MinusPod Documentation
Full documentation for MinusPod. Start with the project README for a quick install, then come here for the details.
Contents
Section titled “Contents”- Contributing - development prerequisites, setup, builds, tests, and pull requests
- Installation - requirements, Whisper/Parakeet sizing, resource measurement, quick start, CPU-only image, Intel hybrid CPU tuning
- Environment Variables - every env var, grouped by how often you touch it
- LLM Providers - Claude Code wrapper, Ollama (local/cloud), OpenRouter, recommended models, pricing
- Transcription - Parakeet (recommended) and faster-whisper backends, choosing between them, GPU compute types, remote APIs (whisper.cpp, Groq, OpenAI), language, timeouts
- Finding Feeds & Usage - podcast search, finding RSS feeds, Audiobookshelf
- Web Interface - the management UI, ad editor workflow, screenshots
- How It Works & Detection Pipeline - the transcription -> detection -> cut pipeline, verification pass, sliding windows, processing queue, validation, pattern learning, audio analysis
- Episode Processing Workflows - visual map of the pipeline, the five processing modes, re-runs, and failure paths
- Videos - a UI tour, an animated walk through the pipeline, and one short video per feature, in portrait and landscape cuts, plus how to rebuild and re-record them
- Configuration & Experiments - settings, per-stage LLM tuning, VAD gap detector, provider keys, ad reviewer, community patterns, scheduled backups
- Security, Storage & Custom Assets - remote access, login lockout, backups, custom ad markers
- Audio Cue Detection - per-feed cue templates, the find-audio-cues scan, cue types, settings, and tuning
- Pattern Cleanup - scheduled or manual LLM review and statistics checks: trims, splits, renames, retirements, false-positive or contamination flags, approve/reject/undo, API
- Upstream Transcript Differential - diffing the Whisper transcript against the publisher’s ad-free
podcast:transcript, how gaps corroborate or hold, settings - System One - native TypeSafe and compatible providers, supported phases, profiles, migration, and statistics
- Failover - standby LLM provider and transcriber, triggers, health probes, recovery, manual control, API
- Download History - the History Downloads tab: what is recorded, filters, optional geo lookup, retention, privacy, API
- Local Feeds - build a feed from your own audio files: creation, uploads, bulk archive import, naming scheme, JSON sidecar schema
- YouTube Feeds - turn a playlist or channel into a feed: SponsorBlock-first processing, captions, yt-dlp lifecycle, cookies
- Configuration Import / Export - export and restore global settings and selected feed configuration
- Intel GPU Transcription (OpenVINO) - offload Whisper to an Intel GPU via OpenVINO Model Server
- API & Webhooks - REST endpoints, webhook events, payload templates, signing
- Podcasting 2.0 - what MinusPod emits, regenerates, and deliberately strips from the Podcast Namespace, and why
- Deployment Runbook - operational runbook
- Releasing & Channels - stable vs edge channels, how releases are tagged and promoted
- Container Security Updates - pinned security patches for media libraries and base packages in the GPU and CPU images
- Glossary - every term the app uses, defined and linked to the docs that cover it
- Design Specs - design documents behind larger features, cited by tests that pin a design decision