# PodMaster

Local podcast transcription, embedding, and semantic search with an MCP server for AI assistant integration. All processing runs on your machine.

## Features

- Local transcription via faster-whisper (CPU or Metal/CUDA)
- Semantic search with sentence embeddings + SQLite vector index
- RSS feed import
- MCP server for Claude Code, Cursor, and other AI assistants
- REST API for GUI integrations
- Full CLI

## Install

```bash
pip install podmaster
pip install "podmaster[mcp]"
pip install "podmaster[api]"
pip install "podmaster[all]"
```

Requires Python 3.11+.

## Quick Start

```bash
podmaster add "https://feeds.example.com/podcast.xml"
podmaster list podcasts
podmaster list episodes
podmaster transcribe --episode <episode-id>
podmaster search "what did they say about AI agents"
podmaster mcp
```

## Models

PodMaster ships a small, vetted set of on-device models. Run `podmaster models` for the full list.

### Transcription (faster-whisper)

| Model | Params | When to use |
|-------|-------:|-------------|
| `base` | 74M | Fast bulk indexing |
| `small` (default) | 244M | Best accuracy/speed balance |
| `medium` | 769M | Higher accuracy, ~3x slower |
| `large-v3` | 1550M | Best accuracy, archival quality |
| `distil-large-v3` | 756M | Near large-v3 at 2x speed (English) |

### Embedding (sentence-transformers)

| Model | Dim | When to use |
|-------|----:|-------------|
| `all-MiniLM-L6-v2` (default) | 384 | Fast, proven baseline |
| `BAAI/bge-small-en-v1.5` | 384 | Higher quality, drop-in swap |
| `all-mpnet-base-v2` | 768 | Highest quality (needs re-index on switch) |

## MCP Integration

```json
{
  "mcpServers": {
    "podmaster": {
      "command": "podmaster",
      "args": ["mcp"]
    }
  }
}
```

Tools: `transcribe_audio`, `search_library`, `get_episode_transcript`, `list_podcasts`, `list_episodes`.

## Development

```bash
git clone https://github.com/summitmktg/podmaster
cd podmaster
pip install -e ".[all,dev]"
pytest tests/
```

## License

MIT