Aptu

crates.io docs.rs REUSE SLSA Level 3 OpenSSF Best Practices

Surface what needs review, skip what does not -- without leaving your terminal.

AI-Powered Triage Utility - AI-powered OSS issue triage and PR review with contribution history tracking.

Aptu is a context-engineering experiment: instead of throwing big models at problems, it crafts tight prompts that let smaller models do the job with fewer tokens and surprising precision.

Benchmarks

Head-to-head comparison of aptu+mercury-2 (Mercury 2, a small diffusion-based LLM by Inception Labs) vs a raw claude-opus-4.6 call (no schema, no rubric, no AST context) across 6 fixtures (3 triage, 3 PR review).

Arm	Quality (mean, /5)	Cost/call	Latency p50
aptu+mercury-2	4.8/5	$0.0011	1,934 ms
raw claude-opus-4.6	2.2/5	$0.0193	16,032 ms

Measured across aptu #737, #850, #1094 (triage) and #1091, #1098, #1101 (PR review); n=1 per fixture.

aptu+mercury-2 is 17x cheaper and 8x faster than a raw claude-opus-4.6 call, while scoring more than twice as high on the structured rubric.

See docs/BENCHMARKS.md for full methodology, fixture breakdown, and C1-C5 scores.

Demo

Features

• AI Triage - Summaries, suggested labels, clarifying questions, and contributor guidance
• Issue Discovery - Find good-first-issues from curated repositories
• PR Analysis - AI-powered pull request review and feedback; aptu pr create --diff <file> applies a patch, commits, and opens a PR
• Structural Graph Context - Injects petgraph BFS blast-radius context into pr review prompts (opt-in, --features graph)
• Model-Tier Routing - Routes large PRs to a higher-capability model tier automatically based on estimated prompt size
• Prompt Customization - Override built-in system prompts per operation or append custom guidance via config
• GitHub Action - Auto-triage incoming issues with labels and comments
• Multiple Providers - Anthropic, Cerebras, Gemini, Groq, OpenRouter (default), Z.AI, and ZenMux; free-tier models available via OpenRouter
• Local History - Track your contributions offline
• Multiple Outputs - Text, JSON, YAML, Markdown, and SARIF
• Claude OAuth - Authenticate with Anthropic via ~/.claude/credentials.json (written by the Claude desktop app); no API key required
• Dependency Enrichment - Automatically fetches upstream release notes for dependency bump PRs (Renovate / Dependabot)
• Observability - Per-review context JSONL (APTU_CONTEXT_FILE) and token usage metrics (APTU_METRICS_FILE) for explainability and budget debugging (see Observability and Environment Variables)
• OpenSSF Best Practices Silver - Fewer than 1% of open source projects reach this level; see Security

Installation

# Homebrew (macOS/Linux)
brew install clouatre-labs/tap/aptu

# Cargo-binstall (fast)
cargo binstall aptu-cli

# Cargo
cargo install aptu-cli

Quick Start

aptu auth login            # Authenticate with GitHub
aptu repo list             # List curated repositories
aptu issue list --repo block/goose          # Browse issues
aptu issue triage block/goose#123    # Triage with AI
aptu issue triage block/goose#123 --dry-run  # Preview
aptu history               # View your contributions

Observability

export APTU_METRICS_FILE=metrics.jsonl
aptu pr review owner/repo#123   # token usage appended to metrics.jsonl per run

See docs/GITHUB_ACTION.md for full field reference.

Security Scanning

Aptu includes built-in security pattern detection for PR reviews. Scanning is performed locally, and no code is sent to external services.

aptu pr review owner/repo#123                       # Review with security scanning
aptu scan-security . --sarif-output findings.sarif  # SARIF for GitHub Code Scanning

See docs/SECURITY_SCANNING.md for SARIF upload and GitHub integration.

Prompt Customization

Aptu's built-in system prompts are compiled into the binary as defaults. You can override them per operation at runtime or append project-specific guidance globally.

See docs/CONFIGURATION.md for file paths, operation names, and examples.

GitHub Action

Auto-triage new issues with AI using any supported provider.

- uses: clouatre-labs/aptu@v0
  with:
    github-token: ${{ secrets.GITHUB_TOKEN }}
    openrouter-api-key: ${{ secrets.OPENROUTER_API_KEY }}

Options: apply-labels, no-comment, skip-labeled, dry-run, model, provider.

See docs/GITHUB_ACTION.md for setup and examples.

Configuration

See docs/CONFIGURATION.md for AI provider setup.

Models

Use aptu models list to discover available models from all configured providers.

Discovering models

aptu models list                                # all providers
aptu models list --provider openrouter          # OpenRouter only

Filtering and sorting

Flag	Description
--provider	Filter to a specific provider
--sort name|context	Sort by name or context window size
--min-context N	Show only models with at least N tokens of context
--filter TEXT	Filter by name or ID (case-insensitive substring match)

Free-tier models

OpenRouter exposes pricing data for each model. Models with zero prompt and completion cost are labeled free in the output. Use --provider openrouter to browse free models.

Security

• SLSA Level 3 - Provenance attestations for all releases
• REUSE/SPDX - License compliance for all files
• Signed Commits - GPG-signed commits required
• Dependency Scanning - Automated updates via Renovate

See SECURITY.md for reporting and verification.

Architecture

Aptu is a multi-crate Rust workspace. See docs/ARCHITECTURE.md for the full crate structure, data flow, and key dependencies.

Roadmap

See docs/ROADMAP.md for the project direction across near-term, medium-term, and long-term horizons.

Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines. See docs/REPO-STANDARDS.md for a full artifact map and rationale covering CI workflows, tooling, and security controls.

License

Apache-2.0. See LICENSE.