NotebookLM Content Creation

Creates NotebookLM Studio content (Audio Overview, Video Overview, Infographics, Slides) and monitors it to completion using a background polling loop.

Requires:

• notebooklm-mcp-cli installed: uv tool install notebooklm-mcp-cli
• Authenticated: nlm login (done on the server already)

Studio Types

⚠️ CLI 与 SKILL 旧版不一致（重要！）

nlm studio create 不支持 --type 参数！每种产出类型是独立的顶级命令：

• nlm video create <notebook_id> --format explainer --language zh-CN --confirm
• nlm infographic create <notebook_id> --detail detailed --orientation landscape --language zh-CN --confirm
• nlm slides create <notebook_id> --format detailed_deck --length default --language zh-CN --confirm

⚠️ Download 命令

• Audio: nlm download audio <notebook_id> --id <artifact_id> -o <path>
• Video: nlm download video <notebook_id> --id <artifact_id> -o <path>
• Infographic: nlm download infographic <notebook_id> --id <artifact_id> -o <path>
• Slides: nlm download slide-deck <notebook_id> --id <artifact_id> -o <path> ⚠️ 不是 nlm download slides！

Workflow

Step 1 — Notebook Selection

List all notebooks:

nlm notebook list

Parse the JSON output for id and title. Match against the user's keyword (case-insensitive substring match). If multiple match, present options with numbers.

If no notebook matches:

• Ask user: "No notebook found matching '[keyword]'. Create a new one or add more sources to an existing notebook?"
• If user confirms new notebook: create with nlm notebook create "<name>"
• Then add sources: nlm source add <notebook_id> --url <url> --wait

Step 2 — Check Existing Artifacts

Before creating new content, check if the notebook already has generated artifacts:

nlm studio status <notebook_id>

If artifacts with status: completed exist, show them to the user and ask:

"This notebook already has completed content. Download existing [type] or generate new content?"

• Download existing: go directly to download step
• Generate new: proceed to Step 3

Step 3 — Pre-Flight Confirmation OR Auto-Execute

Interactive mode (user initiated): Ask all parameters at once. Write in the user's current session language.

Creating [Audio/Video/Infographic/Slides] Overview from "[notebook name]"

Please confirm:

① Content type: [Audio Overview / Video Overview / Infographics / Slides]
② Format: [deep_dive / brief] (default: deep_dive)
③ Length: [short / default / long] (default: default) — not available for Infographics/Slides
④ Language: [BCP-47 code, e.g., en, zh-CN] (default: notebook's detected language or en)
⑤ Output path: [path] (default: ~/ObsidianVault/Default/NotebookLM/<notebook-name>/)

Reply with any changes, or "ok" to proceed with defaults.

Triggered mode (upstream skill chaining): When the agent receives a trigger message containing all required parameters (e.g., from Deep Research), skip user confirmation and auto-execute. The trigger message should include:

• 报告路径 / report_path: path to the source file to upload
• Notebook 名称 / notebook_name: name for the notebook (create if not exists)
• 产出类型: Audio Overview / Video Overview / Infographics / Slides
• 格式: deep_dive / brief / etc.
• 长度: short / default / long
• 语言: BCP-47 code

In triggered mode, the agent should:

1. Create notebook with nlm notebook create "<notebook_name>"
2. Upload source with nlm source add <notebook_id> --file <report_path> --wait
3. Proceed directly to Step 4 (Create Content) with the provided parameters
4. Set up polling and notify user when complete

Step 4 — Create Content

Based on user's confirmed parameters:

Audio:

nlm audio create <notebook_id> --format <format> --length <length> --language <lang> --confirm

Capture the returned Artifact ID.

Video:

nlm video create <notebook_id> --format <format> --style <style> --language <lang> --confirm

Infographics:

nlm infographic create <notebook_id> --detail <level> --orientation <orientation> --language <lang> --confirm

Slides:

nlm slides create <notebook_id> --format <format> --length <length> --language <lang> --confirm

Capture the returned Artifact ID for each.

