--- title: "Database Vector Stores" description: "Database Vector Stores provide persistent indexing, metadata filtering, and approximate nearest neighbor (ANN) retrieval capabilities across relational databases, distributed search engines, in-mem..." last_updated: "2026-10-11T17:37:10.218815+00:00" canonical_url: "https://www.doc0.dev/docs/faa36707-7c28-4f69-a18f-700ff61c704e/technical/vector-stores/database-vector-stores" ---
Relevant source files The following files were used as context for generating this wiki page: - [mem0/vector_stores/oracledb.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/oracledb.py) - [mem0-ts/src/oss/src/vector_stores/oracledb.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/oracledb.ts) - [mem0/vector_stores/valkey.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/valkey.py) - [mem0-ts/src/oss/src/vector_stores/baidu.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/baidu.ts) - [mem0-ts/src/oss/src/vector_stores/neptune_analytics.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/neptune_analytics.ts) - [mem0/vector_stores/redis.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/redis.py) - [mem0/vector_stores/neptune_analytics.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/neptune_analytics.py) - [mem0-ts/src/oss/src/vector_stores/cassandra.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/cassandra.ts) - [mem0-ts/src/oss/src/vector_stores/redis.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/redis.ts) - [mem0/vector_stores/opensearch.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/opensearch.py) - [mem0-ts/src/oss/src/vector_stores/valkey.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/valkey.ts) - [mem0/vector_stores/baidu.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/baidu.py) - [mem0-ts/src/oss/src/vector_stores/opensearch.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/opensearch.ts) - [mem0/vector_stores/mongodb.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/mongodb.py) - [mem0-ts/src/oss/src/vector_stores/elasticsearch.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/elasticsearch.ts) - [mem0-ts/src/oss/src/vector_stores/azure_mysql.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/azure_mysql.ts) - [mem0/configs/vector_stores/oracledb.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/configs/vector_stores/oracledb.py) - [mem0/configs/vector_stores/__init__.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/configs/vector_stores/__init__.py) - [mem0/vector_stores/chroma.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/chroma.py) - [mem0/vector_stores/__init__.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/__init__.py) - [mem0-ts/src/oss/src/vector_stores/mongodb.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/mongodb.ts) - [mem0/vector_stores/azure_ai_search.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/azure_ai_search.py) - [mem0-ts/src/oss/src/vector_stores/weaviate.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/weaviate.ts) - [mem0/vector_stores/faiss.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/faiss.py) - [mem0/vector_stores/weaviate.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/weaviate.py) - [mem0/vector_stores/cassandra.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/cassandra.py) - [mem0/vector_stores/milvus.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/milvus.py) - [mem0-ts/src/oss/src/vector_stores/vectorize.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/vectorize.ts) - [mem0-ts/src/oss/src/vector_stores/pgvector.ts](https://github.com/blade47/mem0-ts/src/oss/src/vector_stores/pgvector.ts) - [mem0/vector_stores/elasticsearch.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/elasticsearch.py)
## Overview Database Vector Stores provide persistent indexing, metadata filtering, and approximate nearest neighbor (ANN) retrieval capabilities across relational databases, distributed search engines, in-memory data structures, and graph analytics systems. Within Mem0, these database backends handle the storage and querying of long-term memory embeddings alongside associated payloads, user tenant contexts, and timestamps. Unlike standalone [Embedded Vector Stores](https://www.doc0.dev/docs/faa36707-7c28-4f69-a18f-700ff61c704e/technical/vector-stores/embedded-vector-stores) or managed [Cloud Vector Backends](https://www.doc0.dev/docs/faa36707-7c28-4f69-a18f-700ff61c704e/technical/vector-stores/cloud-vector-backends), database vector stores integrate embedding storage