A repository for performing analysis on scripture with LLMs. This is a work-in-progress and code will be improved over time.
- .NET 8
- Docker Desktop
- An NVIDIA GPU with minimum 12GB RAM
- Alternatively, you can use a smaller model than R1:14b and then you need less V/RAM
- Run
docker-compose upin the root folder of the project. This will spin up the ollama container with DeepSeek-R1:14b for the console app to call via API. - Set up the books you want to analyze in
output/config.jsonas a JSON array- Alternatively, you can disable the code that checks which books are enabled and it will run against the entire dataset
- Run the Console app to begin analysis
- Update prompt engineering and return fields as necessary to create your own research plan.
- JSON version of the KJV Bible courtesy of the MIT licensed repo at kenyonbowers/BibleJSON
- DeepSeek-R1, MS phi4 used under MIT license as detailed at ollama here and here
- This repository uses the Crossway ESV API to compare that translation with the KJV. A valid application registration there and API token is required to use this functionality, as the ESV is not an openly licensed text. This repository also has a request limiting feature in to comply with Crossway's terms of service.