Step 5 — Set Up Task Directory

Create a temp directory following the polling best practices pattern:

/tmp/notebooklm-studio/
  <YYMMDD-HHmm>_<sanitized-notebook-name>_<studio-type>/
    task.json          ← full task metadata
    progress.json      ← poll count, artifact id, last status
    poll.log           ← each poll attempt
    error.log          ← errors
    done.flag          ← created on success
    <output file>      ← downloaded artifact

Write task.json:

{
  "notebook_id": "<id>",
  "notebook_name": "<name>",
  "artifact_id": "<id>",
  "studio_type": "audio",
  "output_path": "~/ObsidianVault/Default/NotebookLM/<notebook-name>/<output_filename>",
  "poll_interval_seconds": 300,
  "max_polls": 8,
  "created_at": "<ISO timestamp>"
}

Step 6 — Notify User and Launch Background Polling

Notify the user (in current session language):

"Content generation started. I'll monitor it in the background and notify you when it's ready (typically 2–5 minutes). Poll every 5 minutes, max 40 minutes."

Launch the polling script in the background:

cd /tmp/notebooklm-studio/<task-dir>/
nohup bash /tmp/notebooklm-studio/poll.sh > /dev/null 2>&1 &

Step 7 — Polling Script

Write this script to <task-dir>/poll.sh:

#!/bin/bash
set -euo pipefail

TASK_DIR="/tmp/notebooklm-studio/<task-dir>"
cd "$TASK_DIR"

[[ -f done.flag ]] && echo "Already done." && exit 0

CHAT_ID="INJECT_CHAT_ID"  # ← Agent: replace with current Discord channel ID (from inbound_meta.chat_id)
POLL_COUNT=$(grep '"poll_count"' progress.json 2>/dev/null | sed 's/[^0-9]//g') || POLL_COUNT=0
MAX_POLLS=8
INTERVAL=300

log() { echo "[$(date '+%Y-%m-%d %H:%M:%S')] $*" | tee -a poll.log; }

notify_user() {
  local message="$1"
  openclaw message send --channel discord --target "$CHAT_ID" -m "$message" 2>/dev/null || log "WARNING: notification failed"
}

while true; do
  POLL_COUNT=$((POLL_COUNT + 1))

  if [[ $POLL_COUNT -gt $MAX_POLLS ]]; then
    log "TIMEOUT after $MAX_POLLS polls"
    echo "Timeout" >> error.log
    notify_user "❌ NotebookLM 播客生成超时。"
    exit 1
  fi

  log "[Poll $POLL_COUNT/$MAX_POLLS] Checking status..."
  RESULT=$(nlm studio status "$(grep '"notebook_id"' task.json | sed 's/.*: *"\([^"]*\)".*/\1/') 2>&1) || true
  echo "$RESULT" >> poll.log

  # Check if our artifact is completed
  ARTIFACT_STATUS=$(echo "$RESULT" | grep -A5 "\"id\": \"$(grep '"artifact_id"' task.json | sed 's/.*: *"\([^"]*\)".*/\1/')\"" | grep '"status"' | sed 's/.*: *"\([^"]*\)".*/\1/' | head -1)
  log "Artifact status: '$ARTIFACT_STATUS'"

