Booky

• 

  Booky app

• 

  Booky talk

Inspiration

I wanted to create a app that lets users to talk with greatest minds and experts in life, business and science.

What it does

Booky is a book based LLM chat that is connected to simple pgvector RAG. Embeddings are created either from a single book or body of works (like collection of blog posts). It allows users to talk with famous authors, get tips and apply the instructions to their daily life. It also learns about users preferences and goals. Users can chat with text or audio on the go. And they can create podcasts with summarised learnings from discussions to listen later.

How we built it

We build the app using bolt.new with supabase, edge functions, elevenlabs and published to netlify. I am solo creator of the project.

Challenges we ran into

I initially started building with buildship (and n8n) for RAG but quickly realised we can just accomplish reasonable level with Supabase edge functions. And this was really smooth with bolt.new

Accomplishments that we're proud of

Im really proud pulling this off 100% inside bolt.new

What we learned

Code editor is pretty ace too. Ive mostly used bolt to just create project starts, but can actually get much further with the code editor too. Supabase edge functions are quite ideal pair with bolt.new setup.

What's next for Booky

We will start approaching publishers about LLM version book rights and optimising our RAG model and chat for larger body of works. We want to focus on niche professional authors that are applicable for work and can give better guidance than regular chatgpt etc generalist LLMs. We'll extend podcasting capabilities and introduce multi author roundtable chats. Any success in the hackathon will be crucial sign for the project.

Built With

• elevenlabs
• netlify
• supabase
• tailwind
• vite