How rbbobby/audio-text-manager-desktop works
Audio Text Manager is a local Wails desktop app for audio/video transcription and summarization. The main workflow is file selection, audio normalization, queued background processing, Whisper transcription (optionally with speaker labeling), and llama.cpp summarization, with results and job history stored locally. Settings, runtime checks, export, and reprocessing are additional product capabilities. Some connections are documented in the README or visible through the source index; unsampled frontend and subsystem internals are represented only at paths supported by the tree.
Main entry points: pipeline.go Job worker · app.go Wails app API · models.go ASR models · sidecar.go Runtime discovery
Components 18
Desktop interface
- Desktop UI
frontend/dist/app.js - Wails app API
app.go - Settings
internal/config/config.go
Job workflow
- Job worker
internal/pipeline/pipeline.go - Job history
internal/jobs/store.go - Audio preparation
internal/media/media.go - App data location
internal/appdir/appdir.go
Speech processing
- ASR models
internal/asr/models.go - Whisper transcription
internal/asr/whisper.go - Transcript segments
internal/asr/segments.go - Speaker labeling
internal/asr/diarize.go - Build flavor options
internal/flavor/flavor.go
Local runtime
- Summary generation
internal/summary/ollama.go - Summary model
internal/llm/model.go - llama.cpp server
internal/llm/server.go - Runtime discovery
internal/sidecar/sidecar.go - Model downloads
internal/download/download.go
Other components
- Desktop user
Connections 25
- Desktop user → Desktop UI: uses
README.md - Desktop UI → Wails app API: invokes
README.md - Wails app API → Job worker: starts
app.go - Wails app API → Job history: manages jobs
app.go - Wails app API → Audio preparation: prepares uploads
app.go - Wails app API → Settings: loads settings
app.go - Wails app API → App data location: initializes data
app.go - Job worker → Job history: claims and updates
internal/pipeline/pipeline.go - Job worker → Settings: reads snapshot
internal/pipeline/pipeline.go - Job worker → ASR models: ensures weights
internal/pipeline/pipeline.go - Job worker → Whisper transcription: transcribes audio
internal/pipeline/pipeline.go - Whisper transcription → Transcript segments: parses output
internal/asr/whisper.go - Whisper transcription → Speaker labeling: assigns speakers
internal/asr/whisper.go - Job worker → Build flavor options: selects diarization
internal/pipeline/pipeline.go - Job worker → Summary model: ensures model
internal/pipeline/pipeline.go - Job worker → llama.cpp server: starts server
internal/pipeline/pipeline.go - Job worker → Summary generation: requests summary
internal/pipeline/pipeline.go - llama.cpp server → Runtime discovery: prepares process
internal/llm/server.go - Summary model → Runtime discovery: searches bundled model
internal/llm/model.go - Summary model → Model downloads: downloads missing model
internal/llm/model.go - ASR models → Build flavor options: checks preset
internal/asr/models.go - ASR models → Runtime discovery: searches bundled weights
internal/asr/models.go - ASR models → Model downloads: downloads missing weights
internal/asr/models.go - Audio preparation → Runtime discovery: resolves tools
internal/media/media.go - Job history → App data location: stores SQLite data