feat(prompts): attach prompts to storage stores in run configs

.github/workflows/integration-auth-tests.yml

@@ -91,6 +91,9 @@ jobs:
               conversations:
                 table_name: openai_conversations
                 backend: sql_default
+              prompts:
+                namespace: prompts
+                backend: kv_default
           server:
             port: 8321
           EOF

benchmarking/k8s-benchmark/stack-configmap.yaml

@@ -115,13 +115,21 @@ data:
           db: ${env.POSTGRES_DB:=llamastack}
           user: ${env.POSTGRES_USER:=llamastack}
           password: ${env.POSTGRES_PASSWORD:=llamastack}
-      references:
+      stores:
         metadata:
           backend: kv_default
           namespace: registry
         inference:
           backend: sql_default
           table_name: inference_store
+          max_write_queue_size: 10000
+          num_writers: 4
+        conversations:
+          backend: sql_default
+          table_name: openai_conversations
+        prompts:
+          backend: kv_default
+          namespace: prompts
     models:
     - metadata:
         embedding_dimension: 768

benchmarking/k8s-benchmark/stack_run_config.yaml

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

docs/docs/distributions/configuration.mdx

@@ -63,13 +63,21 @@ storage:
     sql_default:
       type: sql_sqlite
       db_path: ${env.SQLITE_STORE_DIR:=~/.llama/distributions/ollama}/sqlstore.db
-  references:
+  stores:
     metadata:
       backend: kv_default
       namespace: registry
     inference:
       backend: sql_default
       table_name: inference_store
+      max_write_queue_size: 10000
+      num_writers: 4
+    conversations:
+      backend: sql_default
+      table_name: openai_conversations
+    prompts:
+      backend: kv_default
+      namespace: prompts
 models:
 - metadata: {}
   model_id: ${env.INFERENCE_MODEL}

docs/docs/distributions/k8s/stack-configmap.yaml

@@ -113,13 +113,21 @@ data:
           db: ${env.POSTGRES_DB:=llamastack}
           user: ${env.POSTGRES_USER:=llamastack}
           password: ${env.POSTGRES_PASSWORD:=llamastack}
-      references:
+      stores:
         metadata:
           backend: kv_default
           namespace: registry
         inference:
           backend: sql_default
           table_name: inference_store
+          max_write_queue_size: 10000
+          num_writers: 4
+        conversations:
+          backend: sql_default
+          table_name: openai_conversations
+        prompts:
+          backend: kv_default
+          namespace: prompts
     models:
     - metadata:
         embedding_dimension: 768

docs/docs/distributions/k8s/stack_run_config.yaml

@@ -113,6 +113,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 registered_resources:
   models:
   - metadata:

llama_stack/core/datatypes.py

@@ -582,6 +582,7 @@ 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

@@ -540,6 +540,7 @@ 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

@@ -257,6 +257,10 @@ 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: []

tests/external/run-byoa.yaml

@@ -25,6 +25,9 @@ storage:
     conversations:
       table_name: openai_conversations
       backend: sql_default
+    prompts:
+      namespace: prompts
+      backend: kv_default
 external_apis_dir: ~/.llama/apis.d
 external_providers_dir: ~/.llama/providers.d
 server:

tests/unit/cli/test_stack_config.py

@@ -44,6 +44,9 @@ def config_with_image_name_int():
             responses:
               backend: sql_default
               table_name: responses
+            prompts:
+              backend: kv_default
+              namespace: prompts
         providers:
           inference:
             - provider_id: provider1

tests/unit/distribution/test_distribution.py

@@ -48,6 +48,7 @@ def _default_storage() -> StorageConfig:
             metadata=KVStoreReference(backend="kv_default", namespace="registry"),
             inference=InferenceStoreReference(backend="sql_default", table_name="inference_store"),
             conversations=SqlStoreReference(backend="sql_default", table_name="conversations"),
+            prompts=KVStoreReference(backend="kv_default", namespace="prompts"),
         ),
     )
 

tests/unit/prompts/prompts/conftest.py

@@ -18,7 +18,7 @@ from llama_stack.core.storage.datatypes import (
     SqlStoreReference,
     StorageConfig,
 )
-from llama_stack.providers.utils.kvstore import kvstore_impl, register_kvstore_backends
+from llama_stack.providers.utils.kvstore import register_kvstore_backends
 
 
 @pytest.fixture
@@ -38,6 +38,7 @@ async def temp_prompt_store(tmp_path_factory):
             metadata=KVStoreReference(backend="kv_test", namespace="registry"),
             inference=InferenceStoreReference(backend="sql_test", table_name="inference"),
             conversations=SqlStoreReference(backend="sql_test", table_name="conversations"),
+            prompts=KVStoreReference(backend="kv_test", namespace="prompts"),
         ),
     )
     mock_run_config = StackRunConfig(
@@ -50,6 +51,6 @@ async def temp_prompt_store(tmp_path_factory):
     store = PromptServiceImpl(config, deps={})
 
     register_kvstore_backends({"kv_test": storage.backends["kv_test"]})
-    store.kvstore = await kvstore_impl(KVStoreReference(backend="kv_test", namespace="prompts"))
+    await store.initialize()
 
     yield store