Quickstart

mmarkdown is built for AI product teams that need reliable source-to-Markdown normalization for agent workflows.

Example request

API_KEY="YOUR_API_KEY"

curl -X POST http://127.0.0.1:8000/v1/convert \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "source_type": "txt",
    "source": "# Weekly Update\nShipped 3 improvements."
  }'
  • Use source_type: "txt" for plain text
  • Use source_type: "url" for web pages
  • Unsupported source types currently return typed NOT_IMPLEMENTED_SOURCE_TYPE

Bulk conversion pattern

  • Bulk processing is done as many per-item POST /v1/convert calls.
  • Run calls in worker queues with bounded concurrency and retries by error code.
  • Use idempotency keys in your worker layer to avoid duplicate processing.

How file inputs work

  • Current live scope is URL + TXT.
  • For files, standard pattern is: upload files to object storage, then send a signed URL to mmarkdown.
  • Priority roadmap for direct file support: PDF, DOCX, CSV/TSV.

Useful endpoints

  • GET /healthz — liveness check
  • GET /v1/metrics — 24h reliability summary
  • GET /v1/metrics/failures — recent typed failures