For Career Switchers with Resume
Login / Sign up Page
Offline & Explained with animation
AI Tutor
Dashboard
Roadmap
Name Of the Project

DreamCatcher — Adaptive Mastery Engine

100% Offline, Self-Hosted Career Mastery Platform

DreamCatcher uses local AI (Ollama) to diagnose skills, create adaptive roadmaps, and teach from your own textbooks — with zero cloud dependencies.

Architecture

┌─────────────────────────────────────────────────────────────┐
│                    DREAMCATCHER                              │
├─────────────────────────────────────────────────────────────┤
│  FRONTEND (Next.js 14)          BACKEND (FastAPI)           │
│  ├─ Dashboard                   ├─ /diagnostic               │
│  ├─ Diagnostic Modal            ├─ /generate-roadmap         │
│  ├─ Roadmap Timeline            ├─ /tutor (RAG)              │
│  ├─ AI Tutor Chat               ├─ /assessment/*             │
│  └─ Assessment UI               ├─ /stats                    │
│                                 ├─ /upload                   │
│                                 └─ /health                   │
├─────────────────────────────────────────────────────────────┤
│  MULTI-AGENT ORCHESTRATION                                   │
│  ├─ 🤖 Strategist (Phi-3)    → Diagnostics & Roadmaps       │
│  ├─ 📚 Scholar (TinyLlama)   → RAG-powered Tutoring         │
│  └─ 📝 Proctor (Phi-3)       → Assessments & Retention      │
├─────────────────────────────────────────────────────────────┤
│  LOCAL INFRASTRUCTURE                                        │
│  ├─ Ollama (localhost:11434)  → Phi-3 + TinyLlama           │
│  ├─ FAISS (CPU)               → Vector search               │
│  ├─ SQLite                    → User data & state           │
│  └─ Sentence-Transformers     → Local embeddings            │
└─────────────────────────────────────────────────────────────┘

Project Structure

dreamcatcher/
├── backend/
│   ├── main.py              # FastAPI routes & API
│   ├── agents.py            # 3-agent orchestration (Strategist, Scholar, Proctor)
│   ├── rag_engine.py        # FAISS + PyMuPDF + Sentence Transformers
│   ├── database.py          # SQLite models & helpers
│   ├── requirements.txt     # Python dependencies
│   └── uploads/             # PDF textbooks storage
│
├── frontend/
│   ├── app/
│   │   ├── layout.tsx       # Root layout
│   │   ├── page.tsx         # Main entry
│   │   └── globals.css      # Tailwind + custom styles
│   ├── components/
│   │   ├── Dashboard.tsx    # Mastery gauge, trajectory chart, roadmap
│   │   ├── DiagnosticModal.tsx  # 15-question adaptive test
│   │   └── AssessmentModal.tsx  # Weekly 10-question assessment
│   ├── package.json
│   ├── tailwind.config.js
│   └── tsconfig.json

Quick Start (VS Code)

Prerequisites

Python 3.10+
Node.js 18+
Ollama installed locally → ollama.com

Step 1: Pull LLMs via Ollama

Ensure your Ollama app is open and running in the background. Open a terminal and pull the required models:

ollama pull phi3:latest
ollama pull tinyllama:latest

Step 2: Open Project in VS Code

Open VS Code.
Go to File > Open Folder and select the dreamcatcher folder.
Open a split terminal (Ctrl + ~ or Terminal > New Terminal, then click the split icon +).

Step 3: Start the Backend (Terminal 1)

In the first terminal tab, navigate to the backend, activate the environment, and start the API:

cd backend
.\venv\Scripts\Activate
python -m uvicorn main:app --host 0.0.0.0 --port 8000 --reload

The backend will run on http://localhost:8000.

Step 4: Start the Frontend (Terminal 2)

In the second terminal tab, navigate to the frontend and start the Next.js app:

cd frontend
npm run dev

The frontend will run on http://localhost:3000.

Step 5: Open the App

Navigate to http://localhost:3000 in your browser. Since both servers are running in dev mode (--reload and run dev), any code changes you make in VS Code will update live in the browser!

📡 API Endpoints

Endpoint	Method	Description
`/`	GET	API status
`/health`	GET	Ollama + RAG health check
`/diagnostic`	POST	Generate 15-question MCQ test
`/diagnostic/submit`	POST	Submit answers, get analysis
`/generate-roadmap`	POST	Create 12-week adaptive roadmap
`/roadmap`	GET	Retrieve current roadmap
`/tutor`	POST	Ask the Scholar (RAG + TinyLlama)
`/tutor/stream`	POST	Stream Scholar explanations in chunks
`/upload`	POST	Upload PDF for RAG ingestion
`/rag/status`	GET	Vector store statistics
`/rag/reindex`	POST	Rebuild FAISS index
`/assessment/generate`	POST	Generate weekly 10-Q assessment
`/assessment/submit`	POST	Submit & update mastery metrics
`/stats`	GET	Dashboard data (success probability, trajectory)
`/stats/detailed`	GET	Extended dashboard stats
`/chat/history`	GET	Tutor chat history

New Features

1. Industrial Learning Roadmaps

The Adaptive Engine generates senior-level, enterprise-ready roadmaps. Milestones focus heavily on real-world engineering concepts like Kubernetes, CI/CD (GitHub Actions, Terraform), Microservices (gRPC, API Gateways), and Distributed Systems (Redis, Kafka).

2. Native Visual AI Tutor

Ask the AI Scholar Tutor to explain complex concepts visually using the @animation command! By typing a query like @animation stack What is a stack?, the Tutor will stream a detailed text explanation alongside a beautiful, purely CSS-driven React animation. Currently supported animations: stack, loop, and array.