Inspiration
This project began with Alan’s keynote at the start of the hackathon. Hearing how he built tools used by millions of people in their daily workflow pushed me to try building something meaningful as well.
The idea for DocFlow.ai came directly from my experience as an intern software engineer. Understanding a large codebase without reliable documentation slows everything down. You can ask teammates for help, but often they’re just as unsure about certain parts of the system. You end up navigating the code on your own while still facing tight deadlines.
Documentation is essential, yet nobody enjoys writing it, and most teams can’t keep it consistently updated. DocFlow.ai is meant to remove that burden by automatically maintaining the documents developers depend on.
What it does
DocFlow.ai monitors your codebase for changes and updates your documentation automatically, ensuring it always reflects the current state of the system.
How we built it
The system is built using Node.js, Express, and webhooks. For model processing, I integrated Gemini and OpenRouter to handle content generation and updates with consistent quality.
Challenges we ran into
Managing ngrok URLs each time the server restarted made early development slow. Structuring and labeling documentation in a coherent way also took several design iterations. Creating architecture diagrams and overview documents required careful planning so the output would be useful and not just text.
With guidance from Wei and the mentors, these challenges became clearer, and I was able to integrate all the necessary features into a working solution.
Accomplishments that we're proud of
I’m proud that I kept pushing even while working solo. Getting a functional demo built - something concrete and usable - was an important milestone and a reminder that progress compounds with persistence.
What we learned
I gained practical experience with webhooks, AI prompt design, documentation generation, and the workflow developers rely on when working with large codebases.
What's next for DocFlow.ai
The next step is integrating with GitHub Apps so teams can install DocFlow.ai directly into their repositories. From there, the focus is on talking to real users, understanding what matters most to them, and iterating based on real feedback.
Link to demo:
Built With
- express.js
- gemeni
- ngrok
- node.js
- openrouter
- python
Log in or sign up for Devpost to join the conversation.