directly into existing enterprise data systems like PostgreSQL via `pgvector`, Elasticsearch, OpenSearch, Redis, Valkey, Apache Cassandra, Amazon Neptune Analytics, Oracle AI Vector Search, MongoDB, Azure AI Search, and Baidu Mochow. Sources: [mem0-ts/src/oss/src/vector_stores/pgvector.ts:217-245](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/pgvector.ts#L217-L245) The primary design challenge solved by these drivers is bridging multi-dimensional vector math with structured relational tables, document schemas, or inverted indices. Because different engines handle vector indexing via distinct internal mechanisms—such as PostgreSQL native `vector` data types, Elasticsearch `dense_vector` fields with HNSW graphs, or Redis Hash keys with RediSearch modules—each backend abstracts the connection lifecycle, schema initialization, parameter binding, and query filtering into a unified `VectorStore` interface. Sources: [mem0-ts/src/oss/src/vector_stores/pgvector.ts:217-245](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/pgvector.ts#L217-L245) ```mermaid flowchart TD A["Memory Core Engine"] --> B["VectorStore Interface"] B --> C["Relational DBs
(pgvector, Azure MySQL, Oracle)"] B --> D["Search Engines
(Elasticsearch, OpenSearch, Azure AI Search)"] B --> E["In-Memory & KV
(Redis, Valkey)"] B --> F["Distributed & Graph
(Cassandra, Neptune Analytics, Milvus, Baidu)"] ``` Sources: [mem0-ts/src/oss/src/vector_stores/pgvector.ts:217-245](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/pgvector.ts#L217-L245) ## Relational Database Integrations: pgvector, Azure MySQL, and Oracle DB Relational database integrations extend traditional SQL tables with native or JSON-serialized vector storage and similarity operators. In PGVector, tables are initialized with a `UUID` primary key, a `vector` data type column sized to `embeddingModelDims`, and a `JSONB` payload column. During nearest neighbor search, `pgvector` executes queries using the `<->` cosine distance operator, sorting results and bounding output sizes with `LIMIT $2`. Parameterized filters are dynamically constructed via `buildFilterConditions()`, translating metadata JSON path lookups into SQL expressions with strict parameter indexing to prevent injection. Sources: [mem0-ts/src/oss/src/vector_stores/pgvector.ts:284-446](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/pgvector.ts#L284-L446) For Azure MySQL, embeddings and JSON payloads are stored as JSON columns and evaluated via application-level `cosineSimilarity()` loops over fetched rows, since MySQL handles vector distance sorting in memory after fetching candidate sets. Sources: [mem0-ts/src/oss/src/vector_stores/azure_mysql.ts:123-211](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/azure_mysql.ts#L123-L211) Oracle AI Vector Search utilizes Oracle 23.4+ native `VECTOR` data types and `JSON` payload columns inside relational tables. Search operations invoke `VECTOR_DISTANCE(vector, :query_vec, ...)` combined with `ORDER BY distance FETCH APPROX FIRST :max_rows ROWS ONLY`. When executing filtered searches, Oracle conditionally omits index hints or injects `VECTOR_INDEX_TRANSFORM` depending on filter presence. Sources: [mem0/vector_stores/oracledb.py:350-402](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/oracledb.py#L350-L402) ```mermaid sequenceDiagram participant App as Memory Client participant PG as PGVector Store participant DB as PostgreSQL Engine App->>PG: search(queryVector, topK, filters) PG->>DB: SELECT id, vector <=> $1::vector, payload WHERE ... ORDER BY distance LIMIT $2 DB-->>PG: Result rows with distance and JSONB payload PG-->>App: Normalized VectorStoreResult[] array ``` Sources: [mem0-ts/src/oss/src/vector_stores/pgvector.ts:414-446](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/pgvector.ts#L414-L446) | Provider | Vector Storage Format | Index / Distance Mechanism | Metadata / Filter Strategy | | :--- | :--- | :--- | :--- | | **pgvector** | `vector(dims)` column | HNSW / DiskANN via `<->` operator | JSONB operators (`payload->>'key'`) | | **Azure MySQL** | `JSON` column | Application-level `cosineSimilarity()` | `JSON_EXTRACT(payload, ?)