Add Milvus integration for vector create and search

docs/_toc.yml

@@ -81,6 +81,7 @@ parts:
           - file: source/reference/vector_databases/qdrant
           - file: source/reference/vector_databases/pgvector
           - file: source/reference/vector_databases/pinecone
+          - file: source/reference/vector_databases/milvus
 
       - file: source/reference/ai/index
         title: AI Engines

docs/source/reference/vector_databases/milvus.rst

@@ -0,0 +1,37 @@
+Milvus
+==========
+
+Milvus is an open-source, distributed vector database designed for similarity search and analytics on large-scale vector data.
+The connection to Milvus is based on the `pymilvus <https://pymilvus.readthedocs.io/en/latest>`_ library.
+
+Dependency
+----------
+
+* pymilvus
+
+Parameters
+----------
+
+To use Milvus you must have a URI to a running Milvus instance. Here are the `instructions to spin up a local instance <https://milvus.io/docs/install_standalone-docker.md>`_.
+If you are running it locally, the Milvus instance should be running on ``http://localhost:19530``. Please be sure that the Milvus version is >= 2.3.0. Below are values that the Milvus integration uses:
+
+* `MILVUS_URI` is the URI of the Milvus instance (which would be ``http://localhost:19530`` when running locally). **This value is required**
+* `MILVUS_USER` is the name of the user for the Milvus instance.
+* `MILVUS_PASSWORD` is the password of the user for the Milvus instance.
+* `MILVUS_DB_NAME` is the name of the database to be used. This will default to the `default` database if not provided.
+* `MILVUS_TOKEN` is the authorization token for the Milvus instance. 
+
+The above values can either be set via the ``SET`` statement, or in the os environment fields "MILVUS_URI", "MILVUS_USER", "MILVUS_PASSWORD", "MILVUS_DB_NAME", and "MILVUS_TOKEN"
+
+
+.. code-block:: sql
+
+   SET MILVUS_URI = 'http://localhost:19530';
+
+
+Create Index
+-----------------
+
+.. code-block:: sql
+
+   CREATE INDEX index_name ON table_name (data) USING MILVUS;

evadb/catalog/catalog_type.py

@@ -117,6 +117,7 @@ class VectorStoreType(EvaDBEnum):
     PINECONE  # noqa: F821
     PGVECTOR  # noqa: F821
     CHROMADB  # noqa: F821
+    MILVUS  # noqa: F821
 
 
 class VideoColumnName(EvaDBEnum):

evadb/evadb_config.py

@@ -36,4 +36,9 @@
     "OPENAI_API_KEY": "",
     "PINECONE_API_KEY": "",
     "PINECONE_ENV": "",
+    "MILVUS_URI": "",
+    "MILVUS_USER": "",
+    "MILVUS_PASSWORD": "",
+    "MILVUS_DB_NAME": "",
+    "MILVUS_TOKEN": "",
 }

evadb/executor/executor_utils.py

@@ -185,6 +185,18 @@ def handle_vector_store_params(
             ),
             "PINECONE_ENV": catalog().get_configuration_catalog_value("PINECONE_ENV"),
         }
+    elif vector_store_type == VectorStoreType.MILVUS:
+        return {
+            "MILVUS_URI": catalog().get_configuration_catalog_value("MILVUS_URI"),
+            "MILVUS_USER": catalog().get_configuration_catalog_value("MILVUS_USER"),
+            "MILVUS_PASSWORD": catalog().get_configuration_catalog_value(
+                "MILVUS_PASSWORD"
+            ),
+            "MILVUS_DB_NAME": catalog().get_configuration_catalog_value(
+                "MILVUS_DB_NAME"
+            ),
+            "MILVUS_TOKEN": catalog().get_configuration_catalog_value("MILVUS_TOKEN"),
+        }
     else:
         raise ValueError("Unsupported vector store type: {}".format(vector_store_type))
 

evadb/interfaces/relational/db.py

@@ -248,7 +248,7 @@ def create_vector_index(
             index_name (str): Name of the index.
             table_name (str): Name of the table.
             expr (str): Expression used to build the vector index.
-            using (str): Method used for indexing, can be `FAISS` or `QDRANT` or `PINECONE` or `CHROMADB`.
+            using (str): Method used for indexing, can be `FAISS` or `QDRANT` or `PINECONE` or `CHROMADB` or `MILVUS`.
 
