How ronincodex/voice-agent works
Voice Agent is a multilingual phone-calling product with a Next.js dashboard and a FastAPI backend. Its main workflow runs from an outbound-call request or inbound phone call, through Vobiz webhooks and a real-time Pipecat conversation, to persisted call data, recordings, and a post-call summary. Separate dashboard, agent-configuration, compliance, and observability capabilities support that workflow. The backend excerpt and SOURCE INDEX show its main orchestration and dependencies; some subsystem links are documented only at a high level in the README.
Main entry points: server.py FastAPI server · agent_pipeline.py Voice pipeline · summarizer.py Post-call summary · supabase_client.py Call and transcript store
Components 33
Dashboard
- Dashboard pages
frontend/app - Call views and forms
frontend/components - API client
frontend/lib/api.ts
API and telephony
- FastAPI server
src/voice_agent/api/server.py - API schemas
src/voice_agent/api/schemas.py - Vobiz adapter
src/voice_agent/telephony/client.py - Call tracking
src/voice_agent/telephony/call_state.py
Conversation pipeline
- Voice pipeline
src/voice_agent/pipeline/agent_pipeline.py - Conversation stages
src/voice_agent/pipeline/nodes.py - Language prompts
src/voice_agent/config/languages.py
Persistence and summaries
- Call and transcript store
src/voice_agent/db/supabase_client.py - Agent configuration
src/voice_agent/db/agent_config.py - Session state store
src/voice_agent/state/redis_store.py - Recording storage
src/voice_agent/storage/r2_client.py - Post-call summary
src/voice_agent/postcall/summarizer.py - Summary fallback
src/voice_agent/observability/fallback.py
Safety and operations
- Input guardrail
src/voice_agent/pipeline/guardrail.py - Conversation validators
src/voice_agent/pipeline/validators.py - Calling window
src/voice_agent/compliance/calling_window.py - PII masking
src/voice_agent/compliance/pii.py - Compliance audit
src/voice_agent/db/audit.py - App settings
src/voice_agent/config/settings.py - Pipeline metrics
src/voice_agent/observability/metrics.py - Idempotency
src/voice_agent/observability/idempotency.py - Retry policies
src/voice_agent/observability/retry.py
Other components
- Dashboard user
- Phone caller
- Vobiz
- Sarvam AI
- Groq
- Supabase
- Upstash Redis
- Cloudflare R2
Connections 36
- Dashboard user → Dashboard pages: uses
README.md - Dashboard pages → Call views and forms: renders
- Dashboard pages → API client: requests data
- API client → FastAPI server: calls API
README.md - Phone caller → Vobiz: calls
README.md - Vobiz → FastAPI server: webhooks
README.md - FastAPI server → Vobiz adapter: initiates calls
src/voice_agent/api/server.py - FastAPI server → Call tracking: tracks calls
src/voice_agent/api/server.py - FastAPI server → Voice pipeline: starts conversation
src/voice_agent/api/server.py - FastAPI server → API schemas: validates API data
src/voice_agent/api/server.py - Voice pipeline → Conversation stages: initializes flow
README.md - Voice pipeline → Language prompts: loads language config
src/voice_agent/pipeline/agent_pipeline.py - Voice pipeline → Session state store: uses session state
src/voice_agent/pipeline/agent_pipeline.py - Voice pipeline → Call and transcript store: uses call data
src/voice_agent/pipeline/agent_pipeline.py - Voice pipeline → Compliance audit: records events
src/voice_agent/pipeline/agent_pipeline.py - Voice pipeline → Pipeline metrics: collects metrics
src/voice_agent/pipeline/agent_pipeline.py - Voice pipeline → Input guardrail: checks input
src/voice_agent/pipeline/agent_pipeline.py - Voice pipeline → Conversation validators: validates turns
src/voice_agent/pipeline/agent_pipeline.py - Voice pipeline → Sarvam AI: uses speech AI
README.md - FastAPI server → Calling window: enforces hours
src/voice_agent/api/server.py - FastAPI server → PII masking: masks transcripts
src/voice_agent/api/server.py - FastAPI server → Agent configuration: reads configuration
src/voice_agent/api/server.py - FastAPI server → Call and transcript store: reads and writes calls
src/voice_agent/api/server.py - FastAPI server → Recording storage: stores recordings
src/voice_agent/api/server.py - Recording storage → Cloudflare R2: uploads recordings
README.md - FastAPI server → Post-call summary: generates summary
src/voice_agent/api/server.py - Post-call summary → Groq: fallback summary
src/voice_agent/postcall/summarizer.py - Post-call summary → Sarvam AI: summarizes transcript
src/voice_agent/postcall/summarizer.py - Post-call summary → Summary fallback: uses fallback
src/voice_agent/postcall/summarizer.py - Call and transcript store → Supabase: persists records
README.md - Session state store → Upstash Redis: stores sessions
README.md - FastAPI server → App settings: loads configuration
src/voice_agent/api/server.py - Vobiz adapter → Retry policies: retries requests
src/voice_agent/telephony/client.py - Call and transcript store → Retry policies: retries database I/O
src/voice_agent/db/supabase_client.py - Recording storage → Retry policies: retries storage I/O
src/voice_agent/storage/r2_client.py - Conversation stages → Idempotency: uses idempotency
src/voice_agent/pipeline/nodes.py