How sanketwork300-hash/voice_chat-agent works
Cloud Support Agent serves browser chat and voice for diagnosing AWS account issues, answering documentation questions, and applying customer-confirmed fixes; it can escalate unresolved cases. The graph follows the supported request and agent paths, including cloud operations, knowledge retrieval, and persistence, and preserves evaluation and observability as product capabilities. README documents the simulated AWS backend and evaluation workflow; source excerpts do not expose evaluation runner internals, so those links are documented rather than inferred from implementation.
Main entry points: agent.py Support agent · Chat.tsx Chat and voice UI · gateway.py Tool gateway · pipeline.py Voice pipeline
Components 26
Browser experience
- Chat and voice UI
frontend/components/Chat.tsx - Chat socket client
frontend/lib/useChatSocket.ts - Voice client
frontend/lib/useVoiceClient.ts
Support runtime
- Chat API
backend/api/chat.py - Voice API
backend/api/voice.py - Voice service
backend/voice/service.py - Voice pipeline
backend/voice/pipeline.py - Support agent
backend/agent/agent.py - Issue taxonomy
backend/agent/taxonomy.py - Agent prompts
backend/agent/prompts.py - Data redaction
backend/security/redaction.py - Tool gateway
backend/tools/gateway.py - Tool catalog
backend/tools/catalog.py - Debug API
backend/api/debug.py
Cloud and knowledge
- LLM provider
backend/providers/llm/__init__.py - AWS domain tools
backend/tools - Cloud provider
backend/cloud/base.py - Simulated AWS
- Knowledge search
backend/rag/knowledge.py - Embedding provider
backend/providers/embeddings/local.py
State and telemetry
- Postgres state
backend/memory/store.py - Prometheus metrics
backend/metrics.py
Evaluation
- Evaluation API
backend/api/evaluation.py - Scenario evaluation
backend/evaluation
Other components
- Customer
- Human support
Connections 29
- Customer → Chat and voice UI: uses
README.md - Chat and voice UI → Chat socket client: sends chat
README.md - Chat socket client → Chat API: WebSocket turns
README.md - Chat API → Support agent: handles turns
backend/api/chat.py - Chat and voice UI → Voice client: starts calls
README.md - Voice client → Voice API: WebRTC offer
README.md - Voice API → Voice service: routes calls
backend/api/voice.py - Voice service → Voice pipeline: builds worker
backend/voice/service.py - Voice pipeline → Support agent: submits turns
backend/voice/pipeline.py - Support agent → Issue taxonomy: classifies and resolves
backend/agent/agent.py - Support agent → Agent prompts: uses prompts
backend/agent/agent.py - Support agent → LLM provider: requests generation
backend/agent/agent.py - Support agent → Tool gateway: dispatches tools
backend/agent/agent.py - Tool gateway → Tool catalog: loads definitions
backend/tools/gateway.py - Tool gateway → AWS domain tools: executes calls
backend/tools/gateway.py - AWS domain tools → Cloud provider: calls operations
backend/tools/catalog.py - Cloud provider → Simulated AWS: AWS API
README.md - Support agent → Knowledge search: searches docs
backend/agent/agent.py - Knowledge search → Embedding provider: embeds queries
backend/rag/knowledge.py - Support agent → Postgres state: stores conversation
backend/agent/agent.py - Tool gateway → Postgres state: records audit
backend/tools/gateway.py - Support agent → Data redaction: redacts data
backend/agent/agent.py - Support agent → Prometheus metrics: records usage
backend/agent/agent.py - Voice pipeline → Prometheus metrics: records latency
backend/voice/pipeline.py - Chat and voice UI → Debug API: requests diagnostics
README.md - Chat and voice UI → Evaluation API: views runs
README.md - Evaluation API → Scenario evaluation: serves evaluations
README.md - Scenario evaluation → Postgres state: stores results
README.md - Support agent → Human support: escalates
README.md