CSRadrX

CSRadrX is an AI-powered computer science intelligence platform that monitors research papers, engineering blogs, open-source repositories, and developer communities to detect meaningful ecosystem updates.

The system processes data using an event-driven pipeline and publishes summarized insights via a dashboard and automated channels.

Live: https://csradrx.live

Current MVP flow:

• fetch GitHub repositories
• filter high-signal repos
• enqueue jobs in Redis
• process jobs in a worker
• store Event, EventAnalysis, and EventScore
• expose results through the API

Current Architecture

• apps/fetcher -> fetches GitHub data and enqueues jobs
• apps/queue -> consumes jobs, writes DB records, runs analysis and scoring
• apps/api -> serves stored data
• packages/db -> Prisma schema and database config
• packages/core -> shared analysis and scoring logic

Prerequisites

Install:

• Node.js
• pnpm
• Redis
• PostgreSQL-compatible database

Environment Setup

Create or update packages/db/.env with:

DATABASE_URL="your_database_url"
REDIS_URL="redis://127.0.0.1:6379"
GITHUB_TOKEN="your_github_token"

You can use packages/db/.env.example as a template.

Install Dependencies

From the repo root:

pnpm install

Start Redis

If installed via Homebrew:

brew services start redis
redis-cli ping

Expected:

PONG

Run The Project

Start the worker first:

cd apps/queue
npx tsx src/index.ts

In a second terminal, run the fetcher:

cd apps/fetcher
npx tsx src/github.ts

Expected fetcher output:

Queued: owner/repo

Expected worker output:

Queue worker ready
Processed job owner/repo: inserted owner/repo

In a third terminal, start the API:

cd apps/api
node --import tsx src/index.ts

API Endpoints

Available endpoints:

• http://localhost:3000/events
• http://localhost:3000/trending
• http://localhost:3000/research

Verification

Check that events, analysis, and scores exist:

SELECT
  e.title,
  ea.summary,
  es.score
FROM "Event" e
LEFT JOIN "EventAnalysis" ea ON ea."eventId" = e.id
LEFT JOIN "EventScore" es ON es."eventId" = e.id
ORDER BY e."createdAt" DESC
LIMIT 20;

Notes

• apps/fetcher does not write to the DB directly
• duplicate events are skipped before enqueue using batched URL checks
• /trending currently returns GitHub events from the last 48 hours ordered by score