/ansari-backend

Ansari is a helper for you to become a better Muslim

Primary LanguagePython

Ansari

Try Ansari now at ansari.chat!

Ansari is an experimental open source project that explores the application of large language models in helping Muslims improve their practice of Islam and non-Muslims develop an accurate understanding of the teachings of Islam.

It is not always correct and can get things wrong. The list below includes some of the issues we’ve seen in working with Ansari.

How can you help?

  • Try Ansari out and let us know your experiences – mail us at feedback@ansari.chat.
  • Help us implement the feature roadmap below.

What can Ansari do?

A complete list of what Ansari can do can be found here.

What is logged with Ansari?

  • Only the conversations are logged. Nothing incidental about your identity is currently retained: no ip address, no accounts etc. If you share personal details with Ansari as part of a conversation e.g. “my name is Mohammed Ali” that will be logged.
  • If you shared something that shouldn’t have been shared, drop us an e-mail at feedback@ansari.chat and we can delete any logs.

The Roadmap

This roadmap is preliminary, but it gives you an idea of where Ansari is heading. Please contact us at feedback@ansari.chat if you’d like to help with these.

  • Add feedback buttons to the UI (thumbs up, thumbs down, explanation)
  • Add “share” button to the UI that captures a conversation and gives you a URL for it to share with friends.
  • Add “share Ansari” web site.
  • Improve logging of chats – move away from PromptLayer.
  • Add Hadith Search.
  • Improve source citation.
  • Add prayer times.
  • Add login support
  • Add personalization – remembers who you are, what scholars you turn to etc.
  • Separate frontend from backend in preparation for mobile app.
  • Replace Gradio interface with Flutter.
  • Ship Android and iOS versions of Ansari.
  • Add notifications (e.g. around prayer times).
  • Add more sources including:
    • Videos by prominent scholars (transcribed into English)
    • Islamic question and answer web sites.
  • Turn into a platform so different Islamic organizations can customize Ansari to their needs.

Acknowledgements

  • Amin Ahmad: general advice on LLMs, integration with vector databases.
  • Hossam Hassan: vector database backend for Qur’an search.
  • Saifeldeen Hadid: testing and identifying issues.