` | | **Oracle DB** | `VECTOR(dims)` column | HNSW / IVF neighbor graph via `VECTOR_DISTANCE` | Oracle JSON relational duality / bind vars | Sources: [mem0-ts/src/oss/src/vector_stores/pgvector.ts:315-349](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/pgvector.ts#L315-L349), [mem0-ts/src/oss/src/vector_stores/azure_mysql.ts:124-189](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/azure_mysql.ts#L124-L189), [mem0/vector_stores/oracledb.py:321-364](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/oracledb.py#L321-L364) ## Search Engines: Elasticsearch, OpenSearch, and Azure AI Search Distributed search and retrieval engines leverage inverted indices alongside vector quantization and HNSW graphs to execute hybrid keyword and semantic searches. Sources: [mem0/vector_stores/opensearch.py:83-135](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/opensearch.py#L83-L135) Elasticsearch and OpenSearch store embeddings in `dense_vector` or `knn_vector` fields configured with `nmslib` or native engines using cosine similarity. When inserting documents, payloads are flattened into `metadata` objects. During vector search, queries construct a `knn` block specifying `query_vector`, `k`, and `num_candidates` (typically `topK * 2`). Metadata filters are validated against a strict regex (`/^[a-zA-Z_][a-zA-Z0-9_]*$/`) and injected into the KNN filter block as boolean term queries. Sources: [mem0-ts/src/oss/src/vector_stores/elasticsearch.ts:134-164](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/elasticsearch.ts#L134-L164), [mem0-ts/src/oss/src/vector_stores/opensearch.ts:160-194](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/opensearch.ts#L160-L194) Azure AI Search configures indices with `SearchField` vector definitions (supporting `Collection(Edm.Single)` or `Collection(Edm.Half)` precision) and optional Scalar or Binary Quantization compression profiles (`ScalarQuantizationCompression`, `BinaryQuantizationCompression`). Filtering is performed via OData filter expressions generated by `_build_filter_expression()`. Sources: [mem0/vector_stores/azure_ai_search.py:105-163](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/azure_ai_search.py#L105-L163) > [!NOTE] > OpenSearch and Elasticsearch require explicit index settings (`index.knn: true` in OpenSearch; `dense_vector` mapping with `index: true` in Elasticsearch) to enable approximate nearest neighbor indexing. Without these settings, searches fall back to exact brute-force scans across all documents. Sources: [mem0/vector_stores/elasticsearch.py:72-95](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/elasticsearch.py#L72-L95) | Search Backend | Vector Mapping Type | Default Distance / Metric | Filter Execution Mechanism | | :--- | :--- | :--- | :--- | | **Elasticsearch** | `dense_vector` | `cosine` | `knn.filter` boolean term clauses | | **OpenSearch** | `knn_vector` | `cosinesimil` | QueryDSL bool / term filters on `payload` | | **Azure AI Search** | `Collection(Edm.Single)` | Cosine via HNSW profile | OData filter expression (`eq`, `and`) | Sources: [mem0/vector_stores/elasticsearch.py:72-95](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/elasticsearch.py#L72-L95), [mem0/vector_stores/opensearch.py:83-107](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/opensearch.py#L83-L107), [mem0/vector_stores/azure_ai_search.py:135-160](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/azure_ai_search.py#L135-L160) ## In-Memory & Key-Value Stores: Redis and Valkey In-memory data stores like Redis and Valkey employ RediSearch and RedisVL modules to index hash structures containing binary-encoded floating-point vector embeddings. Sources: [mem0/vector_stores/redis.py:48-78](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/redis.py#L48-L78) The `RedisDB` and `ValkeyDB` drivers store entries as Redis Hashes using keys prefixed with `mem0:{collection_name}:{id}`. Float32 arrays are converted to raw byte buffers (`np.array(vector, dtype=np.float32).tobytes()` in Python or `Buffer.from(new Float32Array(query).buffer)` in TypeScript) before persistence. Sources: [mem0/vector_stores/redis.py:123-147](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/redis.py#L123-L147), [mem0-ts/src/oss/src/vector_stores/valkey.ts:432-439](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/valkey.ts#L432-L439) Schema definitions declare fields such as `memory_id`, `hash`, `agent_id`, `run_id`, `user_id` as `TAG` types, `memory` and `metadata` as `TEXT` or `TAG`, `created_at` as `NUMERIC`, and `embedding` as a `VECTOR` attribute using FLAT or HNSW algorithms. Sources: [mem0/vector_stores/valkey.py:18-34](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/valkey.py#L18-L34), [mem0/vector_stores/redis.py:20-36](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/redis.py#L20-L36) ```typescript // Example: Building a Valkey / Redis pre-filter expression export function buildRedisFilterExpr(filters?: SearchFilters): string { if (!filters) return "*"; const conditions = Object.entries(toSnakeCase(filters)) .filter(([, value]) => value !