         Returns:
             EvaDBCursor: The EvaDBCursor object.

evadb/parser/evadb.lark

@@ -53,7 +53,7 @@ function_metadata_key: uid
 
 function_metadata_value: constant
 
-vector_store_type: USING (FAISS | QDRANT | PINECONE | PGVECTOR | CHROMADB)
+vector_store_type: USING (FAISS | QDRANT | PINECONE | PGVECTOR | CHROMADB | MILVUS)
 
 index_elem: ("(" uid_list ")"
           | "(" function_call ")")
@@ -424,6 +424,7 @@ QDRANT:                              "QDRANT"i
 PINECONE:                            "PINECONE"i
 PGVECTOR:                            "PGVECTOR"i
 CHROMADB:                            "CHROMADB"i
+MILVUS:                              "MILVUS"i
 
 // Computer vision tasks
 

evadb/parser/lark_visitor/_create_statements.py

@@ -299,6 +299,8 @@ def vector_store_type(self, tree):
             vector_store_type = VectorStoreType.PGVECTOR
         elif str.upper(token) == "CHROMADB":
             vector_store_type = VectorStoreType.CHROMADB
+        elif str.upper(token) == "MILVUS":
+            vector_store_type = VectorStoreType.MILVUS
         return vector_store_type
 
 

evadb/third_party/vector_stores/milvus.py

@@ -0,0 +1,146 @@
+# coding=utf-8
+# Copyright 2018-2023 EvaDB
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import os
+from typing import List
+
+from evadb.third_party.vector_stores.types import (
+    FeaturePayload,
+    VectorIndexQuery,
+    VectorIndexQueryResult,
+    VectorStore,
+)
+from evadb.utils.generic_utils import try_to_import_milvus_client
+
+allowed_params = [
+    "MILVUS_URI",
+    "MILVUS_USER",
+    "MILVUS_PASSWORD",
+    "MILVUS_DB_NAME",
+    "MILVUS_TOKEN",
+]
+required_params = []
+_milvus_client_instance = None
+
+
+def get_milvus_client(
+    milvus_uri: str,
+    milvus_user: str,
+    milvus_password: str,
+    milvus_db_name: str,
+    milvus_token: str,
+):
+    global _milvus_client_instance
+    if _milvus_client_instance is None:
+        try_to_import_milvus_client()
+        import pymilvus
+
+        _milvus_client_instance = pymilvus.MilvusClient(
+            uri=milvus_uri,
+            user=milvus_user,
+            password=milvus_password,
+            db_name=milvus_db_name,
+            token=milvus_token,
+        )
+
+    return _milvus_client_instance
+
+
+class MilvusVectorStore(VectorStore):
+    def __init__(self, index_name: str, **kwargs) -> None:
+        # Milvus URI is the only required
+        self._milvus_uri = kwargs.get("MILVUS_URI")
+
+        if not self._milvus_uri:
+            self._milvus_uri = os.environ.get("MILVUS_URI")
+
+        assert (
+            self._milvus_uri
+        ), "Please set your Milvus URI in evadb.yml file (third_party, MILVUS_URI) or environment variable (MILVUS_URI)."
+
+        # Check other Milvus variables for additional customization
+        self._milvus_user = kwargs.get("MILVUS_USER")
+
+        if not self._milvus_user:
+            self._milvus_user = os.environ.get("MILVUS_USER", "")
+
+        self._milvus_password = kwargs.get("MILVUS_PASSWORD")
+
+        if not self._milvus_password:
+            self._milvus_password = os.environ.get("MILVUS_PASSWORD", "")
+
+        self._milvus_db_name = kwargs.get("MILVUS_DB_NAME")
+
+        if not self._milvus_db_name:
+            self._milvus_db_name = os.environ.get("MILVUS_DB_NAME", "")
+
+        self._milvus_token = kwargs.get("MILVUS_TOKEN")
+
+        if not self._milvus_token:
+            self._milvus_token = os.environ.get("MILVUS_TOKEN", "")
+
+        self._client = get_milvus_client(
+            milvus_uri=self._milvus_uri,
+            milvus_user=self._milvus_user,
+            milvus_password=self._milvus_password,
+            milvus_db_name=self._milvus_db_name,
+            milvus_token=self._milvus_token,
+        )
+        self._collection_name = index_name
+
+    def create(self, vector_dim: int):
+        if self._collection_name in self._client.list_collections():
+            self._client.drop_collection(self._collection_name)
+        self._client.create_collection(
+            collection_name=self._collection_name,
+            dimension=vector_dim,
+            metric_type="COSINE",
+        )
+
+    def add(self, payload: List[FeaturePayload]):
+        milvus_data = [
+            {
+                "id": feature_payload.id,
+                "vector": feature_payload.embedding.reshape(-1).tolist(),
+            }
+            for feature_payload in payload
+        ]
+        ids = [feature_payload.id for feature_payload in payload]
+
+        # Milvus Client does not have upsert operation, perform delete + insert to emulate it
+        self._client.delete(collection_name=self._collection_name, pks=ids)
+
+        self._client.insert(collection_name=self._collection_name, data=milvus_data)
+
+    def persist(self):
+        self._client.flush(self._collection_name)
+
+    def delete(self) -> None:
+        self._client.drop_collection(
+            collection_name=self._collection_name,
+        )
+
+    def query(self, query: VectorIndexQuery) -> VectorIndexQueryResult:
+        response = self._client.search(
+            collection_name=self._collection_name,
+            data=[query.embedding.reshape(-1).tolist()],
+            limit=query.top_k,
+        )[0]
+
+        distances, ids = [], []
+        for result in response:
+            distances.append(result["distance"])
+            ids.append(result["id"])
+
+        return VectorIndexQueryResult(distances, ids)

