100 Databases for AI Apps resources for developers

The industry standard for adding vector similarity search to Postgres. Supports L2 distance, inner product, and cosine distance for embeddings.

Postgres implementation featuring storage/compute separation and an HTTP driver for accessing data from serverless environments without connection limits.

A managed pgvector environment that includes a client-side library for performing vector similarity searches directly from the frontend or edge functions.

SQLite-based edge database that supports vector extensions, optimized for low-latency access in globally distributed applications.

Native vector search capabilities within Cloudflare's distributed SQLite database, allowing for semantic search at the edge.

Enables vector search on unstructured BSON data, allowing developers to store metadata and embeddings in a single document.

A purpose-built library for using Redis as a vector database, ideal for high-speed semantic caching and session-based RAG.

A specialized query caching layer for MySQL that can be used to accelerate the retrieval of AI-related metadata and conversation histories.

High-performance analytical database used for massive-scale vector similarity searches where ingestion speed is a bottleneck.

MySQL-compatible distributed database with built-in vector search support, designed for scaling AI workloads horizontally.

Provides TypeScript-first schema definitions for vector columns and helper functions for distance-based filtering.

Implementation pattern for executing vector similarity searches using raw SQL within Prisma while maintaining type safety for other models.

Type-safe SQL query builder integration for pgvector, allowing for complex joins between relational data and vector results.

A database schema management tool that supports HNSW and IVFFlat index declarations for pgvector in declarative HCL files.

Using custom value transformers to map float arrays to Postgres vector types in NestJS or Node.js applications.

A global database cache and connection pooler that solves the connection exhaustion problem common in serverless AI apps.

The primary integration for Python-based AI backends to manage vector storage within a relational SQLAlchemy model.

CLI tool for generating SQL migrations that correctly handle adding vector dimensions and modifying HNSW index parameters.

A graph-relational database that treats vectors as first-class citizens with a high-level query language (EdgeQL) for RAG.

Exposing pgvector similarity functions as RESTful endpoints by wrapping them in Postgres stored procedures.

Configuring transaction-mode pooling to handle the high volume of short-lived connections from AI worker processes.

A WebSocket-based driver that allows querying Postgres over HTTP, bypassing TCP connection overhead in edge environments.

Optimizing m and ef_construction parameters in pgvector to balance search speed against memory consumption.

Strategy for determining the optimal number of lists for IVFFlat indexes based on total row count for faster search.

Using vector similarity to check if a similar prompt has been answered, reducing LLM API costs and database load.

Splitting large vector tables by time or metadata category to improve index maintenance and query performance.

Reducing the precision of embeddings (e.g., from float32 to int8) to decrease the storage footprint and memory usage in Postgres.

An optimized connection pooler designed specifically for massive scale in serverless Postgres environments.

A serverless vector database with an HTTP API, designed for use cases where zero-configuration and edge compatibility are required.

Transition path for applications that outgrow pgvector, providing a dedicated engine for billion-scale vector search.