Back to work

Document workflow infrastructure

FileMorf

A privacy-first file conversion and document workflow platform with browser conversion, server jobs, R2 storage, OCR, public APIs, recipes, and agentic document operations.

Live with upgrade in progress Next.jsReactHonoPostgresRedisBullMQ

Problem

File workflows look simple at the UI layer, but reliable conversion needs storage, queues, status, artifacts, limits, and repeatable execution.

System

FileMorf splits browser-first conversion from retained server workflows, public API access, and document intelligence layers.

Highlights

  • Browser-side image, PDF, document, spreadsheet, structured data, and archive conversion flows.
  • Server job lifecycle with signed uploads, quotas, status polling, output artifacts, and cleanup workers.
  • Document intelligence with OCR, summaries, highlights, citation-backed Q&A, comparison, and operator workflows.

Key decisions

  • Use signed R2 upload and download URLs so file handling stays bounded and explicit.
  • Separate conversion, render, OCR, workflow, cleanup, and operator lanes behind queues.
  • Persist artifacts, warnings, outputs, and verification records so document work can be reviewed.

Evidence

  • Public API keys, bearer auth, permissions, per-key rate limits, and idempotency fingerprints.
  • Workflow recipes with ordered conversion, OCR, and document steps.
  • Operator infrastructure with typed plans, persisted artifacts, verifier paths, and benchmark coverage.