  if [[ "$ARTIFACT_STATUS" == "completed" ]]; then
    log "Completed. Downloading..."
    OUTPUT_PATH=$(grep '"output_path"' task.json | sed 's/.*: *"\([^"]*\)".*/\1/')
    STUDIO_TYPE=$(grep '"studio_type"' task.json | sed 's/.*: *"\([^"]*\)".*/\1/')
    NOTEBOOK_NAME=$(grep '"notebook_name"' task.json | sed 's/.*: *"\([^"]*\)".*/\1/')
    nlm download audio "$(grep '"notebook_id"' task.json | sed 's/.*: *"\([^"]*\)".*/\1/')" --id "$(grep '"artifact_id"' task.json | sed 's/.*: *"\([^"]*\)".*/\1/')" -o "$OUTPUT_PATH" >> poll.log 2>&1 || true
    if [[ -s "$OUTPUT_PATH" ]]; then
      SIZE=$(du -h "$OUTPUT_PATH" | cut -f1)
      log "Downloaded: $OUTPUT_PATH ($SIZE)"
      touch done.flag
      notify_user "✅ 播客生成完成！[$NOTEBOOK_NAME] 已保存（$SIZE），共 $POLL_COUNT 轮。"
    else
      log "Downloaded file is empty"
      echo "Empty output" >> error.log
      notify_user "⚠️ NotebookLM 播客下载失败，文件为空。"
    fi
    exit 0
  fi

  if [[ "$ARTIFACT_STATUS" == "failed" ]]; then
    log "Generation failed"
    echo "Failed" >> error.log
    notify_user "❌ NotebookLM 播客生成失败。"
    exit 1
  fi

  # Save progress
  sed -i "s/\"poll_count\": [0-9]*/\"poll_count\": $POLL_COUNT/" progress.json
  log "Still in_progress. Sleeping ${INTERVAL}s..."
  sleep "$INTERVAL"
done

Initialize progress.json:

{
  "poll_count": 0,
  "last_poll_at": null,
  "last_poll_result": null
}

Step 8 — Completion

When polling exits (success or failure):

On success:

• Verify file exists and has content
• Move file to confirmed output path if not already there
• Notify user in current session language:

  "✅ [Content type] ready!\n\nNotebook: [name]\nSaved to: [path]\nSize: [size]\nPolls: [N] (~[X] minutes)"

• Leave temp folder for manual inspection

On failure/timeout:

• Notify user:

  "❌ [Content type] generation did not complete.\n\nNotebook: [name]\nReason: [timeout / auth error / generation failed]\n\nOptions:\n1. Re-run with same parameters\n2. Check NotebookLM web UI manually\n3. Clean up temp folder"

• Do NOT auto-retry or delete temp folder

Quick Reference

# List notebooks
nlm notebook list

# Check status
nlm studio status <notebook_id>

# Create audio
nlm audio create <notebook_id> --format deep_dive --length default --confirm

# Create video
# Create video
nlm video create <notebook_id> --format explainer --language zh-CN --confirm

# Create infographic
nlm infographic create <notebook_id> --detail detailed --orientation landscape --confirm

# Create slides
nlm slides create <notebook_id> --format detailed_deck --length default --confirm

# Download (after getting artifact ID)
nlm download [audio|video|infographic|slide-deck] <notebook_id> --id <artifact_id> -o <path>

⚠️ Download slides 的命令是 slide-deck，不是 slides！

Lessons Learned (2026-03-27)

CLI 命令与 SKILL 文档不一致

旧版 SKILL.md 使用的 nlm studio create --type video/infographic/slides 已废弃。实际 CLI 每种产出类型是独立的顶级命令：

Video 的选项也不同

Video 不支持 --length 或 --format deep_dive。正确选项：

• --format explainer/brief/cinematic
• --style auto_select/classic/whiteboard/kawaii/anime/watercolor/retro_print/heritage/paper_craft
• --language zh-CN

并行生成

4 种产出类型可以并行生成（同时提交多个 nlm create 命令），不需要排队。每种产出有独立的 artifact ID，可以独立轮询。

轮询最佳实践

• 轮询间隔：默认 5 分钟（300 秒）。对于 NotebookLM 这种较长任务足够用，且避免频繁请求。
• 超时：40 次 × 1 分钟 = 40 分钟，基本够用
• 完成后通过 openclaw message send --channel discord --target $CHAT_ID 直接投递到 Discord channel，不经过 agent 推理（避免模型超时导致通知丢失）。