Top AI Product

There’s something deeply satisfying about a project that looks at a complicated problem and just… refuses to overcomplicate the solution. [Omni](https://github.com/getomnico/omni) is an open-source workplace AI search and chat platform — think Glean, but self-hosted and free — and its boldest design choice is doing *everything* with Postgres. No Elasticsearch cluster to babysit. No standalone vector database to configure. Just PostgreSQL with ParadeDB extensions handling BM25 full-text search, pgvector semantic search, and all your application data in one place.

It [showed up on Hacker News](https://news.ycombinator.com/item?id=47215427) a couple days ago as a Show HN post and immediately caught fire — 103 points, 29 comments, and a spot on [bestofshowhn.com](https://bestofshowhn.com/)’s March rankings. The HN crowd, predictably, zeroed in on the Postgres-only architecture. One commenter put it well: “Postgres as a search backend is one of those decisions that looks wrong on paper but works really well in practice.” Another dev shared that they’ve been running a similar Postgres-based search setup for six years with thousands of users and are only now approaching its limits.

The actual product connects to the tools your team already uses — Google Drive, Gmail, Slack, Confluence, Jira, HubSpot — and lets you ask questions across all of them through a chat interface. It respects existing permissions from source systems, so people only see what they’re supposed to see. On the LLM side, you can plug in Anthropic, OpenAI, Gemini, or run open-weight models through vLLM. The backend is written in Rust and Python, with a SvelteKit frontend, and you can get the whole thing running with a single `docker compose up`.

Now, it’s not without rough edges. The [docs](https://docs.getomni.co) are still filling out (a few 404s spotted in the wild), and if you’re routing queries through OpenAI or Anthropic’s APIs, the “no data leaves your network” promise gets a bit fuzzy. Fair criticism. But for small to mid-size teams who want workplace search without handing their internal data to a SaaS vendor or wrestling with a multi-service infrastructure nightmare, Omni hits a real sweet spot. Sometimes the boring technology choice is the right one.