Skip to content
Sangam Gupta
All work

Quanta (Sangam AI)

A ChatGPT-style assistant running entirely on a server I set up and maintain.

Status
Demo offline right now
Stack
Qwen2.5, Ollama, OpenWebUI, Nginx, DigitalOcean
Links
Open Quanta

How it fits together

  1. Browser
  2. ai.sangamgupta.in (HTTPS)
  3. Nginx reverse proxy
  4. OpenWebUI
  5. Ollama
  6. Qwen2.5 model

Self-hosted LLM chat platform running Qwen2.5 through Ollama and OpenWebUI, deployed on DigitalOcean behind Nginx with HTTPS and custom system prompts.

The problem

Hosted chat assistants send every prompt to someone else's servers and limit how you can shape the model. I wanted an assistant I fully control, from the model weights to the domain it answers on.

How I built it

  1. Run the open-weight Qwen2.5 model with Ollama on a DigitalOcean droplet.
  2. Put OpenWebUI in front for accounts, chat history and model switching.
  3. Terminate HTTPS at Nginx as a reverse proxy on ai.sangamgupta.in, set up with AI assistance.
  4. Shape its behaviour with domain-specific system prompts and knowledge bases for AI and finance questions.

Where it is now

Live at ai.sangamgupta.in, maintained and updated by me end to end.

Ask about Quanta (Sangam AI)

Answers come from this case study.