Local Chat
Private, on-device chat with your local Ollama models. The browser connects directly to localhost:11434 — no prompts or images leave this machine.
Local proxy keeps your API key on this machine. Direct modes require a free key from Tavily or Brave Search and work best with a tool-calling model like Devstral.
Frames are processed locally by gemma3:4b at the endpoint above.
No messages yet.
Choose a model and start a conversation. Your history is stored locally in this browser.
Not connecting? Setup notes
This page only works in a browser running on the same machine as Ollama, with Ollama running (ollama serve).
- Local preview (
http://localhost:4000) works with zero config — Ollama allowslocalhostorigins by default. - Deployed site (
https://mohankku.github.io) is blocked by Ollama's default CORS policy. Allow it once, then restart Ollama:(For a persistent setting across reboots, addlaunchctl setenv OLLAMA_ORIGINS "https://mohankku.github.io"brew services restart ollamaOLLAMA_ORIGINSunderEnvironmentVariablesin~/Library/LaunchAgents/homebrew.mxcl.ollama.plistinstead oflaunchctl setenv.) - Ollama only listens on
127.0.0.1, so even with the origin allowed, only browsers on this laptop can reach it. Restarting the service unloads models; they reload on first use.
Price watches
The proxy re-checks each watch about every 6 hours. Prices come from search snippets, so treat them as approximate.