CoReasoning Lab

An AI-native learning platform where students do not just get answers from AI, they learn to reason with it.

CoReasoning Lab trains three core skills:

• Framing: turn ambiguous problems into precise, solvable tasks.
• Judging: critique AI output and spot gaps, risks, and mistakes.
• Steering: guide AI iteratively toward better results.

Why This Project Exists

Most AI-in-education products optimize for speed-to-answer. CoReasoning Lab optimizes for quality-of-thinking.

Learners interact with intentionally imperfect AI outputs, then improve them through structured reasoning loops. The platform measures progress across separate cognitive skills instead of a single score.

Learning Flow

1. Framing phase: student refines a raw problem with assumptions, constraints, and clarifications.
2. AI generates an initial solution (with realistic issues to inspect).
3. Judge + Steer cycles: student evaluates quality, then sends targeted corrections.
4. Platform grades Framing, Judging, and Steering independently with feedback.

Feature Snapshot

• Challenge modes: practice and assessment.
• Response modes: multiple-choice or open-ended per phase.
• LLM-backed generation and evaluation with fallback behavior.
• Multi-language content and UI assets (en, he, fr, de, es).
• Role-based workflows: student, instructor, admin.
• Course/challenge management, analytics, and PDF reporting.
• Bulk YAML import for institutions, users, courses, and challenges.

Tech Stack

• Runtime: Node.js 20+
• Backend: Express, Knex, Zod, Passport
• DB: SQLite (dev/test) and PostgreSQL (production)
• AI: OpenAI/Groq integrations with pluggable provider config
• Testing: Jest, Supertest, Playwright
• Deployment: Docker and Render blueprint

Architecture At A Glance

Static Client (HTML/CSS/JS)
        |
        v
Express API Routes (/api/v1/*)
        |
        v
Service Layer (business logic)
        |
  +-----+------------------+
  |                        |
  v                        v
LLM Service + Prompt      Database (Knex)
Engine (YAML prompts)     SQLite / PostgreSQL

Quick Start (Repo Root)

Set environment variables in code/.env.all (copy from .env.example and fill required keys).

npm install
npm run build
npm run db:migrate
npm run db:seed
npm run dev

Then open:

• App: http://localhost:3000
• Health check: http://localhost:3000/api/health

Useful Commands

npm run test:unit
npm run test:integration
npm run test:e2e
npm run test:all
npm run lint
npm run docker:dev

Project Map

code/
  client/        static web UI
  server/        API routes, services, middleware, DB, auth
  prompts/       LLM prompt templates
  content/       i18n source data
  tests/         unit/integration/e2e tests
docs/            concept of operations, spec, audits
scenarios/       course scenario content

Deep Docs

• Product concept: docs/CONOPS.md
• System spec: docs/spec/SPEC.md
• Full technical README: code/README.md
• Developer guide: code/docs/developer-guide.md

Notes

• The hero image in this README was generated with Google Gemini Image API.