== null && value !== undefined) .map(([key, value]) => `@${key}:{${escapeRedisTagValue(value)}}`); return conditions.length > 0 ? conditions.join(" ") : "*"; } ``` Sources: [mem0-ts/src/oss/src/vector_stores/redis.ts:141-147](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/redis.ts#L141-L147) > [!CAUTION] > Tag filter values in RediSearch must be escaped via `escapeTagValue()` or `escapeRedisTagValue()` to neutralize special characters such as `*`, `|`, and whitespace. Unescaped tag values alter query syntax and can allow tenant-isolation filter bypasses. Sources: [mem0/vector_stores/valkey.py:48-57](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/valkey.py#L48-L57), [mem0-ts/src/oss/src/vector_stores/valkey.ts:22-27](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/valkey.ts#L22-L27) ## Distributed & Graph Backends: Cassandra, Neptune Analytics, Milvus, and Baidu Mochow Distributed and specialized graph stores handle massive memory throughput using custom partition keys and vector-optimized storage engines. Sources: [mem0-ts/src/oss/src/vector_stores/cassandra.ts:41-80](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/cassandra.ts#L41-L80) - **Apache Cassandra**: Stores vectors as `list` and payloads as JSON text inside tables structured under configurable keyspaces (default `mem0`). Search queries scan rows in pages (`PAGE_SIZE = 500`), normalize vectors, apply in-memory metadata filters, and maintain top-K results via `pushTopResult()`. - **Amazon Neptune Analytics**: Operates as a serverless graph database where memory entities are stored as graph vertices with label `MEM0_VECTOR_{collection_name}` and identifier `~id`. Vector upserts and similarity searches execute Cypher procedures such as `CALL neptune.algo.vectors.upsert(n, embedding)` and `CALL neptune.algo.vectors.topKByEmbeddingWithFiltering(...)`. - **Milvus / Zilliz**: Configures collections with `AUTOINDEX` for float vectors and `SPARSE_INVERTED_INDEX` with built-in `BM25` functions for hybrid sparse-dense retrieval over auto-tokenized `text` fields. - **Baidu Mochow**: Utilizes typed tables (`STRING` id, `FLOAT_VECTOR` vector, `JSON` metadata) with HNSW vector indices and `BM25SearchRequest` full-text capabilities. Sources: [mem0-ts/src/oss/src/vector_stores/cassandra.ts:143-185](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/cassandra.ts#L143-L185), [mem0/vector_stores/neptune_analytics.py:41-61](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/neptune_analytics.py#L41-L61), [mem0/vector_stores/milvus.py:33-65](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/milvus.py#L33-L65), [mem0-ts/src/oss/src/vector_stores/baidu.ts:352-391](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/baidu.ts#L352-L391) ```mermaid graph LR subgraph Cassandra C1["Scan Rows (Paged)"] --> C2["Normalize Vector & Parse Payload"] C2 --> C3["Filter & pushTopResult()"] end subgraph Neptune N1["Cypher Query Generation"] --> N2["neptune.algo.vectors.topKByEmbeddingWithFiltering"] N2 --> N3["Parse Node & Score Output"] end subgraph Milvus M1["Dense Vectors (AUTOINDEX)"] --> M2["Sparse Vectors (BM25 Function)"] M2 --> M3["Hybrid Search Result Ranking"] end ``` Sources: [mem0-ts/src/oss/src/vector_stores/cassandra.ts:143-185](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/cassandra.ts#L143-L185), [mem0/vector_stores/neptune_analytics.py:159-198](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/neptune_analytics.py#L159-L198), [mem0/vector_stores/milvus.py:66-125](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/milvus.py#L66-L125) ```typescript async search( query: number[], topK: number = 5, filters?: SearchFilters, ): Promise { await this.initialize(); this.assertVectorDimension(query, "Query"); const scored: VectorStoreResult[] = []; await this.scanRows( ` SELECT id, vector, payload FROM ${this.keyspace}.${this.collectionName} `, async (row) => { const vector = this.normalizeVector(row.vector); const payload = this.parsePayload(row.payload); if (!vector || vector.length !