Transcripts

divinci.transcripts manages conversation transcripts in a workspace — listing them, reading them, and adding messages.

Creating and reading transcripts

const transcript = await divinci.transcripts.create({
  workspaceId: "ws_123",
  releaseId: "rel_abc123", // optional
});

const { data } = await divinci.transcripts.list("ws_123", { limit: 20 });
const one = await divinci.transcripts.get(transcript._id);

await divinci.transcripts.delete(transcript._id);

Adding a single message

const message = await divinci.transcripts.addMessage(transcript._id, {
  role: "user",
  content: "How do I reset my password?",
});

Batch ingestion (removed)

Caution

ingestBatch was removed on 2026-08-07. The method throws, and the endpoint answers 410 Gone.

It was the only write path that skipped the message pipeline: it bypassed moderation, de-identification, RAG retrieval and inference, leaving PII redaction to the caller. Because it accepted a caller-supplied message role, a system message written this way was replayed to the model as a real instruction on every later turn.

There is no drop-in replacement. If you run inference yourself, persist it yourself. To have Divinci generate and record a turn, use addMessage, which runs the full pipeline.

Method reference

Method	Description
list(workspaceId, options?)	Paginated transcripts.
get(transcriptId)	Fetch a transcript.
create(options)	Create a transcript (workspaceId, optional releaseId).
addMessage(transcriptId, options)	Append a single message.
ingestBatch(options)	Removed 2026-08-07 — throws. See above.
delete(transcriptId)	Delete a transcript.