Merge branch 'main' into fix-fireworks

llama_stack/cli/stack/_build.py

@@ -45,6 +45,7 @@ from llama_stack.core.utils.dynamic import instantiate_class_type
 from llama_stack.core.utils.exec import formulate_run_args, run_command
 from llama_stack.core.utils.image_types import LlamaStackImageType
 from llama_stack.providers.datatypes import Api
+from llama_stack.providers.utils.sqlstore.sqlstore import SqliteSqlStoreConfig
 
 DISTRIBS_PATH = Path(__file__).parent.parent.parent / "distributions"
 
@@ -294,6 +295,12 @@ def _generate_run_config(
         if build_config.external_providers_dir
         else EXTERNAL_PROVIDERS_DIR,
     )
+    if not run_config.inference_store:
+        run_config.inference_store = SqliteSqlStoreConfig(
+            **SqliteSqlStoreConfig.sample_run_config(
+                __distro_dir__=(DISTRIBS_BASE_DIR / image_name).as_posix(), db_name="inference_store.db"
+            )
+        )
     # build providers dict
     provider_registry = get_provider_registry(build_config)
     for api in apis:

llama_stack/core/library_client.py

@@ -10,7 +10,6 @@ import json
 import logging  # allow-direct-logging
 import os
 import sys
-from concurrent.futures import ThreadPoolExecutor
 from enum import Enum
 from io import BytesIO
 from pathlib import Path
@@ -148,7 +147,6 @@ class LlamaStackAsLibraryClient(LlamaStackClient):
         self.async_client = AsyncLlamaStackAsLibraryClient(
             config_path_or_distro_name, custom_provider_registry, provider_data, skip_logger_removal
         )
-        self.pool_executor = ThreadPoolExecutor(max_workers=4)
         self.provider_data = provider_data
 
         self.loop = asyncio.new_event_loop()

llama_stack/providers/inline/tool_runtime/rag/memory.py

@@ -167,8 +167,18 @@ class MemoryToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, RAGToolRunti
             for vector_db_id in vector_db_ids
         ]
         results: list[QueryChunksResponse] = await asyncio.gather(*tasks)
-        chunks = [c for r in results for c in r.chunks]
-        scores = [s for r in results for s in r.scores]
+
+        chunks = []
+        scores = []
+
+        for vector_db_id, result in zip(vector_db_ids, results, strict=False):
+            for chunk, score in zip(result.chunks, result.scores, strict=False):
+                if not hasattr(chunk, "metadata") or chunk.metadata is None:
+                    chunk.metadata = {}
+                chunk.metadata["vector_db_id"] = vector_db_id
+
+                chunks.append(chunk)
+                scores.append(score)
 
         if not chunks:
             return RAGQueryResult(content=None)
@@ -203,6 +213,7 @@ class MemoryToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, RAGToolRunti
             metadata_keys_to_exclude_from_context = [
                 "token_count",
                 "metadata_token_count",
+                "vector_db_id",
             ]
             metadata_for_context = {}
             for k in chunk_metadata_keys_to_include_from_context:
@@ -227,6 +238,7 @@ class MemoryToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, RAGToolRunti
                 "document_ids": [c.metadata["document_id"] for c in chunks[: len(picked)]],
                 "chunks": [c.content for c in chunks[: len(picked)]],
                 "scores": scores[: len(picked)],
+                "vector_db_ids": [c.metadata["vector_db_id"] for c in chunks[: len(picked)]],
             },
         )