Nitan Podcast：美卡论坛每周中文播客生成器

Nitan Podcast

Weekly Chinese podcast generated from hot discussions on USCardForum.

Nitan Podcast is the public show repository: it owns the podcast identity, feed, episode assets, forum post templates, and the Nitan/USCardForum-specific logic behind each episode. Reusable pipeline stages are provided by the pinned CastForge dependency; this repo remains the canonical home of the show.

Forum thread: 【Nitan Podcast】你的每周美卡论坛精华播客

flowchart LR
  E[Extract via Nitan MCP] --> B[Optional Gemini brief]
  B --> X[Export Markdown]
  X --> NLM[NotebookLM Audio]
  NLM --> MP3[GitHub Pages MP3]
  MP3 --> RSS[Apple / Spotify / RSS]

Subscribe

The public episode archive includes an in-browser player and source notes. See the 18-episode retrospective and August 2026 baseline for what the project learned before the CastForge growth launch.

Public Compatibility Contract

The automation runs through CastForge, but the subscriber-facing surface of this show must stay stable.

• Feed URL stays https://lifan-builds.github.io/nitan-podcast/feed.xml
• Site URL stays https://lifan-builds.github.io/nitan-podcast/
• Episode URLs stay https://lifan-builds.github.io/nitan-podcast/episodes/weekly_meika_YYYY-Www.mp3
• Episode GUIDs keep the nitan-podcast-YYYY-Www format
• docs/feed.xml and docs/episodes/ remain the publication targets

That is what keeps Spotify and Apple Podcasts from seeing the migration as a broken or new feed.

What This Repo Owns

• Nitan MCP extraction and thread-selection logic
• Category voice and Story Brief editorial rules
• Chinese editorial prompts and podcast/forum copy
• Podcast metadata and RSS identity
• docs/feed.xml and docs/episodes/*.mp3
• Show assets such as cover art
• Validation and tests for this show's public contract

What CastForge Owns

• reusable pipeline stages
• shared integrations for LLM and audio tooling
• framework-level orchestration and contracts

This repository owns the weekly schedule, runner setup, workflow gates, publication, and idempotency policy.

See podcast.yaml for the instance contract exposed to CastForge, and docs/castforge-migration.md for the current integration boundary.

Quick Start

git clone https://github.com/lifan-builds/nitan-podcast.git
cd nitan-podcast
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env

requirements.txt installs the pinned CastForge release and show-specific dependencies. Configure MCP_*, optional GEMINI_API_KEY, and NOTEBOOKLM_* in .env.

Run The Podcast Pipeline

python run_pipeline.py --dated
python run_pipeline.py --dated --skip-briefing
python run_pipeline.py --dated --publish-notebooklm
python run_pipeline.py --dated --generate-post --generate-rss --audio-url "https://..."

# Dry run without MCP
EXTRACTION_FIXTURE_PATH=fixtures/sample_extraction.json python run_pipeline.py --skip-briefing --dated

Stdout prints the primary output path. See python run_pipeline.py --help for the full CLI.

NotebookLM Setup

For --publish-notebooklm using notebooklm-py:

1. pip install -r requirements-integrations.txt && playwright install chromium
2. Run notebooklm login
3. Set NOTEBOOKLM_NOTEBOOK_ID in .env

The login stores ~/.notebooklm/storage_state.json; it is separate from a normal Chrome Google sign-in.

Automation

This repo owns its own weekly schedule and GitHub Actions workflow. Pipeline execution is powered by CastForge, installed from the v0.1.3 tag through requirements.txt. The workflow does not install a second CastForge copy.

The workflow runs on a self-hosted macOS runner with three Monday retry windows (6 AM / 12 PM / 6 PM PST). First success publishes the episode; subsequent triggers skip.

Pipeline phases:

1. MCP extract + optional Gemini briefing + export Markdown
2. NotebookLM Audio Overview + download MP3
3. GitHub Release (backup) + copy MP3 to docs/episodes/
4. Generate RSS feed + forum post draft
5. Commit + push (GitHub Pages auto-deploys)
6. Validate live feed

Manual trigger: Actions tab, weekly-podcast, Run workflow.

Development

pytest tests/ -v
python scripts/validate_feed.py

Useful docs:

• AGENTS.md for architecture and conventions
• docs/castforge-migration.md for the current CastForge integration boundary
• podcast.yaml for the instance contract

Project Structure

License

MIT