Building a Unified AI Playground - My Weekend Project with Open WebUI and LiteLLM

Published:

Over the weekend, I setup a frontend and a proxy backend for multiple AI LLM models.

I used two awesome open source projects. Shoutout to the developers:

  1. Open WebUI: https://github.com/open-webui/open-webui - For frontend

  2. LiteLLM: https://github.com/BerriAI/litellm - For the proxy backend

This enables me to:

  1. Hit multiple models (GPT 4o, Claude, etc), from a single frontend instead of jumping multiple sites. Even compare the responses from multiple models for a single prompt.

  2. A single dashboard to analyse the usage and costs across all the models.

  3. Setup a single budget across all models.

I self-hosted this setup on a Hetzner VPS.

Screenshots:

Next Steps

Self host some open source LLMs like Llama.