Merge pull request #393 from partoneplay/main

Add Milvus as vector storage

examples/lightrag_ollama_neo4j_milvus_demo.py

@@ -0,0 +1,51 @@
+import os
+from lightrag import LightRAG, QueryParam
+from lightrag.llm import ollama_model_complete, ollama_embed
+from lightrag.utils import EmbeddingFunc
+
+# WorkingDir
+ROOT_DIR = os.path.dirname(os.path.abspath(__file__))
+WORKING_DIR = os.path.join(ROOT_DIR, "myKG")
+if not os.path.exists(WORKING_DIR):
+    os.mkdir(WORKING_DIR)
+print(f"WorkingDir: {WORKING_DIR}")
+
+# neo4j
+BATCH_SIZE_NODES = 500
+BATCH_SIZE_EDGES = 100
+os.environ["NEO4J_URI"] = "bolt://localhost:7687"
+os.environ["NEO4J_USERNAME"] = "neo4j"
+os.environ["NEO4J_PASSWORD"] = "neo4j"
+
+# milvus
+os.environ["MILVUS_URI"] = "http://localhost:19530"
+os.environ["MILVUS_USER"] = "root"
+os.environ["MILVUS_PASSWORD"] = "root"
+os.environ["MILVUS_DB_NAME"] = "lightrag"
+
+
+rag = LightRAG(
+    working_dir=WORKING_DIR,
+    llm_model_func=ollama_model_complete,
+    llm_model_name="qwen2.5:14b",
+    llm_model_max_async=4,
+    llm_model_max_token_size=32768,
+    llm_model_kwargs={"host": "http://127.0.0.1:11434", "options": {"num_ctx": 32768}},
+    embedding_func=EmbeddingFunc(
+        embedding_dim=1024,
+        max_token_size=8192,
+        func=lambda texts: ollama_embed(
+            texts=texts, embed_model="bge-m3:latest", host="http://127.0.0.1:11434"
+        ),
+    ),
+    graph_storage="Neo4JStorage",
+    vector_storage="MilvusVectorDBStorge",
+)
+
+file = "./book.txt"
+with open(file, "r") as f:
+    rag.insert(f.read())
+
+print(
+    rag.query("What are the top themes in this story?", param=QueryParam(mode="hybrid"))
+)

lightrag/kg/milvus_impl.py

@@ -0,0 +1,88 @@
+import asyncio
+import os
+from tqdm.asyncio import tqdm as tqdm_async
+from dataclasses import dataclass
+import numpy as np
+from lightrag.utils import logger
+from ..base import BaseVectorStorage
+
+from pymilvus import MilvusClient
+
+
+@dataclass
+class MilvusVectorDBStorge(BaseVectorStorage):
+    @staticmethod
+    def create_collection_if_not_exist(
+        client: MilvusClient, collection_name: str, **kwargs
+    ):
+        if client.has_collection(collection_name):
+            return
+        client.create_collection(
+            collection_name, max_length=64, id_type="string", **kwargs
+        )
+
+    def __post_init__(self):
+        self._client = MilvusClient(
+            uri=os.environ.get(
+                "MILVUS_URI",
+                os.path.join(self.global_config["working_dir"], "milvus_lite.db"),
+            ),
+            user=os.environ.get("MILVUS_USER", ""),
+            password=os.environ.get("MILVUS_PASSWORD", ""),
+            token=os.environ.get("MILVUS_TOKEN", ""),
+            db_name=os.environ.get("MILVUS_DB_NAME", ""),
+        )
+        self._max_batch_size = self.global_config["embedding_batch_num"]
+        MilvusVectorDBStorge.create_collection_if_not_exist(
+            self._client,
+            self.namespace,
+            dimension=self.embedding_func.embedding_dim,
+        )
+
+    async def upsert(self, data: dict[str, dict]):
+        logger.info(f"Inserting {len(data)} vectors to {self.namespace}")
+        if not len(data):
+            logger.warning("You insert an empty data to vector DB")
+            return []
+        list_data = [
+            {
+                "id": k,
+                **{k1: v1 for k1, v1 in v.items() if k1 in self.meta_fields},
+            }
+            for k, v in data.items()
+        ]
+        contents = [v["content"] for v in data.values()]
+        batches = [
+            contents[i : i + self._max_batch_size]
+            for i in range(0, len(contents), self._max_batch_size)
+        ]
+        embedding_tasks = [self.embedding_func(batch) for batch in batches]
+        embeddings_list = []
+        for f in tqdm_async(
+            asyncio.as_completed(embedding_tasks),
+            total=len(embedding_tasks),
+            desc="Generating embeddings",
+            unit="batch",
+        ):
+            embeddings = await f
+            embeddings_list.append(embeddings)
+        embeddings = np.concatenate(embeddings_list)
+        for i, d in enumerate(list_data):
+            d["vector"] = embeddings[i]
+        results = self._client.upsert(collection_name=self.namespace, data=list_data)
+        return results
+
+    async def query(self, query, top_k=5):
+        embedding = await self.embedding_func([query])
+        results = self._client.search(
+            collection_name=self.namespace,
+            data=embedding,
+            limit=top_k,
+            output_fields=list(self.meta_fields),
+            search_params={"metric_type": "COSINE", "params": {"radius": 0.2}},
+        )
+        print(results)
+        return [
+            {**dp["entity"], "id": dp["id"], "distance": dp["distance"]}
+            for dp in results[0]
+        ]

lightrag/lightrag.py

@@ -44,6 +44,8 @@ from .kg.neo4j_impl import Neo4JStorage
 
 from .kg.oracle_impl import OracleKVStorage, OracleGraphStorage, OracleVectorDBStorage
 
+from .kg.milvus_impl import MilvusVectorDBStorge
+
 # future KG integrations
 
 # from .kg.ArangoDB_impl import (
@@ -228,6 +230,7 @@ class LightRAG:
             # vector storage
             "NanoVectorDBStorage": NanoVectorDBStorage,
             "OracleVectorDBStorage": OracleVectorDBStorage,
+            "MilvusVectorDBStorge": MilvusVectorDBStorge,
             # graph storage
             "NetworkXStorage": NetworkXStorage,
             "Neo4JStorage": Neo4JStorage,

requirements.txt

@@ -11,6 +11,7 @@ networkx
 ollama
 openai
 oracledb
+pymilvus
 pyvis
 tenacity
 # lmdeploy[all]