<p align="center"> <img src="docs/logo.png" alt="Treliq" width="140" /> </p> <h3 align="center">AI-Powered PR & Issue Triage for Maintainers & Enterprise Teams</h3> <p align="center"> <a href="https://www.npmjs.com/package/treliq"><img src="https://img.shields.io/npm/v/treliq?style=flat-square&color=CB3837&logo=npm" alt="npm version" /></a> <a href="https://www.npmjs.com/package/treliq"><img src="https://img.shields.io/npm/dm/treliq?style=flat-square&color=CB3837" alt="npm downloads" /></a> <a href="https://github.com/mahsumaktas/treliq/actions"><img src="https://img.shields.io/github/actions/workflow/status/mahsumaktas/treliq/ci.yml?branch=main&style=flat-square" alt="CI" /></a> <img src="https://img.shields.io/badge/tests-428_passing-2DA44E?style=flat-square" alt="Tests" /> <img src="https://img.shields.io/badge/signals-21+12-8B5CF6?style=flat-square" alt="33 Signals" /> <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-blue?style=flat-square" alt="License: MIT" /></a> </p>

Treliq scores, deduplicates, and ranks pull requests and issues so maintainers focus on what matters. Available as a CLI, REST server, and GitHub Action.

The Problem

Code review tools (CodeRabbit, Greptile, Copilot) review diffs. None answer the maintainer's real questions:

"These 5 PRs fix the same bug — which one is best?"
"Show me the top 10 PRs I should review today."
"Auto-close all the duplicate PRs and spam issues."

Code Review != PR Triage. Treliq fills the gap.

Quick Start

# Install
npm install -g treliq

# Score all open PRs (21 signals, no API key needed)
npx treliq scan -r owner/repo --no-llm

# Score PRs + issues together
npx treliq scan -r owner/repo --include-issues

# Find duplicate PR clusters
npx treliq dedup -r owner/repo

# Score a single PR with LLM (needs GEMINI_API_KEY)
npx treliq score -r owner/repo -n 123 -f markdown

# Interactive setup wizard
npx treliq init

Output Example

 #  PR                                          Score  Tier      Scored By
 1  feat: add streaming response support        92     critical  sonnet
 2  fix: resolve memory leak in scanner         87     high      sonnet
 3  refactor: extract scoring engine             71     high      haiku
 4  chore(deps): bump express to v5             45     normal    heuristic
 5  docs: update API reference                  28     low       heuristic

Duplicate Clusters:
  Cluster 1 (93% similarity): #42, #67, #89 — "fix rate limiter"
  Cluster 2 (87% similarity): #15, #31 — "add dark mode"

Spam: #99 (empty diff), #101 (AI-generated boilerplate)

Dashboard

Live Demo

Key Features

21 PR Signals + 12 Issue Signals — CI status, test coverage, merge conflicts, spam detection, intent classification, contributor trust, and more (full list)
Cascade Pipeline — Heuristic pre-filter -> Haiku -> Sonnet. ~$13 for 4000 PRs (vs $27 Sonnet-only)
Dual Scoring — ideaScore (is the idea valuable?) + implementationScore (is the code good?) + readinessScore (is it merge-ready?)
Cross-type Dedup — PRs and issues in the same vector space; LLM-verified clusters
Auto-Actions — Close duplicates, close spam, auto-merge, auto-label by intent (dry-run by default)
Multi-Provider LLM — Gemini (free), OpenAI, Anthropic, OpenRouter (200+ models)
Server Mode — Fastify REST API, SSE real-time events, cron scheduler, Slack/Discord webhooks (API docs)
Zero-Cost Mode — --no-llm runs 21-signal heuristic scoring with zero API calls

Architecture

CLI / GitHub Action / REST API
        |
   PR & Issue Scanner (GitHub GraphQL + REST)
        |
   21-Signal Scoring (TOPSIS) + Intent Classification
        |
   Cascade LLM Pipeline (Heuristic -> Haiku -> Sonnet)
        |
   Diff Analysis + Vision Doc Alignment
        |
   Cross-type Dedup (LanceDB embeddings + LLM verify)
        |
   Semantic Issue-PR Matching + Holistic Re-ranking
        |
   Auto-Actions (close dupes, spam, merge, label)
        |
   Output: table / JSON / markdown / dashboard / SSE

See docs/ARCHITECTURE.md for the full Mermaid diagram.

Configuration

| Variable | Required | Description | |----------|----------|-------------| | GITHUB_TOKEN | Yes | GitHub API access | | GEMINI_API_KEY | For LLM | Gemini scoring + embeddings + vision (free tier) | | OPENAI_API_KEY | For LLM | OpenAI scoring + embeddings | | ANTHROPIC_API_KEY | For LLM | Anthropic scoring (embeddings via Gemini/OpenAI fallback) | | OPENROUTER_API_KEY | For LLM | 200+ models via OpenRouter | | TRELIQ_MODEL | No | Override default model for any provider |

Auto-Actions

# Preview (dry-run, safe)
npx treliq scan -r owner/repo \
  --auto-close-dupes --auto-close-spam \
  --auto-merge --merge-threshold 90 \
  --auto-label-intent

# Execute for real
npx treliq scan -r owner/repo \
  --auto-close-dupes --auto-close-spam --auto-merge \
  --auto-label-intent --confirm

Server Mode

# Start with dashboard
npx treliq server -r owner/repo -p 4747

# With webhooks + cron + notifications
npx treliq server -r owner/repo -p 4747 \
  --webhook-secret $WEBHOOK_SECRET \
  --schedule "0 */6 * * *" \
  --slack-webhook $SLACK_URL

See docs/API.md for all endpoints and SSE events.

GitHub Action

name: Treliq PR Triage
on:
  pull_request:
    types: [opened, synchronize, reopened]

permissions:
  contents: read
  pull-requests: write
  issues: write

jobs:
  triage:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: 20
      - run: npm install -g treliq@latest
      - name: Score PR
        id: score
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
        run: |
          BODY=$(treliq score -r ${{ github.repository }} -n ${{ github.event.pull_request.number }} -f markdown)
          echo "body<<EOF" >> $GITHUB_OUTPUT
          echo "$BODY" >> $GITHUB_OUTPUT
          echo "EOF" >> $GITHUB_OUTPUT
      - uses: actions/github-script@v7
        env:
          SCORE_BODY: ${{ steps.score.outputs.body }}
        with:
          script: |
            await github.rest.issues.createComment({
              owner: context.repo.owner,
              repo: context.repo.repo,
              issue_number: context.payload.pull_request.number,
              body: process.env.SCORE_BODY,
            });

Comparison

| Feature | Treliq | CodeRabbit | Greptile | Copilot | |---------|--------|-----------|---------|---------| | PR scoring & ranking | Yes | No | No | No | | Issue triage | Yes | No | No | No | | Duplicate detection | Yes | No | No | No | | Auto-close/merge/label | Yes | No | No | No | | Heuristic-only mode | Yes | N/A | N/A | N/A | | Multi-provider LLM | 4 providers | OpenAI | Proprietary | GitHub | | Self-hosted server | Yes | SaaS | SaaS | SaaS | | Code review | No | Yes | Yes | Yes |

Contributing

See CONTRIBUTING.md for guidelines.

License

MIT (c) Mahsum Aktas

Docs: CHANGELOG | Signals | API | Architecture

Treliq

Install / Use

README