Story

There's LLM Council and similar tools, but they use predefined model lineups. This one is different in a few ways that mattered to me:

*Bring your own models.* Mix Ollama (local), OpenAI, Anthropic, Groq, Google — or any OpenAI-compatible endpoint — in whatever combination you want. A council of DeepSeek-R1 + llama2-uncensored + mistral-nemo is a very different deliberation than GPT-4o + Claude + Gemini.

*Zero server, zero account, zero storage.* The app is purely static. API calls go directly from your browser to providers. Nothing touches a backend. No tokens, no sessions, no analytics. Your API keys never leave your machine.

*Runs on your own hardware.* If you have Ollama, you can run an entire council locally for free. I use a 5-member all-Ollama setup on an RTX 2070 (8GB VRAM) — sequential requests, slow, but completely private.

The deliberation process is 3 stages: 1. All members answer independently 2. Each member critiques anonymized responses from the others 3. A designated Chairman synthesizes a final verdict

A few things I found genuinely interesting: - Reasoning models (DeepSeek-R1, QwQ) emit <think> blocks mid-stream. Stripping these while showing a " Thinking…" indicator keeps the UX clean without losing answer quality. - The Contrarian persona on an uncensored model produces meaningfully different critiques than a safety-tuned model playing the same role. - Peer review across models catches blind spots that a single model arguing with itself won't surface.

GitHub: https://github.com/prijak/Ai-council.git