Home (original) (raw)

Vectra

A local, file-backed vector database with cross-language gRPC access.

Get Started View on GitHub

Vectra works like a local Pinecone or Qdrant: each index is a folder on disk, loaded into memory for fast queries. Filter by metadata using MongoDB-style operators, rank by cosine similarity, and get results in under a millisecond for small indexes.

Key capabilities

| | | | | ---------------------------- | ----------------------------------------------------------------------------- | | Zero infrastructure | Everything lives in a local folder — no servers or managed services | | Fast lookups | Sub-millisecond to low-millisecond latency | | Pinecone-style filtering | MongoDB query operators for metadata filtering | | Multiple embeddings | OpenAI, Azure, OSS endpoints, or local HuggingFace models (no API key) | | Pluggable storage | Filesystem, IndexedDB (browsers), in-memory, or custom backends | | Browser & Electron | Dedicated vectra/browser entry point with local embeddings and IndexedDB | | CLI included | Manage indexes, serve gRPC, watch folders from the terminal | | Cross-language | Built-in gRPC server with bindings for Python, C#, Rust, Go, Java, TypeScript | | Flexible format | JSON (human-readable) or Protocol Buffers (40-50% smaller) |

Choose your path

What’s New in v0.14.x

• Browser & Electron support — vectra/browser entry point with IndexedDBStorage and TransformersEmbeddings
• Local embeddings — LocalEmbeddings and TransformersEmbeddings run HuggingFace models with no API key
• Protocol Buffers — opt-in binary format, 40-50% smaller files
• gRPC server — vectra serve exposes 19 RPCs for cross-language access
• FolderWatcher — auto-sync directories into a document index
• Language bindings — vectra generate scaffolds clients for 6 languages

See the Changelog for breaking changes and migration details.

Documentation