evadb/third_party/vector_stores/utils.py

@@ -15,6 +15,7 @@
 from evadb.catalog.catalog_type import VectorStoreType
 from evadb.third_party.vector_stores.chromadb import ChromaDBVectorStore
 from evadb.third_party.vector_stores.faiss import FaissVectorStore
+from evadb.third_party.vector_stores.milvus import MilvusVectorStore
 from evadb.third_party.vector_stores.pinecone import PineconeVectorStore
 from evadb.third_party.vector_stores.qdrant import QdrantVectorStore
 from evadb.utils.generic_utils import validate_kwargs
@@ -50,5 +51,13 @@ def init_vector_store(
             validate_kwargs(kwargs, required_params, required_params)
             return ChromaDBVectorStore(index_name, **kwargs)
 
+        elif vector_store_type == VectorStoreType.MILVUS:
+            from evadb.third_party.vector_stores.milvus import (
+                allowed_params,
+                required_params,
+            )
+
+            validate_kwargs(kwargs, allowed_params, required_params)
+            return MilvusVectorStore(index_name, **kwargs)
         else:
             raise Exception(f"Vector store {vector_store_type} not supported")

evadb/utils/generic_utils.py

@@ -581,6 +581,16 @@ def try_to_import_chromadb_client():
         )
 
 
+def try_to_import_milvus_client():
+    try:
+        import pymilvus  # noqa: F401
+    except ImportError:
+        raise ValueError(
+            """Could not import pymilvus python package.
+                Please install it with 'pip install pymilvus`."""
+        )
+
+
 def is_qdrant_available() -> bool:
     try:
         try_to_import_qdrant_client()
@@ -605,6 +615,14 @@ def is_chromadb_available() -> bool:
         return False
 
 
+def is_milvus_available() -> bool:
+    try:
+        try_to_import_milvus_client()
+        return True
+    except ValueError:
+        return False
+
+
 ##############################
 ## UTILS
 ##############################

script/formatting/spelling.txt

@@ -551,6 +551,11 @@ MaxPool
 MemeImages
 MemoTest
 MetaData
+MILVUS
+Milvus
+Milvus's
+MilvusClient
+MilvusVectorStore
 MindsDB
 MiniLM
 MnistImageClassifier
@@ -1402,6 +1407,7 @@ memeimages
 metaclass
 metafile
 metainfo
+milvus
 mindsdb
 miniconda
 mins
@@ -1526,6 +1532,7 @@ ptype
 pushdown
 px
 py
+pymilvus
 pymupdfs
 pypi
 pypirc
@@ -1655,6 +1662,7 @@ testCreateIndexName
 testDeleteOne
 testFaissIndexImageDataset
 testFaissIndexScanRewrite
+testMilvusIndexImageDataset
 testIndex
 testIndexAutoUpdate
 testOpenTable

setup.py

@@ -112,6 +112,8 @@ def read(path, encoding="utf-8"):
 
 chromadb_libs = ["chromadb"]
 
+milvus_libs = ["pymilvus>=2.3.0"]
+
 postgres_libs = [
     "psycopg2",
 ]
@@ -123,8 +125,8 @@ def read(path, encoding="utf-8"):
 xgboost_libs = ["flaml[automl]"]
 
 forecasting_libs = [
-    "statsforecast", # MODEL TRAIN AND FINE TUNING
-    "neuralforecast" # MODEL TRAIN AND FINE TUNING
+    "statsforecast",  # MODEL TRAIN AND FINE TUNING
+    "neuralforecast",  # MODEL TRAIN AND FINE TUNING
 ]
 
 imagegen_libs = [
@@ -168,6 +170,7 @@ def read(path, encoding="utf-8"):
     "qdrant": qdrant_libs,
     "pinecone": pinecone_libs,
     "chromadb": chromadb_libs,
+    "milvus": milvus_libs,
     "postgres": postgres_libs,
     "ludwig": ludwig_libs,
     "sklearn": sklearn_libs,