Future TranscriptX integration seam¶
Transcribe does not depend on TranscriptX and must not be integrated into TranscriptX before its 1.0 release.
Handoff¶
transcribe.notebook JSON
↓
future pure adapter (post–TranscriptX 1.0)
↓
whatever TranscriptX's post-1.0 document/import contract requires
Portable export rules: contracts/notebook-export.md.
Non-goals for this seam document¶
Do not lock synthetic timestamps (e.g. page
i→[i, i+1)).Do not lock fake speaker labels (e.g.
PAGE/NOTEBOOK).Those may be one future compatibility strategy among others, but synthetic timing/speakers can leak into timing-, interaction-, or speaker-based analysis as plausible nonsense.
What Transcribe guarantees (export)¶
Portable
format: "transcribe.notebook"interchange with page order, effective/raw/edited text, content fingerprints, provenance, and notebookcontent_revisionwhen presentNo required absolute filesystem paths in the export
Page-first domain (not timed speaker segments)
Organisation tags as
tags: string[]slugs on notebook and page, plus an optionaltag_catalogsnapshot (personal_corpus.tag-catalogdefs) so labels/colours survive interchange
Reverse import (Transcribe 1.6 — not shipped)¶
After Transcribe 1.0, ROADMAP 1.6 may import TranscriptX export files (transcript + optional audio + summary + speaker strings) as contextual evidence around notebooks. That is a file adapter, not a runtime dependency either way, and it does not replace this export seam.
TranscriptX export bundle (files)
↓
future Transcribe context-import adapter (1.6)
↓
ContextCollection / ContextRecord (evidence around notebook pages)
Do not call TranscriptX APIs. Do not re-run ASR. Do not port speaker/audio analysis modules. Pin the TX export contract when TX 1.0 exists; a frozen fixture is allowed if that contract is late. Sequencing: ROADMAP.md After 1.0 § 1.6.