OpenLens – Community Focused Github App

Overview

Sentinel is a GitHub App built to help open-source maintainers reduce noise and surface meaningful contributions. Instead of focusing solely on code review, Sentinel highlights contributor history, semantic similarity across past work, and low-effort PR indicators. The system is designed to make triage faster for repos that receive high volumes of community activity.

This MVP demonstrates a minimal but functional workflow: loading a repository, embedding all open PRs and issues, identifying related work, and presenting actionable context through a streamlined dashboard.

What It Does (MVP)

• Sync any GitHub repository’s open PRs and issues into a workspace.
• Generate vector embeddings for each item using pgvector.
• Detect related or duplicate PRs and issues through semantic search.
• Surface contributor insights (first-time contributor vs returning).
• Flag weak or low-effort contributions based on metadata signals.
• Allow repositories to be loaded or discarded without persistence overhead.

Why It Matters

Maintainers of active open-source projects face a constant volume of noise: duplicate PRs, malformed issues, abandoned attempts, and low-effort contributions. Reviewing them manually erodes time that could be spent improving the project.

• Spam and duplicate PRs often disguise themselves as genuine contributions.
• Context switching across many historical issues slows down triage.
• Maintainers rarely have enough time to understand contributor patterns.

Sentinel uses embedding-based retrieval, contributor profiling, and simple heuristics to reduce triage time and help maintainers focus on meaningful work.

Technical Overview (MVP)

• Backend: FastAPI for webhooks, sync jobs, and insight APIs.
• Frontend: Next.js + Tailwind for the maintainer dashboard.
• Database: PostgreSQL with pgvector for embeddings and semantic lookup.
• Workers: Background tasks for syncing, vectorizing, and cleaning repo data.
• Auth: GitHub App installation flow with repo-scoped permissions.

Data Flow

1. Maintainer installs the GitHub App and selects a repository.
2. A webhook triggers the backend to enqueue a repo-load job.
3. Workers fetch PRs and issues, generate embeddings, and store them in Postgres.
4. Frontend retrieves insights: similar PRs, related issues, contributor history.
5. A cleanup routine removes data when the repo workspace is closed.

Features in MVP

• End-to-end GitHub App integration
• Repo loader with background sync
• pgvector-based embedding and search
• Similar PR/Issue retrieval
• Contributor metadata and history
• Dashboard for insight consumption
• Cleanup endpoints for temporary workspaces

System Architecture

            Backend: FastAPI
            - Webhook receiver
            - Sync jobs and embedding pipeline
            - REST API for insights

            Database: PostgreSQL + pgvector
            - pr_embeddings
            - issue_embeddings
            - contributor history

            Workers:
            - Sync worker
            - Embedding worker
            - Cleanup worker

            Frontend: Next.js Dashboard
            - PR view
            - similarity panels
            - contributor context
            

Future Roadmap

• AI-powered classification of PRs (spam, duplicate, valuable).
• Contributor trust and reputation scoring.
• Organization-wide community health dashboards.
• Long-term semantic history tracking and knowledge graph generation.
• Deeper GitHub App permissions and automated triage suggestions.

Deployment

• Local: Docker Compose (backend, frontend, Postgres).
• Cloud: Supabase for Postgres, Vercel for frontend.
• Workers deployed via Railway or Fly.io.
• Automatic cleanup of temporary repo data after sessions.

Target Users

• Open-source maintainers
• Community managers
• Developer relations teams
• Engineering orgs with external contributors