How I digitize my handwritten Notes: RunnerH + Purpose-Built API

This is a submission for the Runner H "AI Agent Prompting" Challenge

What I Built

I'm currently a student at the University of Ilorin studying computer science. Due to several constraints, most prominently power supply, I use a mix of Notion, plus Google Docs, plus handwritten notes. Originally, I was the only one who used my handwritten notes, but in recent times, particularly in the current (second semester), many of my colleagues have shown interest in them too. This has led to the need for digitizing the notes.

Typically, I would do the following:

1. Take Pictures

   • I take pictures of my handwritten notes with my phone.
   • I follow the actual reading order when taking the pictures:
     • For example, if the note is 5 pages (p1 to p5), the reading order could be: p1, p2, p5, p3, p4.
     • This is because I ask questions in the front pages and write answers at the back.
   • Sometimes a middle page contains both notes and answers, so I take the same page twice and crop each accordingly.

2. Edit the Images

   • I crop each duplicate to show only the part needed based on the order.
   • This way, the picture order already matches the intended PDF order.
   • The original EXIF DateTimeOriginal remains unchanged, so the images are properly ordered in my gallery, as gallery uses that instead of creation time.

3. Go to iLovePDF

   • I go to the iLovePDF website and use the Images to PDF tool.
   • I select all the images for conversion.
   • The images are never correctly ordered regardless of how I select them, unless I select one at a time, which is tedious.

4. Manually Sort Images

   • I end up having to manually check the file names to figure out the right order.
   • This is difficult because edited images usually have later file names.
   • I often use the original handwritten note to guide the sorting process.
   • This step can take up to an hour depending on the size of the note.

5. Generate and Share PDF

   • Once sorted, I generate the PDF and share it with my colleagues.

This entire process is extremely time-consuming and tedious, mainly because of sorting the images correctly before generating the PDF.

Thus, tge idea here was to automate this process. I therefore built a handwritten note management system powered by a Runner H agent.

This project automates the process: images are grouped into "notes", organized via a simple API, and converted into PDFs. The Runner H agent understands how to communicate with this backend via an extensive prompt, acting as an intelligent assistant that can create, manage, and retrieve notes, uploads, and PDF exports.

Demo

This chat illustrates Runner H issuing commands to create and export notes based on handwritten image uploads.

How I Used Runner H

I defined a clear, version-controlled API contract that Runner H reads from a public endpoint. This prompt includes documentation for all backend endpoints, including path, method, and parameter details.

Runner H interprets natural language queries and converts them into valid API calls, such as:

• Creating a note and attaching image uploads
• Renaming a note or one of its PDFs
• Generating a PDF from a selected group of uploads or a previously saved snapshot
• Listing available notes, PDFs, and snapshots

The system uses secure validation (access_token and build_id) to ensure that only the latest version is used and unauthorized requests are rejected. Admin-only endpoints allow for configuration of these values.

Interacting with Runner H feels like talking to a personal assistant who understands your note workflow end-to-end, so it's essentially a UI-like interface just in NL.

Use Case & Impact

This system benefits anyone who's looking to do do automation with AI as interface for an API, particularly using the tools I've used here which include Google Apps Script, Drive, and similar. Cor me personally, instead of dragging files, renaming them, and manually generating documents, I can now use a conversational interface to automate everything.

By combining a backend powered by Google Apps Script with Runner H’s agent interface, the solution creates a minimal-effort workflow: I just upload images, tell the agent what I want, and it handles everything else reliably.

Social Love