What I Built

Susegad Desk is a privacy-first guest check-in tool for small homestays and guesthouses.

When a guest arrives, the owner can take a photo of an ID using a phone on the homestay’s local Wi-Fi. The local app runs the image through Tesseract OCR, then uses Phi-3 Mini through Ollama to turn the recognized text into structured guest details. The owner reviews and edits those suggestions before adding them to a register.

The goal is to replace repetitive manual transcription with a quicker workflow that can run on the homestay’s own computer. It’s designed for small operators who want to avoid per-request AI costs and keep the ID-processing workflow local.

Code

satwikpagi5-wq / susegad_desk3

Privacy-first guest check-in demo for local homestays. Includes a Vercel static demo and a local OCR app.

Susegad Desk

A privacy-first guest check-in prototype for independent homestays. Susegad Desk turns an ID photo into editable guest details using OCR and a local language model, then requires human review before adding anything to the day register.

Prototype and privacy notice: The Vercel version is a fictional-data demo only. It does not accept ID images and does not connect to the OCR service. The full OCR workflow runs on a reception computer on a trusted local network. This is not a legal or compliance product.

Hosted demo

The static Vercel build is served from web/. It demonstrates the review form, session-only guest register, and CSV export with fictional values. It intentionally has no upload field: government ID images must not be sent to this public demo.

Local app

The full local app is served by FastAPI from server.py, with its UI in static/. Images are processed…

View on GitHub

How I Built It

The project has two parts:

A local FastAPI backend that receives an ID image, runs Tesseract OCR, and sends the extracted text to Ollama on the same computer.

A responsive web interface for reviewing the extracted fields and maintaining a session-only guest register. The stack is Python and FastAPI for the backend, Tesseract for OCR, Ollama for local model inference, and HTML, CSS, and JavaScript for the interface. The hosted demo is a separate static site, so it can show the workflow without handling identity documents. The extraction is treated as a suggestion, not a final record: the owner must review the fields before adding them.

Why Does Open Innovation Matter?

Local inference is central to this product. A cloud OCR or AI API would require internet access during processing, introduce recurring usage costs, and send sensitive identity data to a third-party service.
With Phi-3 Mini running locally through Ollama, the homestay can process documents on its own computer without a per-request model bill. Once the model and dependencies are installed, the check-in workflow can continue during an internet outage, as long as the computer and local network remain available.

That makes a one-time installation model more plausible for small homestays than a monthly subscription. It also lets the operator keep the document-processing workflow on hardware they control. This prototype is not a compliance product, and every extracted field still needs to be checked against the source document.

DEV Community

Susegad Desk: Private, Guest Check-In for Small Homestays