
+  model_type: embedding
+- metadata: {}
+  model_id: anthropic/claude-3-5-sonnet-latest
+  provider_id: anthropic
+  provider_model_id: anthropic/claude-3-5-sonnet-latest
+  model_type: llm
+- metadata: {}
+  model_id: anthropic/claude-3-7-sonnet-latest
+  provider_id: anthropic
+  provider_model_id: anthropic/claude-3-7-sonnet-latest
+  model_type: llm
+- metadata: {}
+  model_id: anthropic/claude-3-5-haiku-latest
+  provider_id: anthropic
+  provider_model_id: anthropic/claude-3-5-haiku-latest
+  model_type: llm
+- metadata:
+    embedding_dimension: 1024
+    context_length: 32000
+  model_id: anthropic/voyage-3
+  provider_id: anthropic
+  provider_model_id: anthropic/voyage-3
+  model_type: embedding
+- metadata:
+    embedding_dimension: 512
+    context_length: 32000
+  model_id: anthropic/voyage-3-lite
+  provider_id: anthropic
+  provider_model_id: anthropic/voyage-3-lite
+  model_type: embedding
+- metadata:
+    embedding_dimension: 1024
+    context_length: 32000
+  model_id: anthropic/voyage-code-3
+  provider_id: anthropic
+  provider_model_id: anthropic/voyage-code-3
+  model_type: embedding
+- metadata: {}
+  model_id: gemini/gemini-1.5-flash
+  provider_id: gemini
+  provider_model_id: gemini/gemini-1.5-flash
+  model_type: llm
+- metadata: {}
+  model_id: gemini/gemini-1.5-pro
+  provider_id: gemini
+  provider_model_id: gemini/gemini-1.5-pro
+  model_type: llm
+- metadata:
+    embedding_dimension: 768
+    context_length: 2048
+  model_id: gemini/text-embedding-004
+  provider_id: gemini
+  provider_model_id: gemini/text-embedding-004
+  model_type: embedding
+- metadata:
+    embedding_dimension: 384
+  model_id: all-MiniLM-L6-v2
+  provider_id: sentence-transformers
+  model_type: embedding
+shields:
+- shield_id: meta-llama/Llama-Guard-3-8B
+vector_dbs: []
+datasets: []
+scoring_fns: []
+benchmarks: []
+tool_groups:
+- toolgroup_id: builtin::websearch
+  provider_id: tavily-search
+- toolgroup_id: builtin::rag
+  provider_id: rag-runtime
+- toolgroup_id: builtin::code_interpreter
+  provider_id: code-interpreter
+server:
+  port: 8321