How deepakchaudharigit/npcl-asterisk-ari-assistant works
This repository is a Gemini-powered voice assistant with standalone interaction and an Asterisk telephony mode. The principal documented workflow is an incoming call handled by the real-time ARI server, which tracks a session and sets up external-media audio; audio and conversation are exchanged with Gemini Live for low-latency, interruptible voice responses. Separate audio, AI, tools, and security components provide processing and supporting capabilities. The README also documents standalone speech recognition, Gemini responses, and TTS. Some links reflect README-level capabilities; unsampled implementations and wiring are not inferred.
Main entry points: websocket_gemini_client.py WebSocket Gemini client · realtime_ari_handler.py Realtime ARI handler · security_manager.py Security manager · main.py Standalone assistant
Components 26
Entry points
- Realtime server
src/run_realtime_server.py - Standalone assistant
src/main.py
Telephony and sessions
- Realtime ARI handler
src/voice_assistant/telephony/realtime_ari_handler.py - External media
src/voice_assistant/telephony/external_media_handler.py - Session manager
src/voice_assistant/core/session_manager.py - Legacy ARI handler
src/ari_handler.py - RTP streaming
src/voice_assistant/telephony/rtp_streaming_handler.py
Audio processing
- Realtime audio processor
src/voice_assistant/audio/realtime_audio_processor.py - Advanced audio processor
src/voice_assistant/audio/advanced_audio_processor.py - Speech recognition
src/voice_assistant/audio/speech_recognition.py - Text to speech
src/voice_assistant/audio/text_to_speech.py
AI and tools
- Gemini Live client
src/voice_assistant/ai/gemini_live_client.py - WebSocket Gemini client
src/voice_assistant/ai/websocket_gemini_client.py - Gemini client
src/voice_assistant/ai/gemini_client.py - Function calling
src/voice_assistant/ai/function_calling.py - Assistant prompts
src/voice_assistant/ai/npcl_prompts.py - Weather tool
src/voice_assistant/tools/weather_tool.py
Security
- Security manager
src/voice_assistant/security/security_manager.py - Authentication
src/voice_assistant/security/auth_manager.py - Input validation
src/voice_assistant/security/input_validator.py - Rate limiting
src/voice_assistant/security/rate_limiter.py - Audit logging
src/voice_assistant/security/audit_logger.py
Other components
- Caller
- Standalone user
- Asterisk PBX
- Google Gemini
Connections 25
- Caller → Asterisk PBX: calls
README.md - Asterisk PBX → Realtime ARI handler: sends events
README.md - Realtime server → Realtime ARI handler: starts
src/run_realtime_server.py - Realtime ARI handler → Session manager: creates and updates
src/voice_assistant/telephony/realtime_ari_handler.py - Realtime ARI handler → External media: starts stream
src/voice_assistant/telephony/realtime_ari_handler.py - Realtime ARI handler → Gemini Live client: uses
src/voice_assistant/telephony/realtime_ari_handler.py - External media → Realtime audio processor: processes audio
src/voice_assistant/telephony/external_media_handler.py - External media → Gemini Live client: exchanges audio
src/voice_assistant/telephony/external_media_handler.py - External media → Session manager: tracks session
src/voice_assistant/telephony/external_media_handler.py - Gemini Live client → Google Gemini: streams conversation
README.md - Standalone user → Standalone assistant: speaks to
README.md - Standalone assistant → Gemini client: requests responses
src/main.py - Standalone assistant → Speech recognition: transcribes speech
README.md - Standalone assistant → Text to speech: speaks response
README.md - WebSocket Gemini client → Advanced audio processor: processes audio
src/voice_assistant/ai/websocket_gemini_client.py - WebSocket Gemini client → Function calling: dispatches calls
src/voice_assistant/ai/websocket_gemini_client.py - WebSocket Gemini client → Assistant prompts: uses
src/voice_assistant/ai/websocket_gemini_client.py - WebSocket Gemini client → Weather tool: registers tool
src/voice_assistant/ai/websocket_gemini_client.py - WebSocket Gemini client → RTP streaming: streams audio
src/voice_assistant/ai/websocket_gemini_client.py - Gemini Live client → Realtime audio processor: uses
src/voice_assistant/ai/gemini_live_client.py - Security manager → Authentication: authenticates
src/voice_assistant/security/security_manager.py - Security manager → Rate limiting: checks limits
src/voice_assistant/security/security_manager.py - Security manager → Input validation: validates input
src/voice_assistant/security/security_manager.py - Security manager → Audit logging: records events
src/voice_assistant/security/security_manager.py - Caller → Legacy ARI handler: uses legacy mode
README.md