LCORE-1216: Bump up to llama-stack 0.4.3

Dockerfile

@@ -31,7 +31,6 @@ RUN uv pip install -e .
 
 # Verify llama command is available
 RUN uv run python -c "import llama_stack; print('llama-stack imported successfully')"
-RUN uv run llama --help
 
 # Expose port
 EXPOSE 8321

README.md

@@ -35,7 +35,7 @@ Manual procedure, assuming an existing PyPI API token available:
 ### Prerequisites
 
 - Python >= 3.12
-- Llama Stack == 0.2.22
+- Llama Stack == 0.4.3
 - pydantic >= 2.10.6
 
 ### Installation
@@ -163,13 +163,13 @@ shields:
 
 ```bash
 # Test the redaction shield
-curl -X POST "http://localhost:8321/v1/safety/run_shield" \
+curl -X POST "http://localhost:8321/v1/safety/run-shield" \
   -H "Content-Type: application/json" \
   -d '{
     "shield_id": "redaction-shield",
     "messages": [
       {
-        "role": "user", 
+        "role": "user",
         "content": "My API key is abc123xyz and password is secret456"
       }
     ]
@@ -180,11 +180,8 @@ curl -X POST "http://localhost:8321/v1/safety/run_shield" \
 1. **Create provider directory**
    ```bash
    mkdir -p ./providers.d/inline/safety/
-   mkdir -p ./providers.d/remote/tool_runtime/
    curl -o ./providers.d/inline/safety/lightspeed_question_validity.yaml https://raw.githubusercontent.com/lightspeed-core/lightspeed-            providers/refs/heads/main/resources/external_providers/inline/safety/lightspeed_question_validity.yaml
-   curl -o ./providers.d/inline/safety/lightspeed_question_validity.yaml https://raw.githubusercontent.com/lightspeed-core/lightspeed-            providers/refs/heads/main/resources/external_providers/inline/safety/lightspeed_redaction.yaml
-   curl -o ./providers.d/remote/tool_runtime/lightspeed.yaml https://raw.githubusercontent.com/lightspeed-core/lightspeed-providers/refs/heads/main/resources/external_providers/remote/tool_runtime/lightspeed.yaml
-
+   curl -o ./providers.d/inline/safety/lightspeed_question_validity.yaml https://raw.githubusercontent.com/lightspeed-core/lightspeed-            providers/refs/heads/main/resources/external_providers/inline/safety/lightspeed_redaction.yaml   
    ```
 3. **Add external provider definition**
    ```yaml

lightspeed_stack_providers/providers/inline/agents/lightspeed_inline_agent/README.md

@@ -17,30 +17,32 @@ agents:
   - provider_id: lightspeed_inline_agent
     provider_type: inline::lightspeed_inline_agent
     config:
-      persistence_store:
-        type: sqlite
-        namespace: null
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/agents_store.db
-      responses_store:
-        type: sqlite
-        db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/responses_store.db
+      persistence:
+        agent_state:
+          namespace: lightspeed_agents
+          backend: kv_default
+        responses:
+          table_name: lightspeed_responses
+          backend: sql_default
       tools_filter:
-        # Optional: Whether to enable tools filtering, default value is true  
+        # Optional: Whether to enable tools filtering, default value is true
         enabled: true
         # Optional: The model to use for filtering, the default value is the inference model used
         model_id: ${env.INFERENCE_MODEL_FILTER:=}
-        # Optional: From how much tools we start filtering, default value is 10 
+        # Optional: From how much tools we start filtering, default value is 10
         min_tools: 10
         # Optional: the file path of the system prompt, default value is None
         system_prompt_path: ${env.FILTER_SYSTEM_PROMPT_PATH:=}
-        # Optional: the system prompt if not in a file, 
+        # Optional: the system prompt if not in a file,
         # when system_prompt_path is defined system_prompt will be the content of the indicated file
-        # when system_prompt is empty, the default filtering system prompt is used 
+        # when system_prompt is empty, the default filtering system prompt is used
         system_prompt: ${env.FILTER_SYSTEM_PROMPT:=}
-        # Optional: tools to always include, for example rag tools, to not be filtered out, 
+        # Optional: tools to always include, for example rag tools, to not be filtered out,
         # the default is an empty list
         always_include_tools:
-        - knowledge_search
+          - knowledge_search
+      # Optional: Override temperature for this agent (default: None)
+      chatbot_temperature_override: 1.0
 ...
 external_providers_dir: ~/.llama/distributions/ollama/external_providers/
 ```

lightspeed_stack_providers/providers/inline/agents/lightspeed_inline_agent/__init__.py

@@ -1,11 +1,16 @@
 from typing import Any
 
-from llama_stack.providers.datatypes import Api
+from llama_stack.core.datatypes import AccessRule
+from llama_stack_api import Api
 
 from .config import LightspeedAgentsImplConfig
 
 
-async def get_provider_impl(config: LightspeedAgentsImplConfig, deps: dict[Api, Any]):
+async def get_provider_impl(
+    config: LightspeedAgentsImplConfig,
+    deps: dict[Api, Any],
+    policy: list[AccessRule],
+):
     # Configure litellm to drop unsupported params for models that reject them (e.g., top_p).
     # This is safe to set globally since it only affects models that don't support these params.
     import litellm
@@ -18,10 +23,13 @@ async def get_provider_impl(config: LightspeedAgentsImplConfig, deps: dict[Api,
         config,
         deps[Api.inference],
         deps[Api.vector_io],
-        deps[Api.safety],
+        deps.get(Api.safety),
         deps[Api.tool_runtime],
         deps[Api.tool_groups],
-        [],
+        deps[Api.conversations],
+        deps[Api.prompts],
+        deps[Api.files],
+        policy,
     )
     await impl.initialize()
     return impl