lmfetch

Effortlessly turn codebase into context for your LLMs.

lmfetch is a lightning-fast CLI tool that fetches, chunks, supports local files and remote GitHub repositories, and ranks code context. It's designed to fit the most relevant code into your token budget, so your LLM can understand the implementation details without the noise.

Install

Mac & Linux

curl -fsSL https://raw.githubusercontent.com/Abdulmumin1/lmfetch/main/install.sh | bash

Windows (PowerShell) Coming soon.

Usage

Fetch local code

lmfetch . "how does authentication work"

Fetch from GitHub

lmfetch https://github.com/vercel/ai "explain tool calling"

Common Options

# Set a token budget (default: 50k)
lmfetch . "query" -b 10k

# Force processing of large files
lmfetch . "query" --force-large

# Clear cache
lmfetch --clean-cache

# Output context to file
lmfetch . "query" -o context.md

How It Works

1. Smart Chunking: Understands AST (classes, functions) for Python, TS, Go, Rust, and more.
2. Hybrid Ranking: Combines keyword matching, vector embeddings, and dependency graph centrality.
3. HyDE Reasoning: Generates hypothetical code to match your intent, not just your keywords.
4. Privacy: Runs locally. Embeddings use your own API keys (OpenAI/Google).

Architecture

See ARCHITECTURE.md for a deep dive into the pipeline.

License

MIT