notebooklm-equivalent — Source-Grounded Research & Audio-Overview Engine

> HONEST LABEL (خط أحمر مصداقية): Google NotebookLM has NO public developer API > (verified live 2026-08-09: no self-serve keys/endpoints; Enterprise APIs only via > GCP Agentspace early-access/OAuth — not our key path). This skill is NOT > NotebookLM and calls no NotebookLM endpoint. It is a functional equivalent > built from our real arsenal (Gemini 3.1 Pro grounded analysis + our approved TTS > gateway for the audio overview).

What it actually does

Ingest a source (raw text / local file / URL / stdin), then produce source-grounded outputs — it uses ONLY the provided material and explicitly flags what the source does not cover (NotebookLM's signature behavior):

summary — faithful key points + conclusions
studyguide — concepts, definitions, review Q&A
faq — top questions with grounded answers
timeline — chronological sequence (or "no temporal sequence found")
podscript — two-host audio-overview script (feeds our TTS)

Audio overview (podcast)

podscript produces the two-host script. Actual audio is generated by our approved gateway — do NOT re-implement TTS:


python3 skills/notebooklm-equivalent/notebook.py podscript --source <src> --minutes 10 > /tmp/pod.txt
bash scripts/media_say.sh "$(cat /tmp/pod.txt)"   # Gemini Charon chain; Arabic auto-tashkeel

Grounding fidelity (verified in 5-level critique)

Fictional "Zorblax metal" source → summarized strictly from source, zero outside facts added, flagged missing temperature scale.
All modes cite short source phrases; refuse to invent uncovered details.

Usage


python3 skills/notebooklm-equivalent/notebook.py summary    --source <file|url|-> [--json]
python3 skills/notebooklm-equivalent/notebook.py studyguide --source <file|url|->
python3 skills/notebooklm-equivalent/notebook.py faq        --source <file|url|->
python3 skills/notebooklm-equivalent/notebook.py timeline   --source <file|url|->
python3 skills/notebooklm-equivalent/notebook.py podscript  --source <file|url|-> [--minutes N]
python3 skills/notebooklm-equivalent/notebook.py selftest

Notes

Uses GEMINI_API_KEY with automatic fallback to GEMINI_API_KEY_US (both Pro-tier, identical catalogs — mutual backup).
Source cap ~120k chars/call. For PDFs, extract text first (pdftotext / our pdf tool).
For a full REST NotebookLM-style SaaS alternative there is AutoContent API (paid, 3rd-party) — not used here to keep KCPC/private data in-house.

Verified

2026-08-09: py_compile OK · selftest SELFTEST_OK (live Gemini) · 5-level critique all passed.