Merge bfe06ae00f into sapling-pr-archive-ehhuang

llama_stack/core/datatypes.py

@@ -589,6 +589,7 @@ can be instantiated multiple times (with different configs) if necessary.
         _ensure_backend(stores.inference, sql_backends, "storage.stores.inference")
         _ensure_backend(stores.conversations, sql_backends, "storage.stores.conversations")
         _ensure_backend(stores.responses, sql_backends, "storage.stores.responses")
+        _ensure_backend(stores.prompts, kv_backends, "storage.stores.prompts")
         return self
 
 

llama_stack/core/prompts/prompts.py

@@ -11,7 +11,6 @@ from pydantic import BaseModel
 
 from llama_stack.apis.prompts import ListPromptsResponse, Prompt, Prompts
 from llama_stack.core.datatypes import StackRunConfig
-from llama_stack.core.storage.datatypes import KVStoreReference
 from llama_stack.providers.utils.kvstore import KVStore, kvstore_impl
 
 
@@ -40,11 +39,10 @@ class PromptServiceImpl(Prompts):
         self.kvstore: KVStore
 
     async def initialize(self) -> None:
-        # Use metadata store backend with prompts-specific namespace
-        metadata_ref = self.config.run_config.storage.stores.metadata
-        if not metadata_ref:
-            raise ValueError("storage.stores.metadata must be configured in run config")
-        prompts_ref = KVStoreReference(namespace="prompts", backend=metadata_ref.backend)
+        # Use prompts store reference from run config
+        prompts_ref = self.config.run_config.storage.stores.prompts
+        if not prompts_ref:
+            raise ValueError("storage.stores.prompts must be configured in run config")
         self.kvstore = await kvstore_impl(prompts_ref)
 
     def _get_default_key(self, prompt_id: str) -> str:

llama_stack/core/stack.py

@@ -563,6 +563,7 @@ def run_config_from_adhoc_config_spec(
                 metadata=KVStoreReference(backend="kv_default", namespace="registry"),
                 inference=InferenceStoreReference(backend="sql_default", table_name="inference_store"),
                 conversations=SqlStoreReference(backend="sql_default", table_name="openai_conversations"),
+                prompts=KVStoreReference(backend="kv_default", namespace="prompts"),
             ),
         ),
     )

llama_stack/core/storage/datatypes.py

@@ -271,6 +271,10 @@ class ServerStoresConfig(BaseModel):
         default=None,
         description="Responses store configuration (uses SQL backend)",
     )
+    prompts: KVStoreReference | None = Field(
+        default=None,
+        description="Prompts store configuration (uses KV backend)",
+    )
 
 
 class StorageConfig(BaseModel):

llama_stack/distributions/ci-tests/run.yaml

@@ -247,6 +247,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models: []
   shields:

llama_stack/distributions/dell/run-with-safety.yaml

@@ -109,6 +109,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models:
   - metadata: {}

llama_stack/distributions/dell/run.yaml

@@ -105,6 +105,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models:
   - metadata: {}

llama_stack/distributions/meta-reference-gpu/run-with-safety.yaml

@@ -122,6 +122,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models:
   - metadata: {}

llama_stack/distributions/meta-reference-gpu/run.yaml

@@ -112,6 +112,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models:
   - metadata: {}

llama_stack/distributions/nvidia/run-with-safety.yaml

@@ -111,6 +111,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models:
   - metadata: {}

llama_stack/distributions/nvidia/run.yaml

@@ -100,6 +100,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models: []
   shields: []

llama_stack/distributions/open-benchmark/run.yaml

@@ -142,6 +142,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models:
   - metadata: {}

llama_stack/distributions/postgres-demo/run.yaml

@@ -87,6 +87,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models:
   - metadata: {}

llama_stack/distributions/starter-gpu/run.yaml

@@ -250,6 +250,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models: []
   shields:

llama_stack/distributions/starter/run.yaml

@@ -247,6 +247,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models: []
   shields:

llama_stack/distributions/template.py

@@ -259,6 +259,10 @@ class RunConfigSettings(BaseModel):
                 backend="sql_default",
                 table_name="openai_conversations",
             ).model_dump(exclude_none=True),
+            "prompts": KVStoreReference(
+                backend="kv_default",
+                namespace="prompts",
+            ).model_dump(exclude_none=True),
         }
 
         storage_config = dict(

llama_stack/distributions/watsonx/run.yaml

@@ -115,6 +115,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models: []
   shields: []