recollection

Transform any content into structured learning courses.
Automatically analyze and generate personalized lessons from web articles, PDFs, and videos.

Features

Content Loading:

• 🌐 Web URLs - Articles, blog posts, documentation
• 📄 PDF Documents - Local files and remote URLs
• 🎥 YouTube Videos - Automatic transcript extraction
• 📝 Text Files - Markdown and plain text support

Course Generation:

• 🎯 Smart Lesson Design - LLM analyzes content and creates optimal lesson structure
• 📚 Genre-Aware Strategies - Different approaches for tutorials, documentaries, news, and analysis
• ✅ Learning Objectives - Clear, actionable goals for each lesson
• 🎓 Key Takeaways - Essential insights and skills to master
• 🔗 Source Traceability - Links back to original content
• 📊 Multi-Content Support - Combine multiple sources into unified courses

Quick Start

# Install dependencies
uv sync

# Set up API keys
cp .example.env .env
# Edit .env and add your OPENAI_API_KEY or ANTHROPIC_API_KEY

# Launch the interactive TUI app 🎨
uv run python examples/app.py

The TUI provides a beautiful interactive interface for:

• 📥 Loading content from URLs
• 🎓 Generating courses with AI
• 📖 Viewing and managing courses
• 💾 Saving/loading courses

Usage

from src.content.loader.magic import load
from src.course import generate_course
from src.config import get_config

# Load configuration
config = get_config()
analysis_llm = config.create_llm("summarization")
course_llm = config.create_llm("course_generation")

# Load content
content = load(analysis_llm, "https://realpython.com/python-decorators/")

# Generate course
course = generate_course(llm=course_llm, contents=[content])

# Save course
course.to_json_file("my_course.json")

Documentation

For detailed information, see Course Generation Guide

Configuration

Configure models in config.yaml:

models:
  summarization:
    model_id: gpt-4o-mini
    temperature: 0.3
    max_tokens: 4096
    timeout: 60.0

  course_generation:
    model_id: gpt-4o-mini
    temperature: 0.7
    max_tokens: 4096
    timeout: 90.0

Supported Models:

• Anthropic: claude-3-5-sonnet-20241022, claude-3-5-haiku-20241022
• OpenAI: gpt-4o, gpt-4o-mini, gpt-3.5-turbo

How It Works

1. Load Content - Automatically detect and load from various sources
2. Analyze - LLM classifies genre and extracts key topics
3. Generate - Strategy pattern creates optimal lesson structure
4. Structure - Organize into lessons with objectives and takeaways

Examples

# Interactive TUI app (recommended) 🎨
uv run python examples/app.py

# Command-line examples
uv run python examples/demo.py https://example.com/article
uv run python -m examples.test_course_generation
uv run python -m examples.generate_course_from_url

Project Structure

src/
├── content/          # Content loading and analysis
│   ├── loader/       # Magic loader with auto-detection
│   └── analysis/     # Genre classification and topics
├── course/           # Course generation system
│   ├── generator.py  # Main course generation
│   ├── merger.py     # Multi-content merging
│   └── strategies/   # Genre-specific strategies
└── config.py         # Configuration management

examples/
├── app.py            # 🎨 Interactive TUI application
├── demo.py           # Content loading demo
├── test_course_generation.py
└── generate_course_from_url.py

Requirements

• Python 3.13+
• API key for Anthropic or OpenAI

What's Next

• 🎯 Activity Generation - Quizzes, challenges, and games
• 📊 Progress Tracking - Monitor learner progress
• 🔄 Daily Challenges - Automated practice generation
• 🤖 Adaptive Learning - Adjust difficulty based on performance

Contributing

Contributions welcome! Please feel free to submit issues and pull requests.

License

MIT License - see LICENSE for details