PHOENIX OMEGA
Tech-lead onboarding

Audiobooks + podcasts

EXECUTED-REAL (audiobook samples) / SPECCED (podcast proposed)

Audiobook evidence stops at chapter samples; no full-length book has been put before the operator. The podcast pipeline is proposed and its implementation scripts are explicitly missing.

What this pipeline produces

A full-length, listenable audiobook file assembled from all chapters, with its script-quality evidence; for podcasts, episode MP3s, stable metadata, and a feed.xml package.

Where it really stands

Acceptance layer: EXECUTED-REAL (audiobook samples) / SPECCED (podcast proposed)

Evidence: artifacts/audiobook/flagship_gen_z_anxiety/

Audiobook evidence stops at chapter samples; no full-length book has been put before the operator. The podcast pipeline is proposed and its implementation scripts are explicitly missing.

Run it yourself

REQUIRES-LIVE-RUN
find scripts/audiobook -maxdepth 2 -type f -name '*.py' -print
rg -n "ArgumentParser|add_argument|__main__" scripts/audiobook

If a runner and its CLI contract appear, use only its reported flags and confirm it emits under artifacts/audiobook/. If no runner appears, the supplied spec’s comparator loop remains architecture rather than an executable entrypoint.

For podcast, the authority says scripts/podcast/*.py is missing. There is no honest run command until implementation lands.

The map

Verified path in supplied contextRole
docs/audiobook_ops_manual.mdPublishing operations reference
docs/AUDIOBOOK_PIPELINE_SPEC.mdComparator-loop contract
scripts/audiobook/Audiobook implementation area
config/audiobook_script/Comparator and quality config
config/video/audiobook_style.yamlAudiobook visual/style config
artifacts/audiobook/flagship_gen_z_anxiety/Existing chapter WAV evidence
docs/PODCAST_PIPELINE_INTEGRATION_SPEC.mdProposed podcast authority
scripts/podcast/Proposed implementation area

What is broken right now

landing

Render chapters 4–12 on NIM Magpie, assemble a full-length audiobook, and obtain operator review.

unknown

The supplied context does not identify an audiobook CLI filename or verified end-to-end flags.

landing

Podcast assembly, render, feed, upload, admin API, UI, and weekly workflow are specified but not implemented.

The traps

  • Chapter-1 samples do not prove a full-length audiobook. Measure and review the assembled customer file.
  • Do not describe the podcast pipeline as active: the canonical authority map marks it proposed.
  • A comparator PASS concerns script adaptation; it does not substitute for listening to the full rendered audio.
  • A gate PASS is at most structurally clear; never collapse it into an unlayered “all checks pass” claim.

Who to ask

Pipeline owner

Pearl_Dev (audiobook) · Pearl_Prime (podcast, proposed)

Coordination

Pearl_PM owns scope, sequencing, and status claims through artifacts/coordination/ACTIVE_WORKSTREAMS.tsv.

Architecture

Pearl_Architect owns architecture decisions and cap rulings through docs/PEARL_ARCHITECT_STATE.md.

Your first week

1
Inventory the live audiobook CLIs and document their exact help output without inventing flags.
2
Render chapters 4–12 through NIM Magpie and retain per-chapter WAV evidence.
3
Assemble one full-length audiobook and verify duration, bytes, chapter order, and audible joins.
4
Put the full file before the operator and record the acceptance outcome.