ConvertIt

AI-Powered Document Simplification Engine

ConvertIt transforms complex technical documents (URLs, PDFs, text files) into easy-to-understand educational materials tailored for specific audiences. It uses local or remote LLMs, RAG for context management, and AI-powered image generation.

✨ Key Features

Feature	Description
Multi-Source Input	URLs, PDF files, TXT files via drag-and-drop
5 Teaching Personas	5th Grader, High School, Undergrad, Professional, Executive
RAG Knowledge Base	Index reference documents to enhance output quality
Hybrid LLM Routing	Local LLM for simple tasks, remote for quality-critical
Semantic Chunking	Heading-aware splitting with context carryover
AI Illustrations	Generate relevant images via ComfyUI or DALL-E
Real-time Logs	Live debug console for monitoring

🏗 Architecture

┌─────────────┐     ┌──────────────┐     ┌─────────────┐
│   Django    │────▶│  LangGraph   │────▶│   Output    │
│   Web UI    │     │  Workflow    │     │  (MD/PDF)   │
└─────────────┘     └──────────────┘     └─────────────┘
       │                   │
       ▼                   ▼
┌─────────────┐     ┌──────────────┐
│   Settings  │     │  LLM Engine  │
│   & Config  │     │ (Local/API)  │
└─────────────┘     └──────────────┘

Tech Stack:

• Backend: Django + LangGraph workflow
• LLM: LiteLLM + Ollama (local) or OpenAI/Anthropic (remote)
• RAG: ChromaDB + LlamaIndex
• Frontend: Tailwind CSS + Lucide Icons
• Vision: ComfyUI / DALL-E

🚀 Quick Start

Prerequisites

• Python 3.10+
• Ollama (for local LLM)
• Git

Installation

# Clone
git clone https://github.com/williamQ96/convertit.git
cd convertit

# Virtual environment
python -m venv .venv
.\.venv\Scripts\Activate.ps1  # Windows
source .venv/bin/activate      # Mac/Linux

# Install dependencies
pip install -r requirements.txt

# Or manual install:
pip install django litellm ollama chromadb langgraph instructor jinja2 pymupdf python-dotenv requests llama-index llama-parse

Configuration

cp .env.example .env

Edit .env:

LLM_PROVIDER=local
OLLAMA_BASE_URL=http://localhost:11434
OPENAI_API_KEY=sk-...          # For remote LLM
CHROMA_DB_PATH=./chroma_db
RAG_FOLDER=./document/convertit/database

Run

# Start Ollama (if using local)
ollama serve

# Run Django server
python manage.py runserver

Open http://localhost:8000

📁 Project Structure

convertit/
├── agents/            # LangGraph workflow & prompts
│   ├── workflow.py    # Clean → Glossary → Rewrite → Critic → Images
│   └── prompts.py     # Persona-specific prompts
├── core/              # Core services
│   ├── engine.py      # LLM engine with task-based routing
│   ├── ingestion.py   # URL/PDF parsing (Firecrawl + PyMuPDF)
│   ├── indexer.py     # RAG document indexer
│   └── vision.py      # Image generation
├── converter/         # Django app
│   ├── views.py       # API endpoints
│   └── urls.py        # Route configuration
├── database/          # ChromaDB vector store
├── web_ui/            # Django project settings
├── templates/         # HTML templates
└── static/            # CSS, JS, generated images

🔧 Configuration Options

Teaching Personas

Persona	Description
🎒 5th Grader	Simple language, fun analogies
📚 High School	Clear explanations with terminology
🎓 Undergraduate	Academic rigor, proper context
💼 Professional	Polished technical writing
📊 Executive	TL;DR with key takeaways

Output Options (Toggleable)

• 📝 Include code examples
• 📊 Add summary table
• 💡 Highlight key takeaways
• 📖 Include glossary section

Vision Strategy

• AI Gen: Generate new images with AI
• Hybrid: Original + AI enhancements
• Original: Keep original images
• Text Only: No images

🧠 LLM Pipeline

Raw Content → Clean → Glossary → Rewrite → Critic → [Loop] → Images → Output

Optimizations:

• Semantic Chunking: 6K char threshold, heading-aware splitting
• Context Carryover: Summary passed between chunks
• Hybrid Routing: Local LLM for clean/glossary, remote for rewrite/critic

📡 API Endpoints

Endpoint	Method	Description
/	GET	Main UI
/convert/	POST	Start conversion
/api/settings/	POST	Save settings
/api/index/	POST	Index RAG documents
/logs/	GET	Stream logs

📝 Development Log

See development_log.md for detailed progress tracking.

Recent Updates (Feb 2026):

• Phase 11: LLM Pipeline Optimization
• Phase 10: RAG Document Indexing
• Phase 9: UI Configuration Refinement
• Django migration from Chainlit

📄 License

MIT