Skip to content

Jayyk09/AfterThought

BranchesTags

Repository files navigation

AfterThought

AI-Powered Podcast & YouTube Summarization

AfterThought automatically extracts transcripts from Apple Podcasts and YouTube videos, generates AI-powered summaries using Google Gemini, and outputs beautifully formatted Obsidian-compatible markdown files optimized for graph view.

Features

  • 📊 Smart Episode Discovery: Automatically finds recently listened podcast episodes from Apple Podcasts
  • 🎯 Fuzzy Channel Matching: Filter by podcast channel name with intelligent fuzzy matching
  • 🎬 YouTube Support: Summarize YouTube videos with available transcripts (no API key needed)
  • 🤖 AI Summarization: Leverages Google Gemini API for high-quality summaries
  • 📝 Obsidian Integration: Creates markdown files optimized for graph view with wiki links and tags
  • 🔍 Tracking: Avoids re-processing content you've already summarized
  • ⚡ Incremental Processing: Run regularly to keep your notes up-to-date
  • 🔄 Efficient: Word-level TTML transcript parsing with speaker identification

Requirements

  • Python 3.8+
  • macOS (for Apple Podcasts database access)
  • Google Gemini API key (Get one free here)
  • Obsidian (optional, but recommended for viewing summaries)

Installation

Quick Install (Recommended)

Option 1: Using pipx (global access, isolated environment)

git clone https://github.com/Jayyk09/AfterThought.git
cd AfterThought
brew install pipx
pipx install .

Option 2: Using pip (editable install)

git clone https://github.com/Jayyk09/AfterThought.git
cd AfterThought
pip install -e .

Option 3: Traditional venv (manual activation)

git clone https://github.com/Jayyk09/AfterThought.git
cd AfterThought
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

📖 See INSTALL.md for detailed installation instructions and shell wrapper setup.

Configure

cp .env.example .env
# Edit .env with your settings

Required settings:

  • GEMINI_API_KEY - Get free key at Google AI Studio
  • OBSIDIAN_OUTPUT_PATH - Path to your Obsidian vault

Apple Podcasts paths are auto-detected!

Usage

Basic Usage

Summarize episodes played in the last 7 days:

python -m afterthought

Filter by Podcast Channel

Use fuzzy matching to filter by channel name:

python -m afterthought --channel "All-In"
python -m afterthought -c "Lex Fridman"

Custom Date Range

Summarize episodes from the last 30 days:

python -m afterthought --days 30
python -m afterthought -d 14

Force Re-processing

Re-process already summarized episodes:

python -m afterthought --force
python -m afterthought -c "All-In" --force

Dry Run

Preview what would be processed without making changes:

python -m afterthought --dry-run

Show Statistics

View processing statistics from your tracking database:

python -m afterthought --stats

Verbose Output

Enable detailed logging:

python -m afterthought --verbose

Auto-Fetch Missing Transcripts

Automatically trigger playback in Podcasts app to download missing transcripts:

python -m afterthought --fetch-missing
python -m afterthought -c "History of Rome" --fetch-missing

When enabled, AfterThought will:

  1. Detect episodes without transcripts
  2. Open the episode in Apple Podcasts app
  3. Play it briefly to trigger transcript download
  4. Wait 10 seconds for download
  5. Retry processing the episode

This is useful for episodes you've played on iOS that don't have transcripts cached on your Mac yet.

Combined Options

python -m afterthought -c "All-In" -d 30 -f -v
python -m afterthought --channel "History" --fetch-missing --verbose

CLI Options

Option Short Description
--channel -c Fuzzy match podcast channel name
--days -d Episodes played in last N days (default: 7)
--force -f Re-process already summarized episodes
--fetch-missing Auto-fetch missing transcripts by triggering playback
--youtube -y Summarize a YouTube video by URL
--dry-run Show what would be processed without executing
--verbose -v Enable verbose output
--stats Show processing statistics and exit
--help -h Show help message

Output Format

AfterThought creates markdown files organized by podcast channel:

~/Documents/Obsidian/Podcasts/
├── All-In Podcast/
│   ├── E150 Tech Trends 2026.md
│   └── E151 AI Regulation Debate.md
├── Lex Fridman Podcast/
│   ├── #123 - Sam Altman OpenAI.md
│   └── #124 - Andrew Huberman Neuroscience.md
└── ...

