Skip to main content

VecLabs

Decentralized vector memory for AI agents. Rust HNSW core. Solana on-chain provenance. 88% cheaper than Pinecone.

Overview

Most vector databases are centralized infrastructure you rent access to. Your data lives on their servers. You trust their uptime, their pricing, and their word that nothing has changed. VecLabs is built differently. Vectors are encrypted with your Solana wallet key and stored on decentralized storage. After every write, a 32-byte Merkle root is posted to Solana β€” a cryptographic fingerprint of your entire collection, immutable and publicly verifiable. The query engine is a Rust HNSW implementation with no garbage collector, delivering consistent sub-5ms latency that Python and Go-based engines cannot match under load. The result: a vector database that is faster, cheaper, and verifiable by anyone β€” without trusting VecLabs.

Key Features

Sub-5ms Query Latency

Rust HNSW implementation with no garbage collector for consistent performance under load

On-Chain Verification

Merkle root proofs on Solana provide cryptographic verification of collection integrity

88% Cost Savings

Decentralized storage on Shadow Drive eliminates cloud infrastructure markups

Wallet-Based Encryption

AES-256-GCM encryption with keys derived from your Solana wallet

Performance Benchmarks

Measured on Apple M2, 16GB RAM. 100K vectors, 384 dimensions, top-10 query.
Databasep50p95p99Monthly Cost (1M vectors)
VecLabs1.9ms2.8ms4.3ms~$8
Pinecone s1~8ms~15ms~25ms$70
Qdrant~4ms~9ms~15ms$25+
Weaviate~12ms~25ms~40ms$25+

Live on Solana Devnet

Quick Example

import { SolVec } from 'solvec';

const sv = new SolVec({ network: 'devnet' });
const collection = sv.collection('agent-memory', { dimensions: 1536 });

await collection.upsert([{
  id: 'mem_001',
  values: [...],
  metadata: { text: 'User prefers dark mode' }
}]);

const results = await collection.query({ vector: [...], topK: 5 });

// Verify collection integrity against on-chain Merkle root
const proof = await collection.verify();
console.log(proof.solanaExplorerUrl);

Architecture

Three layers. Each does only what it is best at.
SolVec SDK (TypeScript / Python)
.upsert()  .query()  .delete()  .verify()
      |           |           |
      v           v           v
 Rust HNSW    Shadow Drive   Solana
 (in memory)  (encrypted     (32-byte
 sub-5ms p99   vectors)       Merkle root)
 Speed Layer  Storage Layer   Trust Layer

Installation

Install the TypeScript or Python SDK

Quickstart

Get up and running in 5 minutes

API Reference

Explore the complete API documentation

Why VecLabs?

Rust HNSW with no garbage collector delivers consistent sub-5ms p99 latency that Python and Go-based engines cannot match. No latency spikes under concurrent load.
Every collection has a Merkle root posted to Solana after every write. Any party can independently verify the exact state of your agent’s memory without trusting VecLabs.
Decentralized storage means no cloud infrastructure markups. **8/monthfor1Mvectorsβˆ—βˆ—vs8/month for 1M vectors** vs 70/month on Pinecone s1.
The SolVec API is intentionally shaped to match Pinecone’s client. Migration is three line changes. Existing LangChain and AutoGen integrations work with minimal modifications.

Current Status

This is alpha software. The API surface is stable and will not change. Backend persistence is in progress.
ComponentStatus
Rust HNSW coreβœ… Complete β€” 31 tests, 4.3ms p99 at 100K vectors
AES-256-GCM encryptionβœ… Complete
Merkle tree + proof generationβœ… Complete
Solana Anchor programβœ… Live on devnet β€” 6/6 tests passing
TypeScript SDK⚠️ Alpha β€” npm install solvec@alpha
Python SDK⚠️ Alpha β€” pip install solvec --pre
Shadow Drive persistence🚧 In progress β€” vectors currently in-memory
WASM Rust bridge🚧 In progress β€” SDK uses JS fallback
Mainnet deploymentπŸ“… Planned
LangChain integrationπŸ“… Planned

Next Steps

1

Install the SDK

Choose TypeScript or Python and install the packageInstallation Guide β†’
2

Follow the quickstart

Create your first collection and run your first query in 5 minutesQuickstart β†’
3

Understand the architecture

Learn how VecLabs combines Rust HNSW, Shadow Drive, and SolanaArchitecture β†’
4

Explore the API

Dive into the complete TypeScript and Python SDK referenceAPI Reference β†’

Build docs developers (and LLMs) love