== this.dimension) { return; } const item: CassandraVector = { id: String(row.id), vector, payload, }; if (!this.filterVector(item, filters)) { return; } this.pushTopResult( scored, { id: item.id, payload: item.payload, score: this.cosineSimilarity(query, item.vector), }, topK, ); }, Math.max(topK, CassandraDB.PAGE_SIZE), ); return scored; } ``` Sources: [mem0-ts/src/oss/src/vector_stores/cassandra.ts:143-185](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/cassandra.ts#L143-L185) ## Configuration, Connection Management, and Lazy Loading Database vector store initializers implement lazy connection loading to prevent optional peer dependencies (such as `@elastic/elasticsearch`, `pymongo`, `cassandra-driver`, `redis`, or `oracledb`) from breaking base package imports when a provider is not actively in use. Sources: [mem0-ts/src/oss/src/vector_stores/elasticsearch.ts:63-65](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/elasticsearch.ts#L63-L65), [mem0-ts/src/oss/src/vector_stores/mongodb.ts:35-37](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/mongodb.ts#L35-L37) Clients and connection pools are initialized on demand during the first call to `initialize()`, which caches its execution promise (`_initPromise`) to prevent race conditions during concurrent startup sequences. Sources: [mem0-ts/src/oss/src/vector_stores/opensearch.ts:120-126](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/opensearch.ts#L120-L126) ```typescript // Example: Lazy peer loading pattern in OpenSearch and Elasticsearch async initialize(): Promise { if (!this._initPromise) { this._initPromise = this._doInitialize(); } return this._initPromise; } ``` Sources: [mem0-ts/src/oss/src/vector_stores/opensearch.ts:120-126](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/opensearch.ts#L120-L126) | Configuration Property | Target Stores | Purpose & Default Behavior | | :--- | :--- | :--- | | `collection_name` / `tableName` | All Stores | Namespace or table identifier (defaults to `mem0` or `memories`) | | `embeddingModelDims` / `dimension` | All Stores | Vector dimensionality (defaults to `1536`) | | `autoRefresh` | OpenSearch | Controls `indices.refresh()` post-insert (default `false` for Serverless compatibility) | | `useAzureCredential` | Azure MySQL, Azure AI Search | Authenticates via `@azure/identity` DefaultAzureCredential | Sources: [mem0/configs/vector_stores/oracledb.py:41-69](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/configs/vector_stores/oracledb.py#L41-L69), [mem0/vector_stores/opensearch.py:62-81](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/opensearch.py#L62-L81), [mem0-ts/src/oss/src/vector_stores/azure_mysql.ts:34-46](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/azure_mysql.ts#L34-L46) ## Error Handling, Sanitization, and Security Invariants Database vector stores enforce strict input sanitization and parameter binding guards to protect against injection attacks and malformed embeddings. Sources: [mem0/vector_stores/mongodb.py:150-165](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/mongodb.py#L150-L165) - **Filter Key Validation**: Drivers such as Elasticsearch, OpenSearch, pgvector, and Baidu enforce strict regular expression validation on filter keys (`/^[a-zA-Z_][a-zA-Z0-9_]*$/`) to prevent arbitrary query injection through metadata keys. - **Operator Blocking**: MongoDB drivers reject filter values that contain dictionaries (`$ne`, `$gt`, etc.) to prevent NoSQL query operator injection. - **Dimension Checking**: Insert and search operations validate that input vector lengths match configured embedding dimensions before issuing database commands. Sources: [mem0/vector_stores/mongodb.py:150-165](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/mongodb.py#L150-L165), [mem0-ts/src/oss/src/vector_stores/elasticsearch.ts:26-44](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/vector_stores/elasticsearch.ts#L26-L44), [mem0/vector_stores/opensearch.py:25-53](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/opensearch.py#L25-L53) > [!WARNING] > Bypassing filter key validation or passing unvalidated user input directly into MongoDB query objects or SQL string templates can permit query manipulation. Always use the provided validation helpers (`_validate_filter`, `validateFilterValue`) prior to query construction. Sources: [mem0/vector_stores/mongodb.py:150-165](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/vector_stores/mongodb.py#L150-L165) ## Related - [Vector Store Architecture](https://www.doc0.dev/docs/faa36707-7c28-4f69-a18f-700ff61c704e/technical/vector-stores/vector-store-architecture) - [Cloud Vector Backends](https://www.doc0.dev/docs/faa36707-7c28-4f69-a18f-700ff61c704e/technical/vector-stores/cloud-vector-backends) ## Sitemap See the full [sitemap](https://www.doc0.dev/docs/faa36707-7c28-4f69-a18f-700ff61c704e/llms.txt) for all pages in this wiki.