01
Retrieve
Your question hits pgvector and pulls the most relevant markdown chunks from Sethy's notes — not the open web.
RAG chatbot
Hi, I'm Recall
Ask me anything about Sethy's work, projects, and ideas. Every answer is grounded in his actual notes — and your conversation saves itself, no account needed.
recall · chat
What is Sethy building right now?
A Nuxt 4 RAG prototype on Postgres + pgvector, with an anonymous-friendly chat that remembers each visitor's history...
Which model backs it?
Chat streams through LLM; embeddings use a separate model.
ready · 4 sources cited
How it works
Three steps, fully visible
No black box. Retrieval, reasoning, and persistence are all part of the same surface — you can see what Recall reads before it answers.
02
Reason
The chat model reads those chunks, decides if they're enough, and writes a grounded answer in plain language.
03
Remember
Every conversation is saved under a private visitor ID — reload, come back tomorrow, pick up exactly where you left off.
A real exchange
Answered, not guessed
Every response cites the notes it read. If Recall can't find anything relevant, it says so — instead of inventing an answer.
What tools does Sethy reach for most when shipping a side project?
recall
Three things show up again and again in Sethy's notes: Bun as the runtime and package manager, Postgres with pgvector for storage and retrieval, and Nuxt UI for the surface itself.
The pattern across recent projects is consistent — small, well-typed API layer, a few server-rendered pages, and a chat surface that uses the same primitives. Less is more.
sources stack.md projects.md notes/2026-q2.md
Built with
Boring tools, deliberately
- Nuxt 4
- Postgres + pgvector
- Better Auth
- AI SDK
- Comark
- Drizzle