feat: Qdrant inline provider (#1273)

# What does this PR do?
Removed local execution option from the remote Qdrant provider and
introduced an explicit inline provider for the embedded execution.
Updated the ollama template to include this option: this part can be
reverted in case we don't want to have two default `vector_io`
providers.

(Closes #1082)

## Test Plan
Build and run an ollama distro:
```bash
llama stack build --template ollama --image-type conda
llama stack run --image-type conda ollama
```

Run one of the sample ingestionapplicatinos like
[rag_with_vector_db.py](https://github.com/meta-llama/llama-stack-apps/blob/main/examples/agents/rag_with_vector_db.py),
but replace this line:
```py
    selected_vector_provider = vector_providers[0]
```
with the following, to use the `qdrant` provider:
```py
    selected_vector_provider = vector_providers[1]
```

After running the test code, verify the timestamp of the Qdrant store:
```bash
% ls -ltr ~/.llama/distributions/ollama/qdrant.db/collection/test_vector_db_*
total 784
-rw-r--r--@ 1 dmartino  staff  401408 Feb 26 10:07 storage.sqlite
```

[//]: # (## Documentation)

---------

Signed-off-by: Daniele Martinoli <dmartino@redhat.com>
Co-authored-by: Francisco Arceo <farceo@redhat.com>

llama_stack/providers/inline/vector_io/qdrant/__init__.py

@@ -0,0 +1,19 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+from typing import Dict
+
+from llama_stack.providers.datatypes import Api, ProviderSpec
+
+from .config import QdrantVectorIOConfig
+
+
+async def get_adapter_impl(config: QdrantVectorIOConfig, deps: Dict[Api, ProviderSpec]):
+    from llama_stack.providers.remote.vector_io.qdrant.qdrant import QdrantVectorIOAdapter
+
+    impl = QdrantVectorIOAdapter(config, deps[Api.inference])
+    await impl.initialize()
+    return impl

llama_stack/providers/inline/vector_io/qdrant/config.py

@@ -0,0 +1,23 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the terms described in the LICENSE file in
+# the root directory of this source tree.
+
+
+from typing import Any, Dict
+
+from pydantic import BaseModel
+
+from llama_stack.schema_utils import json_schema_type
+
+
+@json_schema_type
+class QdrantVectorIOConfig(BaseModel):
+    path: str
+
+    @classmethod
+    def sample_run_config(cls, __distro_dir__: str) -> Dict[str, Any]:
+        return {
+            "path": "${env.QDRANT_PATH:~/.llama/" + __distro_dir__ + "}/" + "qdrant.db",
+        }

llama_stack/providers/registry/vector_io.py

@@ -92,6 +92,14 @@ def available_providers() -> List[ProviderSpec]:
             ),
             api_dependencies=[Api.inference],
         ),
+        InlineProviderSpec(
+            api=Api.vector_io,
+            provider_type="inline::qdrant",
+            pip_packages=["qdrant-client"],
+            module="llama_stack.providers.inline.vector_io.qdrant",
+            config_class="llama_stack.providers.inline.vector_io.qdrant.QdrantVectorIOConfig",
+            api_dependencies=[Api.inference],
+        ),
         remote_provider_spec(
             Api.vector_io,
             AdapterSpec(

llama_stack/providers/remote/vector_io/qdrant/config.py

@@ -23,7 +23,6 @@ class QdrantVectorIOConfig(BaseModel):
     prefix: Optional[str] = None
     timeout: Optional[int] = None
     host: Optional[str] = None
-    path: Optional[str] = None
 
     @classmethod
     def sample_run_config(cls, **kwargs: Any) -> Dict[str, Any]:

llama_stack/providers/remote/vector_io/qdrant/qdrant.py

@@ -6,7 +6,7 @@
 
 import logging
 import uuid
-from typing import Any, Dict, List, Optional
+from typing import Any, Dict, List, Optional, Union
 
 from numpy.typing import NDArray
 from qdrant_client import AsyncQdrantClient, models
@@ -16,12 +16,13 @@ from llama_stack.apis.inference import InterleavedContent
 from llama_stack.apis.vector_dbs import VectorDB
 from llama_stack.apis.vector_io import Chunk, QueryChunksResponse, VectorIO
 from llama_stack.providers.datatypes import Api, VectorDBsProtocolPrivate
+from llama_stack.providers.inline.vector_io.qdrant import QdrantVectorIOConfig as InlineQdrantVectorIOConfig
 from llama_stack.providers.utils.memory.vector_store import (
     EmbeddingIndex,
     VectorDBWithIndex,
 )
 
-from .config import QdrantVectorIOConfig
+from .config import QdrantVectorIOConfig as RemoteQdrantVectorIOConfig
 
 log = logging.getLogger(__name__)
 CHUNK_ID_KEY = "_chunk_id"
@@ -99,17 +100,19 @@ class QdrantIndex(EmbeddingIndex):
 
 
 class QdrantVectorIOAdapter(VectorIO, VectorDBsProtocolPrivate):
-    def __init__(self, config: QdrantVectorIOConfig, inference_api: Api.inference) -> None:
+    def __init__(
+        self, config: Union[RemoteQdrantVectorIOConfig, InlineQdrantVectorIOConfig], inference_api: Api.inference
+    ) -> None:
         self.config = config
-        self.client = AsyncQdrantClient(**self.config.model_dump(exclude_none=True))
+        self.client: AsyncQdrantClient = None
         self.cache = {}
         self.inference_api = inference_api
 
     async def initialize(self) -> None:
-        pass
+        self.client = AsyncQdrantClient(**self.config.model_dump(exclude_none=True))
 
     async def shutdown(self) -> None:
-        self.client.close()
+        await self.client.close()
 
     async def register_vector_db(
         self,
@@ -123,6 +126,11 @@ class QdrantVectorIOAdapter(VectorIO, VectorDBsProtocolPrivate):
 
         self.cache[vector_db.identifier] = index
 
+    async def unregister_vector_db(self, vector_db_id: str) -> None:
+        if vector_db_id in self.cache:
+            await self.cache[vector_db_id].index.delete()
+            del self.cache[vector_db_id]
+
     async def _get_and_cache_vector_db_index(self, vector_db_id: str) -> Optional[VectorDBWithIndex]:
         if vector_db_id in self.cache:
             return self.cache[vector_db_id]