0

The use case is to deploy self-hosted Retrieval Augmented Generation solution for question-answering based on the documents available in the company. There are many such paid solutions in the cloud, but I'm looking for open source alternatives which can be self-hosted.

Requirements:

  1. Answer should be generated based on the up-to-date documents.
  2. Integration with different data sources (e.g., GitHub, Slack, etc.)
  3. Answers should be generated using Large Language Models.
  4. There should be a possibility to use custom LLMs (e.g., using ollama, llama.cpp, etc.)
  5. There should be a support for users which have different access level to the documents.
  6. Document retriever should be customizable to some degree (e.g., different embedding models, time decay for documents, etc.)

The closest solution I found which meets above criteria is Danswer, but still, they don't have access management.

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.