Ataraxy-Labs/weave

Entity-level git merge driver. Resolves false conflicts git invents when independent agents edit the same file. ~95% reduction vs. line-based merge.

weave – Entity‑level semantic merge for Git

What it is

  • A Rust‑based tool that replaces Git’s line‑by‑line merge with a semantic merge that works on functions, classes, JSON keys and other language entities. It parses the three versions of a file (base, ours, theirs) with tree‑sitter, matches entities by name/scope, and merges them intelligently.

Why it matters

  • When multiple AI agents (or developers) edit the same file, Git often flags a conflict even though the changes are independent (e.g., two new functions added side‑by‑side). weave resolves those false conflicts automatically, letting agents continue without human intervention.

Key features

  • Entity‑level 3‑way merge – works on functions, classes, object keys, etc., across 38 supported languages.
  • Deterministic & stateless – given the same three revisions it always produces the same result, similar to git merge-file.
  • Conflict context – when a real conflict occurs, the marker includes the entity name, type, and a short “refused_by” reason, making manual resolution easier.
  • weave land – a command that validates the merged tree, optionally runs a custom resolver, and stages only files that pass structural checks (PROVEN, VERIFIED, REFUSED).
  • Integration with Git and Jujutsu – provides a weave-driver merge driver and a weave setup command that writes the necessary .gitattributes rules automatically.
  • CRDT‑backed coordination layer (separate weave‑crdt crate) for live multi‑agent editing before the changes hit Git.
  • Benchmarks & tests – 441 unit/integration tests, plus extensive benchmark suites comparing against Git and Mergiraf on real‑world repositories (git, Flask, CPython, Go, TypeScript).

Installation

# Homebrew (macOS / Linux)
brew install weave

# From source (requires Rust toolchain)
git clone https://github.com/Ataraxy-Labs/weave
cd weave
cargo install --path crates/weave-cli      # `weave` command
cargo install --path crates/weave-driver   # driver invoked by Git

Typical workflow

  1. Run weave setup (or --global, --local, --off) to install the merge driver for all supported file types.
  2. Merge as usual: git merge <branch>.
  3. If weave produces conflict markers, use weave explain <file> for detailed entity‑level info.
  4. After editing, run weave check to verify the tree.
  5. Finish with weave land (optionally providing a custom resolver) to stage the resolved files.

Ecosystem & related projects

  • sem – semantic version control.
  • inspect – semantic code review.
  • opensessions – tmux sidebar for coding agents.
  • weave‑crdt – a CRDT layer for live multi‑agent collaboration (mentioned but not part of this repo).

Maturity

  • Latest stable release 0.5.3 (as of Sep 2026).
  • CI runs on Linux and Windows; all tests pass (441 + tests).
  • Benchmarked on thousands of real merge commits with published results.

License

  • Dual‑licensed under MIT OR Apache‑2.0.

Bottom line weave is a production‑ready, open‑source solution for reducing false merge conflicts in codebases where many automated agents (or developers) edit the same files. By merging at the level of language entities rather than raw lines, it lets teams ship changes faster and with fewer manual interventions.

Related

  • Project
    WeaveMindAI/weft

    Weft is a compiled, graph‑based language and runtime for orchestrating AI workflows. It lets you connect LLM calls, agents, APIs, databases, and human‑in‑the‑loop steps as nodes in a typed graph, validates the design before execution, persists long‑running jobs, and records a journal of all data flows. Integrated with the AI assistant “Tangle”, it supports automatic code generation, testing, and versioned runs. Early‑stage but usable, with Docker‑based installer, VS Code extension, and examples like a WhatsApp support bot.

  • Dispatch
    Sem: Semantic Code Understanding via Git Entities

    Sem is a CLI tool that provides semantic, entity-level understanding of Git repositories, improving AI agent accuracy by 2.3x over raw line diffs.

  • Project
    feder-cr/git-ai

    gitai is an open‑source, server‑centric version‑control system modeled on Anthropic’s Scion design. It moves Git history queries, merges, conflict resolution and verification to a stateless server backed by a relational DB and object storage, enabling thousands of short‑lived AI coding agents to fetch only needed objects, land changes concurrently via compare‑and‑swap, and automatically resolve conflicts with any LLM provider. The project includes a CLI (`graft`), a Go SDK, optional caching daemons, and full documentation. It is at version 0.1.0, MIT‑licensed, and ready for community contributions.

  • Project
    Ataraxy-Labs/sem

    A semantic version control tool built on Git that diffs code at the entity level (functions, classes, methods) rather than by lines, designed for both humans and AI agents.

  • Dispatch
    Foremerge 0.5.0 – Intent‑Based Conflict Detection for Parallel Coding Agents

    Foremerge 0.5.0 introduces a local‑first protocol that lets AI coding agents publish their intended semantic changes before editing, allowing deterministic, advisory conflict warnings that Git alone cannot provide.