Omnivore summarizer

LLM-powered summarizer for Omnivore App articles using Ollama and LangChain.

For now works for subscriptions only.

Installation

(Optional) Install uv package manager

Then run:

make init # or copy the pip commands from the Makefile if no uv

After that:

1. Fill in .env-example and rename it to .env.
   • OMNIVORE_API_KEY - your Omnivore API key
   • OLLAMA_MODEL_ID - your Ollama model ID
2. Create summarized and read later labels in your Omnivore App (will be automated in the future)
3. (Optional) Create saved searches in your Omnivore App for the labels you created
4. Download LLaMa3 model:
   • Ollama LlaMa3
   • also works with other models, but tokenization estimation might be off (for now)
5. run ollama serve to start the Ollama server

Usage

make run # or python main.py

Flow description

Work in progress!

1. Fetches all subscriptions from the last 24 hours
2. Filters out artcles that are larger then context of the model (8192)
3. Summarizes the articles with later options to read later or archive
4. TBA

Tech Stack

• Omnivore App (link)
• omnivoreQL - graphQL API for Omnivore APP (link)
• Ollama (link)
• LangChain (link)