Markdown File Structure

Each episode summary is optimized for Obsidian's graph view with extensive linking and tagging:

Frontmatter (YAML):

---
type: podcast-summary
cssclass: podcast
title: "The Punic Wars"
aliases:
  - "The Punic Wars"
podcast: "[[The History of Rome]]"
author: "Mike Duncan"
date: 2007-10-15
listened: 2026-01-07
duration: "18:45"
tags:
  - podcast
  - the-history-of-rome
transcript_available: true
ai_model: "gemini-2.0-flash-exp"
---

Summary Content (Obsidian-Optimized):

  • Wiki Links: All concepts, people, places, and events wrapped in [[double brackets]]

    • Examples: [[Roman Republic]], [[Hannibal]], [[Battle of Cannae]]
    • Creates interconnected nodes in graph view

  • Tags: Categorization with #hashtags

    • Periods: #AncientRome, #LatinAmerica
    • Themes: #MilitaryHistory, #PoliticalPhilosophy
    • Regions: #Mediterranean, #Europe

  • Mermaid Diagrams: Visual timelines and relationships

    timeline
    title Punic Wars Timeline
    264-241 BC : First Punic War
               : Rome vs Carthage naval battles
    218-201 BC : Second Punic War
               : Hannibal crosses Alps
    149-146 BC : Third Punic War
               : Destruction of Carthage
    Loading

  • Concise Structure:

    • Summary: 2-3 bullet core narrative
    • Historical Context: Background with nested relationships
    • Key Events: Chronological developments with heavy linking
    • Notable Quotes: 2-3 significant quotes

No fluff. Dense information. Maximum graph connectivity.

Architecture

Project Structure

AfterThought/
├── config.py                      # Pydantic configuration
├── afterthought/
│   ├── cli.py                    # CLI interface (Click)
│   ├── db/
│   │   ├── podcast_db.py         # Apple Podcasts SQLite queries
│   │   └── tracking_db.py        # Processed episodes tracking
│   ├── parsers/
│   │   └── ttml_parser.py        # TTML XML transcript parsing
│   ├── summarizer/
│   │   └── gemini_client.py      # Google Gemini API client
│   ├── output/
│   │   └── markdown_writer.py    # Obsidian markdown generation
│   └── utils/
│       ├── fuzzy_match.py        # Fuzzy string matching
│       ├── date_utils.py         # Date/time utilities
│       └── logging_config.py     # Logging setup

Data Flow

Apple Podcasts DB → Filter (date/channel) → Check Tracking DB
                                                     ↓
                                              Not processed?
                                                     ↓
                                        Load TTML → Parse → Summarize (Gemini)
                                                                ↓
                                                    Write Markdown → Update Tracking

Key Technologies

  • Pydantic: Type-safe configuration with validation
  • Click: Command-line interface framework
  • Google Gemini: AI summarization (gemini-2.0-flash-exp)
  • thefuzz: Fuzzy string matching for channel names
  • SQLite: Apple Podcasts database (read-only) + tracking database

Configuration

Environment Variables

All configuration is managed via .env file:

Variable Required Default Description
GEMINI_API_KEY Yes - Google Gemini API key
OBSIDIAN_OUTPUT_PATH Yes - Output directory for markdown files
APPLE_PODCASTS_DB_PATH No Auto-detected Path to MTLibrary.sqlite
TTML_CACHE_PATH No Auto-detected Path to TTML transcript cache
TRACKING_DB_PATH No ~/.afterthought/tracking.db Tracking database path
GEMINI_MODEL No gemini-2.0-flash-exp Gemini model to use
DEFAULT_DAYS_FILTER No 7 Default days to look back
PRESERVE_SPEAKERS No true Preserve speaker IDs in transcripts

Apple Podcasts Database

AfterThought automatically detects the Apple Podcasts database location:

~/Library/Group Containers/[ID].groups.com.apple.podcasts/Documents/MTLibrary.sqlite

The [ID] varies by system but is auto-detected. The database is opened in read-only mode to ensure safety.

Transcript Cache

TTML transcript files are cached by Apple Podcasts at:

~/Library/Group Containers/[ID].groups.com.apple.podcasts/Library/Cache/Assets/TTML/

Note: Only ~64% of episodes have transcripts available. Episodes without transcripts will be skipped.

Troubleshooting

"Could not auto-detect Apple Podcasts database"

Solution: Manually set the path in .env:

APPLE_PODCASTS_DB_PATH=~/Library/Group\ Containers/243LU875E5.groups.com.apple.podcasts/Documents/MTLibrary.sqlite

Find your actual path:

ls ~/Library/Group\ Containers/ | grep podcasts

"Invalid API key"

Solution:

  1. Get a new API key from Google AI Studio
  2. Update .env file with the new key
  3. Ensure there are no quotes or spaces around the key

"No transcript available"

Cause: Not all podcast episodes have transcripts. About 36% of episodes lack transcript data.

Solution: This is normal. The tool will skip these episodes and continue processing others.

"Transcript file not found"

Cause: The transcript hasn't been downloaded by Apple Podcasts yet.

Solution: Play the episode for a few seconds in Apple Podcasts to trigger transcript download, then run AfterThought again.

