The Story:
Got laid off earlier this year. Instead of just sending out applications, I decided to build something I wish existed: an AI tool that automatically generates documentation from code.
What It Does:
Connects to your GitHub repos and creates comprehensive documentation using GPT-4 and tree-sitter code parsing.
Features:
- Supports 6 languages: Python, JavaScript, Java, Ruby, Go, C++
- Generates API docs, function/class references, and technical guides
- GitHub OAuth integration
- Analyzes code structure intelligently
Try it: https://docs-generator-two.vercel.app
Tech Stack
- Backend: FastAPI (Python)
- Frontend: React + TypeScript + Vite
- Database: PostgreSQL
- AI: OpenAI GPT-4
- Hosting: Railway + Vercel
- Auth: GitHub OAuth
Why I Built This
Documentation is tedious. Every developer I know (including myself) hates writing docs but knows they're critical. I wanted to see if AI could actually do this well.
What I'm Looking For
I need 5 honest beta testers before I productize this.
Questions I need answered:
- Does this actually solve a problem?
- What's missing?
- Would you pay for this?
- What features would make it genuinely useful?
The Build Journey
Built in about a month. Some interesting technical challenges:
- Getting tree-sitter working with all 6 languages (dependency hell)
- Designing prompts that generate useful docs (not just code summaries)
- Handling large repos without timeout issues
Try It Out
Demo: https://docs-generator-two.vercel.app
Would love your brutally honest feedback in the comments!
Top comments (1)
It shines on web backends, SDKs/CLIs, and infra tools-we’ve run it from tiny repos to a chunky 12k‑file polyglot monorepo, English live with translations in beta. Export to Markdown for Docusaurus/MkDocs/Sphinx, run via GitHub Actions, review diffs in the editor, and keep your code locked down with minimal OAuth scopes, encryption, and optional ephemeral runs. We’re onboarding five beta testers-if you’ve got a medium‑to‑large repo and want CI/export integration, give me a shout.