PPTHEC

PPTHEC is a sanitized starter kit for a PPT production workflow:

source notes/images -> source_material.pdf -> prompt/library profile -> NotebookLM Part A/B slide decks -> exported PDFs -> merged PDF -> optional speaker script

This repository intentionally does not include private cookies, Google tokens, Chrome profiles, NotebookLM auth files, real Xiaohongshu note data, generated PPT/PDF outputs, or local machine paths.

What This Repo Provides

• A clean input-pack builder that turns local note folders into NotebookLM-ready folders.
• Example jiaowu and yihu prompt-library templates.
• Runtime docs for NotebookLM, visual profiling, Part A/B generation, polling, export, and resume.
• Speaker-script delivery rules and a validator that blocks prompt/markdown/internal-metadata leakage.
• A secret scanner to reduce the chance of publishing private tokens or local paths.
• Example PowerShell commands that users can adapt to their own NotebookLM automation code.

What This Repo Does Not Provide

• No bundled Xiaohongshu cookies or bypass logic.
• No NotebookLM account, browser profile, or auth state.
• No GRSAI/OpenAI-compatible API key.
• No real source images, PDFs, PPTs, run states, logs, or generated outputs.

Quick Start

1. Install Python dependencies:

python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -e .[dev]

2. Copy environment template:

Copy-Item .env.example .env

3. Put exported note folders under examples/local_notes/, one folder per note.

Each note folder can contain images or an existing source_material.pdf.

4. Build NotebookLM input folders:

ppthec prepare-inputs --source-dir examples/local_notes --output-dir runs/demo/notebooklm_inputs --library-type jiaowu

5. Check local runtime:

ppthec doctor

6. Run your NotebookLM batch runner.

This repo ships a generic example at scripts/run_workflow.example.ps1. You must point it to your own NotebookLM automation implementation.

7. If your workflow also generates speaking scripts, validate the final deliverable text before copying it to the customer/teacher output folder:

ppthec validate-speaker-script runs/demo/final_output/note-a_speaker_script.txt

Input Folder Contract

After prepare-inputs, every note becomes:

notebooklm_inputs/
  note-title/
    source_material.pdf
    batch_requirements.txt
    manifest.json

batch_requirements.txt is a per-note instruction card. It tells downstream prompt generation which library to use and what scenario boundaries to preserve.

Prompt Libraries

• templates/jiaowu/dynamic_visual.md: education, class meeting, psychology class, parent meeting, teacher-facing content.
• templates/yihu/dynamic_visual.md: medical, nursing, health education, patient education, clinical training.

Force the intended library explicitly when possible:

ppthec prepare-inputs --library-type jiaowu

Safety Before Publishing

Run:

ppthec scan-secrets .

The scanner blocks common local paths, cookies, tokens, auth files, and generated runtime folders. It is not a replacement for manual review, but it catches the mistakes that usually leak private workflow state.

Recommended NotebookLM Runtime Shape

ingest workers: 2-3
submit workers: 1-2
poll workers: 1
export workers: 1-2
max active deck jobs: 12 or lower
submit interval: 10-30 seconds

When NotebookLM reports daily Slides limit or rate limit, stop submitting new deck jobs. Keep polling/exporting already submitted jobs.

License

MIT. Replace it before publishing if your team needs a different license.