zDocs.ai

https://zdocs.ai is live

With my SaaS, you can now

• login with Google, or username/password
• create a set of documents, upload files
• files will be OCR'd
• you can then query your files content
• in the background I prompt Gemini LLM and feed the answers back to you !

Check it out !

Here I went to start a new MVP. Allowing myself 1 month to build this during weekends and nights.

Along the years, I have tailored my perfect tech stack for this. Here we go:

• Django/Python full stack web framework. No frontend framework, please...
• Docker
• HTMX for the cool dynamic stuff... what you don't know yet about htmx ?
• Langchain for the RAG stuff
• A nice bootstrap template for the UI
• Google Cloud Run
• Google Documents AI
• Google Vertex AI
• OpenAI

In my day to day job, as I am playing with ChatGTP and Gemini, I often wanted to be able to upload a set of documents (not just 1 document) and be able to ask the LLM about something in that document... and perhaps include context from another set of documents also.

Strangly enough, I could not find a simple tool out there to do that. I ended up getting into coding a full AI pipeline, including:

• any document type data extraction (using GCP Document AI)
• RAG (Retrieval Augmented Generation) flow

This worked pretty well... and seemed like an opportunity to build a SaaS around it !