cjm-transcript-decomp-qt

Qt shell for the decomp-batch setup app — the second workflow TUI migrated to the PySide6 lane (DEC c4b0d6e5 leg 3, architecture 6c574c89). Imports the pure spine of cjm-transcript-decomp-tui (run/propset/training-run indexes, group_batches, SegmentStack + the fold/probe helpers, sidecar state, discovery) and shares its CLI ladder (build_parser/resolve_split_flags/resolve_settings/hand_off), so Qt-confirmed batches hand off to cjm-transcript-decomp-core byte-identically. The read-only graph seat, the VAD/FA/sentence probe chain, and the P propose pre-stage run under Qt via DecompSession — a kit LoopThreadSession owning the tui SegmentStack, resolving to the shell through queued Signals. Audio = SpanPlayer on QMediaPlayer (position-window span playback for fine segments AND gap rows).

Modules

• cjm_transcript_decomp_qt.__init__
• cjm_transcript_decomp_qt.app — The Qt decomp-batch shell: the same batch queueing over transcription-run
• cjm_transcript_decomp_qt.cli — Console-script driver for the Qt shell: the SAME argument surface,
• cjm_transcript_decomp_qt.panes — Pure row/label builders for the decomp Qt shell — the paint logic, Qt-free.
• cjm_transcript_decomp_qt.propset_picker — Modal proposal-set lifecycle picker (the b20cb911 propset-unarchive
• cjm_transcript_decomp_qt.session — One read-only decomp inspection seat + the propose pre-stage behind a
• cjm_transcript_decomp_qt.trainrun_picker — Modal training-run picker for the propose pre-stage (work item a13ebc66):

API

cjm_transcript_decomp_qt.app

• DecompWindow class — Batch-setup window; exit with plan set = the CLI hands off headless.
• fmt_span function — Status text for a playing span (source-coordinate seconds).

cjm_transcript_decomp_qt.cli

• main function — Resolve the shared setup surface, run the Qt batch window, hand off.

cjm_transcript_decomp_qt.panes

• batch_html function — The hand-off fold panel: how many core invocations the ordered picks
• config_diffs function — The form's deltas from the RUN's recorded config, as key=value chips.
• decomp_header function
• decomp_rows function — One row per decomp run: id, timestamp, source/segment totals, source
• drill_header_html function — Drilled decomp run: id + the transcription manifest it consumed.
• drill_source_rows function — One row per extended source: title, fine-segment count, Source node.
• form_rows function — One row per config field; a value differing from the RUN's recorded
• probe_detail function — The focused predicted entry in full, plain text.
• probe_rows function — The predicted skeleton painted with the SEGMENTS grammar (direct
• probe_summary_html function — The probe readout: predicted vs committed counts + the recovered spans
• probe_target_html function — The frozen probe target: which coarse chunk the sweep runs over.
• probeview_header_html function — The predicted-skeleton headline: which config produced it, over which
• run_detail_html function — Focused-run detail (614dd647): WHAT was transcribed + WHICH graph it
• run_rows function — One row per transcription run: ORDERED queue position (the batch is a
• runs_header function — The listing header: corpus size + where the manifests come from.
• segment_detail function — The focused entry IN FULL as plain text (setPlainText — no truncation,
• segment_rows function — The committed fine spine with FOCUSABLE interleaved gap rows — the gap
• segments_header_html function — The drilled source's headline: row/gap/missing counts + the recorded
• trainrun_rows function — One picker row per training run (work item a13ebc66): the manifest
• vadform_header_html function — Form headline + the probe-only doctrine + the split-preview state.

cjm_transcript_decomp_qt.propset_picker

• PropsetPicker class — One source's proposal-set ring as a native lifecycle picker.

cjm_transcript_decomp_qt.session

• DecompSession class — The loop-thread seat pair for the decomp shell.
• bench_gestures function — The decomp shell's canonical gesture for the scale harness (design detail 4389a579): the

cjm_transcript_decomp_qt.trainrun_picker

• TrainingRunPicker class — The propose-model picker: a header row (kit modal_header, mouse

Dependencies

Depends on: PySide6, cjm-context-graph-layer, cjm-substrate, cjm-substrate-qt-kit, cjm-transcript-decomp-core