Monitor Bluesky posts in real time, filter them by custom keywords, and analyze matching conversations with Gemini . Built for the Vibe Code-a-thon , where it won Best App Using External Knowledge.

Why build this?

I wanted to know what people were actually talking about on Bluesky — not just scroll the feed, but follow topics I care about and understand the sentiment around them as posts arrive.

Bluflare ingests the Bluesky Firehose through Jetstream , stores matching posts, and runs LLM analysis to summarize themes and sentiment. The UI is available in English and French, with light and dark mode.

Hackathon

This project started at the Vibe Code-a-thon and was submitted on Devpost . It won the Best App Using External Knowledge award — pulling live Bluesky data through Jetstream and enriching it with Gemini analysis.

UI

You could define interests/keywords, watch the firehose stream in near real time, and review LLM-generated summaries and sentiment for matching posts.

Architecture

The backend is a Rust service using Axum , SQLx with SQLite, and a Jetstream websocket client. The frontend is a React SPA built with shadcn/ui , TanStack Query , and TanStack Router .

Configuration is handled through Figment with BLUFLARE__-prefixed environment variables, and optional cookie-based authentication for admin endpoints.