cjm-transcript-correction-tui 0.0.1

Transcript-correction TUI driver: keyboard-first terminal review surface (cursor-windowed segment walk, VAD-chunk audio auto-play, fidelity + boundary-shift corrections, agents-propose/humans-confirm triage) over cjm-transcript-correction-core's operation vocabulary. The first born-on-graph library of the self-hosting graph arc.

cjm-transcript-correction-tui

The minimal presentation driver for cjm-transcript-correction-core: a document-order segment walk over the shared transcript graph with chunk audio playback, held-key boundary-shift gestures, inline text correction, and review markers — the human pass-1/pass-2 correction loop's UI. Kept a separate library from the headless core per the non-minimal-drivers-separate-libs rule; every correction write appends through to the workflow's sidecar journal.

Modules

• cjm_transcript_correction_tui.__init__ — Keyboard-first TUI driver for the transcript-correction workflow — a presentation
• cjm_transcript_correction_tui.app
• cjm_transcript_correction_tui.audio
• cjm_transcript_correction_tui.cli
• cjm_transcript_correction_tui.spine

API

cjm_transcript_correction_tui.app

• CorrectionApp class — The correction loop, v0 thinnest slice: document-order segment walk with
• load_tui_state function — Read the per-graph TUI sidecar state (last-focused positions).
• save_tui_state function — Merge one source's last-focused position into the sidecar state file.

cjm_transcript_correction_tui.audio

• ChunkPlayer class — Persistent-output-stream VAD-chunk player — the focus-walk auto-play engine.
• load_chunk function — Read one VAD chunk's samples from the model-input WAV — frame-sliced, sample-accurate.
• stretch function — Pitch-preserving time-stretch (WSOLA, numpy-only) — the playback-speed engine.

cjm_transcript_correction_tui.cli

• build_parser function — The TUI driver's argument surface (mirrors correction-core's run/review args).
• main function — Parse args, run the correction loop (the app owns the event loop + teardown).

cjm_transcript_correction_tui.spine

• ChunkRef class — Where one Segment's VAD-chunk audio lives: the model-input WAV + the chunk-local span.
• SpineView class — One Source's effective correction spine, cursor-windowed for the TUI.
• plan_boundary_shift function — Plan a ONE-WORD boundary shift (the [ / ] gesture unit).

Dependencies

Depends on: cjm-context-graph-layer, cjm-transcript-correction-core, numpy, sounddevice, soundfile, textual