"Gemini API rate limit exceeded"

Solution: AfterThought includes automatic retry with exponential backoff. If rate limits persist:

  • Wait a few minutes between runs
  • Process fewer episodes at once (use --days 1)
  • Check your API quota at Google AI Studio

Permission Errors

Solution: Ensure AfterThought has permission to:

  • Read Apple Podcasts database: ~/Library/Group Containers/...
  • Write to Obsidian directory: Your OBSIDIAN_OUTPUT_PATH
  • Write tracking database: ~/.afterthought/

Development

Running Tests

# Install dev dependencies
pip install pytest pytest-cov

# Run tests
pytest tests/

Code Structure

  • Each module is independent and testable
  • Context managers for database connections
  • Type hints throughout for better IDE support
  • Pydantic for configuration validation

Adding Custom Prompts

Edit afterthought/summarizer/gemini_client.py and modify DEFAULT_PROMPT_TEMPLATE to customize the summarization style.

FAQ

Q: Does this work with other podcast apps?

A: Currently only Apple Podcasts is supported, as it provides word-level TTML transcripts. Support for other apps would require different transcript sources.

Q: Can I use a different AI model?

A: Yes! Set GEMINI_MODEL in your .env to any supported Gemini model:

  • gemini-2.0-flash-exp (default, fast and cheap)
  • gemini-1.5-pro (more capable, higher cost)
  • gemini-1.5-flash (balanced)

Q: How much does Gemini API cost?

A: Gemini 2.0 Flash has a generous free tier:

  • 1,500 requests per day (free)
  • ~4M tokens per day (free)

For most users, AfterThought stays within free limits.

Q: Can I export to formats other than Markdown?

A: Currently only Markdown is supported. The Markdown files are plain text and can be easily converted to other formats using tools like Pandoc.

Q: Does this modify my Apple Podcasts data?

A: No. AfterThought opens the Apple Podcasts database in read-only mode. It never modifies your podcast library.

Contributing

Contributions are welcome! Please:

  1. Fork the repository
  2. Create a feature branch (git checkout -b feature/amazing-feature)
  3. Commit your changes (git commit -m 'Add amazing feature')
  4. Push to the branch (git push origin feature/amazing-feature)
  5. Open a Pull Request

License

MIT License - see LICENSE file for details

Acknowledgments

  • Apple Podcasts for providing high-quality TTML transcripts
  • Google Gemini for powerful AI summarization
  • The Python community for excellent libraries

Support

Changelog

v0.1.0 (2026-01-07)

  • Initial release
  • Core functionality: discovery, parsing, summarization, markdown output
  • Support for fuzzy channel matching
  • Tracking database to avoid re-processing
  • CLI with multiple options
  • Comprehensive documentation

YouTube Support

AfterThought also supports summarizing YouTube videos with available transcripts!

Summarize YouTube Videos

# Summarize a YouTube video by URL
afterthought --youtube "https://www.youtube.com/watch?v=VIDEO_ID"
afterthought -y "https://youtu.be/VIDEO_ID"

How It Works

  • Fetches existing YouTube captions (auto-generated or manual)
  • Works on ~70% of videos (those with captions enabled)
  • No API key needed for transcripts (uses YouTube's public caption endpoint)
  • Generates Obsidian-optimized summaries with wiki links and tags
  • Saves to YouTube/Channel Name/ folders in your Obsidian vault
  • Emphasizes visual diagrams for educational content

Supported URL Formats

afterthought -y "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
afterthought -y "https://youtu.be/dQw4w9WgXcQ"
afterthought -y "dQw4w9WgXcQ"  # Just the video ID

When It Works

  • ✅ Educational/Tutorial videos (~95% have captions)
  • ✅ Podcasts uploaded to YouTube (~85%)
  • ✅ Talks/Interviews (~90%)
  • ✅ News/Documentary content (~95%)

When It Fails

  • ❌ Captions disabled by creator
  • ❌ Private/age-restricted videos
  • ❌ Very new videos (captions not processed yet)
  • ❌ Some music videos (~40% have captions)

YouTube-Specific Options

# Dry run to check if transcript is available
afterthought -y "VIDEO_URL" --dry-run

# Force re-process a video
afterthought -y "VIDEO_URL" --force

# Verbose output with token usage
afterthought -y "VIDEO_URL" --verbose

Output Format

YouTube summaries are saved in YouTube/Channel Name/ folders:

~/Documents/Obsidian/Podcasts/
├── All-In Podcast/
│   └── E150 Tech Trends 2026.md
├── YouTube/
│   ├── 3Blue1Brown/
│   │   └── Linear transformations and matrices.md
│   ├── Veritasium/
│   │   └── The Physics of Black Holes.md
│   └── ...

Each summary includes:

  • Wiki links for concepts, people, technologies
  • Tags for topics and domains
  • Mermaid diagrams (emphasized for educational content)
  • Concise bullet-point format optimized for Obsidian graph view
  • Only summarizes transcript content (no extra context added)

Built with ❤️ for podcast enthusiasts who love taking notes

About

No description, website, or topics provided.

Resources

Readme
Activity

Stars

1 star

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •  

Languages