Hey PH! I'm launching Barongsai — an open-source Python framework for building AI search agents. The problem: Teams want Grok/Perplexity-style AI search but can't send proprietary data to third-party services. Enterprise compliance, data residency, cost control — there are real reasons to self-host. What Barongsai does: - Searches the web (DuckDuckGo or Brave), fetches content, synthesizes answers with clickable citations - Plug in any LLM — OpenAI, Ollama, vLLM, or run HuggingFace models locally with 4-bit quantization (zero token costs) - Ships with Open WebUI integration — docker compose up gives you a full chat interface - OpenAI-compatible API (/v1/chat/completions) — drop into any existing workflow - Enterprise middleware built-in: circuit breakers, rate limiting, structured logging, auth What it's NOT: A hosted service. It's a framework + reference app. You own the stack, you own the data. Built with Python 3.11+, FastAPI, Pydantic v2, fully async. MIT licensed. I open-sourced barongs.ai so teams and enterprises can self-host AI search without vendor lock-in.

barongs.ai

host Grok/Perplexity-style AI search, any LLM, your infra

host Grok/Perplexity-style AI search, any LLM, your infra