AI Unified Web Search

I’m putting together a lightweight AI-powered search engine that mimics Perplexity in spirit, but keeps the experience even more compact: one query, one answer box, complete with its supporting citations, plus hand-picked video and image links shown in the same view.

The core of the project is the retrieval pipeline. For every user question, the engine should first pull high-quality blog posts—the primary citation source I want to foreground—then generate a concise textual answer that includes inline citation markers. After that, the same query should surface a short list of relevant videos and images so the result page feels truly “all-in-one.” Although the system will eventually cover many subjects, right now I just need it to handle general-knowledge topics reliably.

I’m looking for an MVP that:
• accepts a natural-language query,
• returns an AI-generated answer with numbered blog citations presented as clean URLs,
• appends a compact carousel (or similar) of video and image links drawn from open APIs, and
• responds in a couple of seconds on average.

You’re free to choose the stack—Python with FastAPI, Node, LangChain, Pinecone, Elastic, or anything comparable—as long as it’s clear, documented, and easy for me to extend later. If you have demo code that already fetches and ranks blog content, we can start from there.

Delivery is a working repo plus a brief walkthrough screencast. Once the basics are stable we can explore polishing the UI, adding more citation types, or scaling the index, but for now the focus is nailing this unified, citation-first search flow.