How khushi-070906/dwani works
LDST/DwaniLive is a local-network speech-translation host: a presenter streams microphone audio, the server segments and transcribes it, translates for each subscribed language, and broadcasts captions to attendees. The README also documents reverse-direction Q&A, optional glossary adaptation and semantic caching, accessibility settings, and a desktop app. The graph follows the README and source-index references; some implementation excerpts are partial or unavailable, so uncertain wiring is omitted.
Main entry points: server.py Session server · gui.py Desktop app · decision_engine.py Translation decision · pipeline.py Audio segmentation
Components 21
Session hosting
- Session server
server.py - Session state
session.py - Caption delivery
server.py
Speech translation
- Audio segmentation
pipeline.py - Whisper ASR
backends.py - NLLB translation
backends.py - NLLB tokenizer
nllb_tokenizer.py
Translation extensions
- Semantic cache
translation_cache.py - Translation decision
decision_engine.py - Glossary adaptation
glossary.py - Dynamic glossary
dynamic_glossary.py - Persistent memory
persistent_memory.py
Attendee interaction
- Attendee client
- Reverse Q&A
qa_pipeline.py - Accessibility settings
accessibility.py
Desktop and licensing
- Desktop app
gui.py - Runtime environment
appenv.py - License enforcement
licensing.py - License service
license_server.py
Other components
- Presenter
- Attendees
Connections 24
- Presenter → Session server: starts session
README.md - Attendees → Session server: join locally
README.md - Session server → Session state: manages session
server.py - Presenter → Audio segmentation: streams audio
README.md - Session server → Audio segmentation: segments audio
README.md - Audio segmentation → Whisper ASR: transcribes segments
README.md - Session server → NLLB translation: translates languages
README.md - NLLB translation → NLLB tokenizer: uses tokenizer
backends.py - Session server → Caption delivery: broadcasts captions
README.md - Caption delivery → Attendee client: streams captions
README.md - Attendees → Attendee client: selects language
README.md - Session server → Reverse Q&A: routes questions
server.py - Attendees → Reverse Q&A: submits question
README.md - Session server → Accessibility settings: reports settings
server.py - Session server → Translation decision: configures decisions
decision_engine.py - Translation decision → Semantic cache: checks translations
decision_engine.py - Translation decision → Glossary adaptation: checks terminology
decision_engine.py - Session server → Glossary adaptation: configures glossary
server.py - Session server → Dynamic glossary: configures adaptation
server.py - Session server → Persistent memory: configures memory
server.py - Desktop app → Session server: launches server
gui.py - Desktop app → Runtime environment: uses runtime paths
appenv.py - Desktop app → License enforcement: checks license
gui.py - License service → License enforcement: issues licenses
license_server.py