zotero-notebooklm

name	zotero-notebooklm
description	Use when the user wants Codex to organize newly added Zotero papers into a NotebookLM-powered Chinese briefing, Audio Overview, or Video Overview. This skill orchestrates DailyInfo's zotero-brief capability, notebooklm-py authentication, Zotero PDF hydration, upload, generation, download, and manual fallback steps.

Zotero NotebookLM Briefing

Use this skill when the user asks Codex to turn Zotero papers into a NotebookLM briefing, podcast/audio overview, or video overview.

DailyInfo is the capability layer. Codex is the operator. The user may manually complete Google/NotebookLM login or browser steps when required.

For full new-machine setup and human/agent handoff details, read docs/zotero-notebooklm.md or docs/zotero-notebooklm.zh.md when onboarding a machine or when auth/PDF access is unclear.

Contract

• Do not use DailyInfo's OpenRouter summarizer for this workflow.
• Use dailyinfo zotero-brief as the primary command surface.
• Use notebooklm-py via uv run --extra notebooklm ...; do not assume the global notebooklm command is installed.
• Treat NotebookLM auth as manual-friendly: if auth is missing, give the exact notebooklm login command and explain that Codex can continue after login.
• Prefer NotebookLM sources over separate LLM calls: upload copied PDFs plus source_index.md, then ask NotebookLM for the Chinese briefing and optional audio/video overview.
• If automation fails, inspect notebooklm.json and MANUAL_NOTEBOOKLM_STEPS.md; report the next concrete action instead of treating the run as a total failure.

Standard Workflow

1. Confirm environment from the DailyInfo repo root:

   $env:UV_CACHE_DIR='D:\Code\dailyinfo\.tmp\uv-cache'
   uv run --extra notebooklm dailyinfo zotero-brief --help
   uv run --extra notebooklm notebooklm doctor
   

2. If notebooklm doctor reports missing auth, ask the user to complete login with the same profile directory that the run will use:

   $env:NOTEBOOKLM_HOME='D:\Code\dailyinfo\.tmp\notebooklm'
   $env:UV_CACHE_DIR='D:\Code\dailyinfo\.tmp\uv-cache'
   uv run --extra notebooklm notebooklm login --browser chrome
   

   In a normal user terminal, the user may omit UV_CACHE_DIR. In Codex sandboxed runs, keep NOTEBOOKLM_HOME inside the workspace if the default home directory is not writable.

3. Run a small PDF hydration smoke test when PDFs are on Google Drive or another cloud placeholder path:

   $env:DAILYINFO_DATA_ROOT='D:\Code\dailyinfo\.tmp\dailyinfo-run'
   $env:NOTEBOOKLM_HOME='D:\Code\dailyinfo\.tmp\notebooklm'
   $env:UV_CACHE_DIR='D:\Code\dailyinfo\.tmp\uv-cache'
   uv run --extra notebooklm dailyinfo zotero-brief --collection water --date 2026-05-27 --manual-only --force --limit 1 --open-missing-pdfs --pdf-wait-seconds 0 --notebooklm-home D:\Code\dailyinfo\.tmp\notebooklm
   

   Then read notebooklm.json. open_target should show a zotero://open-pdf/library/items/... URI when hydration was attempted.

4. Run the full workflow:

   $env:DAILYINFO_DATA_ROOT='D:\Code\dailyinfo\.tmp\dailyinfo-run'
   $env:NOTEBOOKLM_HOME='D:\Code\dailyinfo\.tmp\notebooklm'
   $env:UV_CACHE_DIR='D:\Code\dailyinfo\.tmp\uv-cache'
   uv run --extra notebooklm dailyinfo zotero-brief --collection water --date 2026-05-27 --artifact audio --force --open-missing-pdfs --notebooklm-home D:\Code\dailyinfo\.tmp\notebooklm
   

   Adjust --collection, --date, and --artifact none|audio|video|both to match the user's request.

5. Verify by reading the output directory:

   • source_index.md: paper metadata and NotebookLM instructions.
   • pdfs/: copied PDFs. Empty is acceptable only if the user will complete upload manually.
   • briefing.md: generated Chinese briefing or a placeholder with the blocker.
   • notebooklm.json: notebook/source/artifact ids and errors.
   • audio_overview.mp3 or video_overview.mp4: present only after successful download.
   • MANUAL_NOTEBOOKLM_STEPS.md: browser fallback steps.

Failure Handling

• Auth missing: run or ask the user to run notebooklm login with the same NOTEBOOKLM_HOME, then rerun dailyinfo zotero-brief.
• Zotero not reachable: tell the user to open Zotero Desktop and ensure the local API on 127.0.0.1:23119 is enabled.
• PDFs missing or access denied: rerun with --open-missing-pdfs; if Codex still cannot read cloud-drive paths, the user should hydrate/open the files in Zotero or run the same command from a normal terminal.
• NotebookLM upload/generation failed: use the already written source_index.md, pdfs/, and briefing_prompt.md in the NotebookLM web UI, then save results into the output directory.

Reporting

Report:

• Date, collection, number of papers found, and output directory.
• Whether NotebookLM auth was ready.
• How many PDFs copied and which blocker remains, if any.
• Whether briefing/audio/video artifacts were created.
• The next manual step if automation stopped.