feat: split API and provider specs into separate llama-stack-api pkg

Extract API definitions, models, and provider specifications into a
standalone llama-stack-api package that can be published to PyPI
independently of the main llama-stack server.

Motivation

External providers currently import from llama-stack, which overrides
the installed version and causes dependency conflicts. This separation
allows external providers to:

- Install only the type definitions they need without server dependencies
- Avoid version conflicts with the installed llama-stack package
- Be versioned and released independently

This enables us to re-enable external provider module tests that were
previously blocked by these import conflicts.

Changes

- Created llama-stack-api package with minimal dependencies (pydantic, jsonschema)
- Moved APIs, providers datatypes, strong_typing, and schema_utils
- Updated all imports from llama_stack.* to llama_stack_api.*
- Preserved git history using git mv for moved files
- Configured local editable install for development workflow
- Updated linting and type-checking configuration for both packages
- Rebased on top of upstream src/ layout changes

Testing

Package builds successfully and can be imported independently.
All pre-commit hooks pass with expected exclusions maintained.

Next Steps

- Publish llama-stack-api to PyPI
- Update external provider dependencies
- Re-enable external provider module tests

Signed-off-by: Charlie Doern <cdoern@redhat.com>

tests/unit/server/test_access_control.py

@@ -8,10 +8,10 @@ from unittest.mock import MagicMock, Mock, patch
 
 import pytest
 import yaml
+from llama_stack_api.apis.datatypes import Api
+from llama_stack_api.apis.models import ModelType
 from pydantic import TypeAdapter, ValidationError
 
-from llama_stack.apis.datatypes import Api
-from llama_stack.apis.models import ModelType
 from llama_stack.core.access_control.access_control import AccessDeniedError, is_action_allowed
 from llama_stack.core.datatypes import AccessRule, ModelWithOwner, User
 from llama_stack.core.routing_tables.models import ModelsRoutingTable

tests/unit/server/test_auth.py

@@ -144,7 +144,7 @@ def middleware_with_mocks(mock_auth_endpoint):
     middleware = AuthenticationMiddleware(mock_app, auth_config, {})
 
     # Mock the route_impls to simulate finding routes with required scopes
-    from llama_stack.schema_utils import WebMethod
+    from llama_stack_api.schema_utils import WebMethod
 
     routes = {
         ("POST", "/test/scoped"): WebMethod(route="/test/scoped", method="POST", required_scope="test.read"),

tests/unit/server/test_resolver.py

@@ -9,9 +9,10 @@ import sys
 from typing import Any, Protocol
 from unittest.mock import AsyncMock, MagicMock
 
+from llama_stack_api.apis.inference import Inference
+from llama_stack_api.providers.datatypes import InlineProviderSpec, ProviderSpec
 from pydantic import BaseModel, Field
 
-from llama_stack.apis.inference import Inference
 from llama_stack.core.datatypes import Api, Provider, StackRunConfig
 from llama_stack.core.resolver import resolve_impls
 from llama_stack.core.routers.inference import InferenceRouter
@@ -25,7 +26,6 @@ from llama_stack.core.storage.datatypes import (
     SqlStoreReference,
     StorageConfig,
 )
-from llama_stack.providers.datatypes import InlineProviderSpec, ProviderSpec
 from llama_stack.providers.utils.kvstore import register_kvstore_backends
 from llama_stack.providers.utils.sqlstore.sqlstore import register_sqlstore_backends
 

tests/unit/server/test_sse.py

@@ -9,8 +9,8 @@ import logging  # allow-direct-logging
 from unittest.mock import AsyncMock, MagicMock
 
 import pytest
+from llama_stack_api.apis.common.responses import PaginatedResponse
 
-from llama_stack.apis.common.responses import PaginatedResponse
 from llama_stack.core.server.server import create_dynamic_typed_route, create_sse_event, sse_generator