Ahnlich: In-memory vector database

A simple, intuitive API

Go from connection to semantic search in four steps, in the language you already work in.

$pip install ahnlich-client-py

connect.py

01 / 04

import asyncio
from grpclib.client import Channel
from ahnlich_client_py.grpc.services.ai_service import AiServiceStub


async def main():
    async with Channel(host="127.0.0.1", port=1370) as channel:
        client = AiServiceStub(channel)
        # ready to talk to Ahnlich

Everything you need to ship semantic search

One fast, open-source engine that handles it all: embed, index, filter, and retrieve, so you can focus on your product, not your infrastructure.

Blazing fast

A lock-free, in-memory core built in Rust with SIMD-accelerated similarity search for low-latency retrieval at scale.

AI built in

A first-class AI proxy runs ONNX embedding models for you. Send raw text or images and let Ahnlich handle the vectors.

Hybrid search

Combine semantic vector search with structured metadata predicates to filter results with surgical precision.

Pluggable models & metrics

Bring your own embedding models and choose the similarity metric (cosine, Euclidean, dot product) per store.

Durable by design

Snapshot persistence with deterministic hashing keeps your data consistent across restarts and upgrades.

Polyglot clients

gRPC and Protocol Buffers power native SDKs for Rust, Python, and Node.js so you can build in any stack.

Featured Example

Build your own music recognition

Index your music library and search by playing audio clips. Like Shazam or YouTube's audio search, powered by ahnlich-ai with CLAP embeddings and ahnlich-db for semantic similarity.

Try the Audio Search Demo View Source Code

Record & Match

Capture audio from your microphone and find matching songs in your library

Long Audio Support

Index songs up to 10 minutes with automatic chunking and overlap

High Accuracy

50-60% similarity scores on clean recordings with CLAP embeddings

Quickstart

Up and running in seconds

Install a single binary and start indexing. No cluster to babysit, no heavyweight dependencies. Just a fast vector store ready for your embeddings.

Read the Quickstart Explore client SDKs →

bash — ahnlich

# Run the Ahnlich DB server

$docker run -d -p 1369:1369 ghcr.io/deven96/ahnlich-db:latest

# ...or spin up the AI proxy for automatic embeddings

$docker run -d -p 1370:1370 ghcr.io/deven96/ahnlich-ai:latest

Start building smarter search today

Grab a release for Mac or Linux, or build from source. Ahnlich is free and open source with an active community.

Download now Star on GitHub