Mimir

Drink from the well of knowledge in the AI era.

AI intelligence collection and organization system. Automatically collects, processes, and archives information from RSS, YouTube, and Twitter.

Features

• Multi-source collection: RSS feeds, YouTube channels
• Processing pipeline: content cleaning, quality assessment, deduplication, classification, knowledge extraction
• Optional LLM enhancement: summarization, translation, categorization
• Notion integration for archiving
• Cost control with budget limits

Quick Start

Install

git clone https://github.com/rpings/Mimir.git
cd Mimir
pip install -r requirements.txt

Configure

1. Set environment variables:

   export NOTION_TOKEN="your_token"
   export NOTION_DATABASE_ID="your_database_id"
   # Optional: LLM features
   export OPENAI_API_KEY="sk-..."

2. Edit configuration files:

   • configs/sources/rss.yaml - RSS feeds
   • configs/sources/rules.yaml - Classification rules
   • configs/config.yml - Main config

Run

python main.py          # Sync mode
python main_async.py     # Async mode

Configuration

See configs/config.yml for full configuration options.

Key settings:

• Processing pipeline processors (cleaning, quality, verification, etc.)
• LLM settings (optional, requires API key)
• Notion database configuration

LLM Setup (Optional)

To enable LLM features:

1. Set OPENAI_API_KEY environment variable
2. Enable in configs/config.yml:

   llm:
     enabled: true
     provider: openai
     model: gpt-4o-mini

For local models, set base_url in config or LLM_BASE_URL environment variable.

Development

# Setup
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Test
pytest tests/

# Code quality
black src/ tests/
ruff check src/ tests/

License

MIT License - see LICENSE file.