Sefer APIs

Torah source infrastructure, through one key

The Sefer APIs expose 4.25M Hebrew segments, 8.7M graph relationships, and a production-shaped transcription pipeline. Every request passes through a scoped edge key and an auditable account usage ledger.

The graph has two deliberately different trust surfaces:

• Citations are citations-of-record: curated links, or explicit extractions carrying the exact Hebrew evidence from the source.
• Leads are AI-inferred relationships. They retain provenance and a model_score, and are never presented as verified citations.

What you can do

• Resolve English-canonical or Hebrew-daf references to canonical Hebrew text.
• Retrieve strict incoming and outgoing citations.
• Explore inferred relationships without confusing them with citations-of-record.
• Search the Hebrew corpus and translate selected passages.
• Upload Torah audio, run asynchronous transcription, poll the job, and retrieve or permanently delete the result.
• Inspect your own usage and dollar-credit balance.

curl -X POST https://sefer-graph-worker.mordechaipotash.workers.dev/v1/resolve \
  -H "Authorization: Bearer $SEFER_KEY" \
  -H "Content-Type: application/json" \
  -d '{"refs":["Genesis 1:1","Berakhot מב., י"],"include":["text","citers"]}'

Start with the Quickstart, or go directly to Transcription.