ai-blame

Extract provenance from AI agent execution traces.

Like git blame, but for AI-assisted edits. See which AI model wrote each line of code.

Quick Command Reference

Why ai-blame? AI coding assistants modify your files, but git blame only shows who committed the changes—not which AI model wrote them. ai-blame fills this gap.

Installation

Using Python Package Managers (Recommended)

The easiest way to install ai-blame is using Python package managers. No Rust toolchain required!

# Using uv (recommended)
uv add --dev ai-blame

# Using pip
pip install ai-blame

# Using pipx (for global installation)
pipx install ai-blame

This installs a pre-built binary that works exactly like the Rust version—fast, reliable, and dependency-free.

Using Cargo (Rust)

# Install from crates.io
cargo install ai-blame

# Or install from source
git clone https://github.com/ai4curation/ai-blame
cd ai-blame
cargo install --path .

From Pre-built Binaries

Download the latest release from the releases page.

Quick Start

# Check what traces are available
ai-blame stats

# View timeline of all AI actions
ai-blame timeline

# Preview what would be added (stdout report)
ai-blame report --initial-and-recent

# Apply changes (writes annotations / sidecars)
ai-blame annotate --initial-and-recent

# Filter to specific files
ai-blame annotate --pattern ".py"

Demo

Watch a complete walkthrough of all commands:

Shows setup, discovery, line-level blame analysis, and annotation workflows using real traces from ai-blame development.

Performance: Caching

The tool includes DuckDB caching enabled by default to speed up repeated runs. Trace files are parsed once and results are cached in .ai-blame.ddb in your trace directory.

Caching enabled by default:

# Cache is automatically used
ai-blame stats

# Rebuild cache (delete existing cache and re-parse)
ai-blame stats --rebuild-cache

# Disable cache for a specific run
ai-blame stats --no-cache

Expected speedup:

• First run: ~55 seconds (builds cache)
• Subsequent runs (unchanged traces): ~3-5 seconds (90% faster 🚀)
• Incremental updates: Proportional to changed files

Cache behavior:

• Claude traces: All-or-nothing invalidation (if any trace file changes, all are re-parsed due to cross-file UUID dependencies)
• Codex traces: Per-file invalidation (each session/file is independent)
• Staleness detection: Modified time + file size comparison

Cache management:

# Delete the cache file to reset
rm .ai-blame.ddb

# Or use the CLI flag to rebuild
ai-blame stats --rebuild-cache

# Disable caching globally
export AI_BLAME_NO_CACHE=1

Desktop App

The Tauri-based desktop app provides a visual interface for exploring AI-assisted code edits.

Features:

• Blame Viewer — Browse files with line-by-line AI attribution and details panel
• Timeline — Chronological view of all AI edits with navigation to source files
• Transcripts — Search and explore AI conversation sessions with full message content
• Settings — Configure project paths and caching options

# Run the desktop app
cd src-tauri && cargo run --release

See the Desktop App documentation for full details.

Documentation

This repo uses MkDocs (Material) for user/CLI documentation.

python -m venv .venv
source .venv/bin/activate
pip install -r docs/requirements.txt
mkdocs serve

Output Examples

YAML/JSON files — Append directly

# config.yaml
name: my-project
version: 1.0

edit_history:
  - timestamp: "2025-12-01T08:03:42+00:00"
    model: claude-opus-4-5-20251101
    agent_tool: claude-code
    action: CREATED

Code files — Sidecar or comments

# main.py (with comment policy)

def hello():
    print("Hello, world!")

# --- edit_history ---
# - timestamp: '2025-12-01T08:03:42+00:00'
#   model: claude-opus-4-5-20251101
#   action: CREATED
# --- end edit_history ---

Or use sidecar files: main.py → main.history.yaml

Configuration

Create .ai-blame.yaml in your project root:

defaults:
  policy: sidecar
  sidecar_pattern: "{stem}.history.yaml"

rules:
  - pattern: "*.yaml"
    policy: append
  - pattern: "*.json"
    policy: append
    format: json
  - pattern: "*.py"
    policy: comment
    comment_syntax: hash
  - pattern: "tests/**"
    policy: skip

Supported Agents

Differences from Python Version

This Rust port maintains CLI compatibility with the Python version but offers significant improvements:

• Better performance - 10-100x faster trace parsing and file processing
• Static typing - Compile-time guarantees for correctness
• Single binary - No runtime dependencies
• Memory safety - Rust's ownership system prevents common bugs
• Easy Python installation - Install via pip or uv with pre-built wheels

For Python Users

You can still install ai-blame using Python package managers:

uv add --dev ai-blame
# or
pip install ai-blame

This installs the same high-performance Rust binary—no Rust toolchain needed! The CLI commands remain the same, so it's a drop-in replacement for the Python version.

Note: The Python API from the original version is not available in this Rust port. The CLI provides all functionality. If you need programmatic access, please open an issue describing your use case.

Development

# Run tests
cargo test

# Run with debug output
RUST_LOG=debug cargo run -- report

# Build for release
cargo build --release

# Format code
cargo fmt

# Lint code
cargo clippy

Repository Structure

This repository uses a Cargo workspace to organize the CLI and Tauri UI components:

ai-blame/
├── src/                 # Core library (ai-blame crate)
│   ├── lib.rs          # Library root
│   ├── main.rs         # CLI binary entry point
│   ├── cli.rs          # CLI command parsing (feature-gated)
│   ├── blame.rs        # Line-level blame computation
│   ├── config.rs       # Configuration loading
│   ├── extractor.rs    # Trace file parsing
│   ├── models.rs       # Data models
│   └── updater.rs      # File annotation logic
├── src-tauri/           # Tauri desktop app (ai-blame-ui crate)
│   ├── Cargo.toml      # UI-specific dependencies
│   ├── tauri.conf.json # Tauri configuration
│   ├── src/main.rs     # Tauri backend (invokes core library)
│   └── icons/          # Application icons for bundling
├── ui/                  # Static HTML/CSS/JS frontend
│   ├── index.html      # Main UI layout
│   ├── app.js          # UI logic
│   └── styles.css      # Styling
├── tests/               # Integration tests
├── Cargo.toml           # Workspace root + core library config
└── tools/               # Development utilities
    └── generate_icons.py

Workspace Design

• Core Library (ai-blame): Contains all reusable logic for parsing traces, computing blame, and updating files
• CLI Binary: Built with the cli feature flag (enabled by default)
• Tauri UI (ai-blame-ui): Depends on the core library with default-features = false to avoid pulling in CLI-only dependencies like clap

This structure follows Tauri best practices by keeping the Tauri binary in src-tauri (required for tauri dev/build to work) while sharing code through the core library.

Feature Flags

• cli (default): Enables CLI command parsing with clap. Disable with --no-default-features when using only the library API.

License

BSD-3-Clause

Contributing

Contributions welcome! This is a port of the Python ai-blame project.

PRs especially welcome for:

• Additional agent support (Cursor, Aider, Copilot, etc.)
• Performance improvements
• Bug fixes
• Documentation improvements