Weavecoder

A Rust coding-agent CLI with a native Agent Swarm for multiple fast parallel requests, prioritizing local models, and an embedded Code Knowledge Graph (tree-sitter) that indexes your entire project codebase.

Features

• Agent Swarm — multiple requests in parallel, natively orchestrated
• Local models first — Ollama, LM Studio, oMLX/llama.cpp, or any OpenAI-compatible endpoint
• Code Knowledge Graph (CKG) — indexes code symbols and relationships with tree-sitter:
  • wvc init — scans the project, extracts functions/classes/imports/calls, and stores them in SQLite + FTS5
  • wvc code-search <query> — hybrid search (FTS5 + embeddings + dependency graph)
  • Graph traversal: "who calls X", "what does Y depend on"
  • Incremental indexing: only re-indexes modified files
• Local embeddings (all-MiniLM-L6-v2) for semantic search by meaning

Installation

The installer scripts are always fetched from the project repository on GitHub — the single source of truth — never from any web domain.

macOS & Linux

curl -fsSL https://raw.githubusercontent.com/nicolasramos-es/weavecoder/main/install.sh | bash

Windows 11 (PowerShell 5.1+)

irm https://raw.githubusercontent.com/nicolasramos-es/weavecoder/main/install.ps1 | iex

The installer verifies the SHA-256 checksum of the binary against the digest published in the release and aborts with a clear error if they don't match. If the repository is private, export GITHUB_TOKEN (or WVC_GITHUB_TOKEN) before running it.

Need Homebrew, source builds, provider setup, or want an agent to set it up for you? Keep reading — Quick Start and From source.

From source

git clone https://github.com/nicolasramos-es/weavecoder.git
cd weavecoder
cargo build --release --bin wvc
# → target/release/wvc

Quick Start

# 1. Connect a local model (e.g. Ollama)
brew install ollama && ollama pull llama3.2
wvc login --provider ollama

# 2. Chat with the agent
wvc --provider ollama --model llama3.2 run 'hello'

# 3. Index a project and search it with the Code Knowledge Graph
wvc init /path/to/project --db ckg.db
wvc code-search "parseConfig" --db ckg.db

Architecture

Crate	Responsibility
wvc-code-graph	Code Knowledge Graph: tree-sitter (Go/Py/TS/Rust), SQLite+FTS5, embeddings, petgraph graph, hybrid search
wvc-embedding	Local embeddings (all-MiniLM-L6-v2, tract-onnx)
wvc-swarm-core	Agent swarm orchestration
wvc-app-core	Agent core (tools, sessions, server)

License

MIT — see LICENSE.

This project builds on the exceptional work of Jeremy Huang (wvc, MIT), on top of which we've added new features and improved the product. The original copyright notice is preserved in full in LICENSE.