How myleslandais/jupyter works
The README identifies ASR model evaluation as the product’s primary goal: evaluate models against datasets, calculate quality and performance metrics, and store comparisons for a leaderboard. The supplied tree does not include the documented adapter or metrics modules, so those capabilities are represented as README-documented subsystems rather than assigned invented source paths. The visible source supports dataset-handler discovery, PostgreSQL result storage, and RunPod management; connections among these and the evaluation workflow are omitted where no call is shown.
Main entry points: client.py RunPod client · manager.py Dataset manager · postgres_storage.py Result storage · runpod_manager.py RunPod manager
Components 14
ASR evaluation
- Evaluation framework
- Model adapters
- Metrics engine
- ASR models
Datasets and results
- Dataset manager
src/datasets/manager.py - Dataset handlers
src/datasets/handlers - Result storage
src/providers/asr/storage/postgres_storage.py
Compute infrastructure
- RunPod manager
src/infra/runpod_manager.py - RunPod client
src/runpod/client.py - VibeVoice manager
src/infra/vibe_voice_manager.py - Comfy client
src/comfy/client.py
Other components
- Evaluator
- PostgreSQL
- RunPod
Connections 9
- Evaluator → Evaluation framework: requests evaluation
- Evaluation framework → Model adapters: uses adapters
- Model adapters → ASR models: transcribes audio
- Evaluation framework → Metrics engine: calculates metrics
- Dataset manager → Dataset handlers: discovers handlers
src/datasets/manager.py - Result storage → PostgreSQL: reads and writes
src/providers/asr/storage/postgres_storage.py - RunPod manager → RunPod client: uses client
src/infra/runpod_manager.py - RunPod client → RunPod: runs commands
src/runpod/client.py - VibeVoice manager → Comfy client: imports client
src/infra/vibe_voice_manager.py