my-AI-chat

A small, fully-customizable UI for chatting with your local Ollama models. Plain HTML/JS frontend + a thin FastAPI proxy. No build step, no cloud.

Run

./run.sh
# then open http://localhost:8010

(First run creates a venv and installs FastAPI/uvicorn/httpx. Needs Ollama running on localhost:11434.)

What's customizable

• System prompt + params — system prompt, temperature, top-p, max tokens, editable live per chat (⚙ Settings).
• Personas / presets — name a bundle of model + prompt + params, switch with one click. Stored in data/presets.json (survives restarts; edit by hand too).
• Theme — 4 base themes (Midnight / Paper / Terminal / Sunset) plus a color picker for every token, font size, and corner roundness. Saved in the browser.
• Multi-model compare — ⊞ Compare sends one prompt to several models at once, side by side.

Layout

backend/app.py      FastAPI: /api/models, /api/chat (streaming), /api/presets
frontend/index.html the whole UI (theme tokens live in :root CSS vars)
data/presets.json   saved personas

Tweaking it yourself

• Colors/spacing: all visual values are CSS variables in :root at the top of frontend/index.html — change defaults there, or add a theme to the THEMES object in the script.
• New backend / model source: point OLLAMA in backend/app.py elsewhere, or swap the /api/chat body to an OpenAI-compatible endpoint.
• Default port: edit run.sh (--port 8010).