VectorLane

Searchable local memory for AI agents.

VectorLane is a local vector memory store designed for AI agents. It provides fast, offline-capable semantic search over text, documents, and data using embeddings. No cloud APIs required — everything runs on your machine.

Architecture

Documents, MemoryLane exports, and ContextLane exports flow into the Ingest pipeline. Content is chunked, embedded, and stored in the local vector store. Search returns ranked results with citations via CLI, HTTP API, or MCP server.

Related MCP server: OmniHub

Features

• Fully local — No external API calls, no data leaves your machine

• Offline embeddings — Uses local-hash (256-dim) by default, with OpenAI/HuggingFace options

• Multiple backends — JSONL (default), SQLite, in-memory

• MCP integration — Native Model Context Protocol support for AI assistants

• REST API — Full HTTP API on port 3090 (configurable)

• CLI — Complete command-line interface for all operations

• SDK — JavaScript/TypeScript and Python client libraries

• Citation tracking — Automatic source attribution for search results

• Multi-source ingestion — Text, URLs, files, and bulk imports

Quick Start

Install

# npm
npm install -g @talocode/vectorlane

# pip
pip install talocode-vectorlane

Initialize and Use

# Start the server
vectorlane serve

# Initialize a project
vectorlane init

# Ingest some text
vectorlane ingest-text "The quick brown fox jumps over the lazy dog."

# Search
vectorlane search "fox"

Or use the SDK

import { VectorLane } from '@talocode/vectorlane';

const vl = new VectorLane();
await vl.init();
await vl.ingestText('The quick brown fox jumps over the lazy dog.');
const results = await vl.search('fox');
console.log(results);

from vectorlane import VectorLane

vl = VectorLane()
vl.init()
vl.ingest_text("The quick brown fox jumps over the lazy dog.")
results = vl.search("fox")
print(results)

Architecture

+-----------+     +-----------+     +-----------+
|  CLI /    |---->|  REST API |---->|  Vector   |
|  SDK      |     |  :3090    |     |  Store    |
+-----------+     +-----------+     +-----------+
                         |                |
                  +------+------+  +------+------+
                  |  Embedding  |  |  Backend    |
                  |  Engine     |  |  (JSONL/   |
                  |  (local/    |  |   SQLite)  |
                  |   openai)   |  +-------------+
                  +-------------+

Installation

See docs/INSTALL.md for detailed installation instructions.

Requirements

• Node.js 18+ (for npm package)

• Python 3.9+ (for pip package)

• No external dependencies required for basic usage

Quick Install

# npm
npm install -g @talocode/vectorlane

# pip (Python SDK)
pip install talocode-vectorlane

CLI Reference

See docs/CLI.md for the full CLI reference.

Core Commands

Collection Commands

Configuration Commands

Import Commands

SDK Reference

See docs/SDK.md for the full SDK reference.

JavaScript/TypeScript

import { VectorLane } from '@talocode/vectorlane';

const vl = new VectorLane({ port: 3090 });

// Initialize
await vl.init();

// Create a collection
await vl.collection.create('docs');

// Ingest text
await vl.ingestText('Your text content here', { collection: 'docs' });

// Ingest a URL
await vl.ingestUrl('https://example.com/article', { collection: 'docs' });

// Search
const results = await vl.search('your query', { collection: 'docs', limit: 5 });

// Get stats
const stats = await vl.collection.stats('docs');

Python

from vectorlane import VectorLane

vl = VectorLane(port=3090)

# Initialize
vl.init()

# Create a collection
vl.collection.create("docs")

# Ingest text
vl.ingest_text("Your text content here", collection="docs")

# Ingest a URL
vl.ingest_url("https://example.com/article", collection="docs")

# Search
results = vl.search("your query", collection="docs", limit=5)

# Get stats
stats = vl.collection.stats("docs")

REST API

See docs/API.md for the full API reference.

Quick Reference

MCP Integration

See docs/MCP.md for MCP configuration.

VectorLane provides native MCP (Model Context Protocol) support. Add to your MCP config:

{
  "mcpServers": {
    "vectorlane": {
      "command": "vectorlane",
      "args": ["mcp"]
    }
  }
}

Available MCP Tools

Vector Store

See docs/VECTOR_STORE.md for backend details.

Embeddings

See docs/EMBEDDINGS.md for embedding options.

Search

See docs/SEARCH.md for search capabilities.

vectorlane search "machine learning"
vectorlane search "API docs" --collection docs --limit 10
vectorlane search "error handling" --threshold 0.7

Chunking

See docs/CHUNKING.md for chunking strategies.

• Fixed-size — Default, 512 tokens per chunk with 50 token overlap

• Sentence — Splits on sentence boundaries

• Paragraph — Splits on paragraph boundaries

• Recursive — Hierarchical splitting with fallbacks

Citations

See docs/CITATIONS.md for citation tracking.

Every search result includes source attribution:

{
  "id": "abc123",
  "text": "The quick brown fox...",
  "score": 0.95,
  "citation": {
    "source": "document.txt",
    "page": 1,
    "offset": 0,
    "timestamp": "2026-07-15T10:30:00Z"
  }
}

Integrations

See docs/INTEGRATIONS.md for integration guides.

• MemoryLane — Import and sync conversation history

• ContextLane — Import and sync context documents

• MCP-compatible tools — Works with Claude, Cursor, and other MCP clients

• LangChain — VectorLane retriever integration

• LlamaIndex — VectorLane vector store integration

Configuration

Configuration is stored in ~/.vectorlane/config.json:

{
  "port": 3090,
  "backend": "jsonl",
  "embedding": "local-hash",
  "storage_path": "~/.vectorlane/data",
  "default_collection": "default",
  "chunk_size": 512,
  "chunk_overlap": 50
}

vectorlane config get port
vectorlane config set port 3091
vectorlane config list

Storage

All data is stored locally in ~/.vectorlane/:

~/.vectorlane/
  config.json          # Configuration
  data/                # Vector store data
    collections/       # Collection data
    embeddings/        # Cached embeddings
  logs/                # Application logs

Troubleshooting

See docs/TROUBLESHOOTING.md for common issues.

Quick Fixes

Server won't start

vectorlane doctor

Port in use

vectorlane config set port 3091
vectorlane serve

Reset everything

vectorlane clear
vectorlane init

Roadmap

See docs/ROADMAP.md for the development roadmap.

v0.1.0 (Current)

• Core vector store with JSONL backend

• local-hash embedding model

• CLI with all core commands

• REST API

• MCP integration

• Python and JavaScript SDKs

v0.2.0

• SQLite backend

• OpenAI embedding support

• HuggingFace embedding support

• LangChain integration

v0.3.0

• Hybrid search (keyword + semantic)

• Multi-modal embeddings (images)

• Distributed mode

• Web UI dashboard

Contributing

1. Fork the repository

2. Create a feature branch (git checkout -b feature/amazing-feature)

3. Commit your changes (git commit -m 'Add amazing feature')

4. Push to the branch (git push origin feature/amazing-feature)

5. Open a Pull Request

License

MIT License - see LICENSE for details.

Support

• Documentation: docs/

• Issues: GitHub Issues

• npm: @talocode/vectorlane