I got tired of building custom OCR pipelines for every document type and fixing parsers whenever a PDF layout changed. So I built Parsley.
Parsley uses LLMs to parse PDFs or images into structured data. You define the fields you want and it extracts them by meaning rather than fixed positions or labels.
The web app uses your own API keys with providers like Google or OpenRouter. Documents go directly to them. The app is stateless and doesn't store files or keys.
Features:
- Custom or auto-generated schemas
- PDF (including password-protected) and image support
- Multiple LLM providers
- JSON or CSV output
- Stateless API for n8n or Zapier
- Free, rate-limited demo mode
I use it in n8n to parse invoices and bank statements straight into spreadsheets. It's been much less brittle than traditional OCR.
I build full-stack apps from the first spark of an idea all the way to deployment. I enjoy working with modern web stacks (the React family) and solid backends using Node/Bun, PostgreSQL, Docker, GCP, or even a single VPS.
I also work heavily with AI using LLM APIs, MCP, and RAG to add useful functionality to products when it makes sense. If you're exploring something new or want to bounce around ideas, I'm always happy to talk.