How dua-saeed641/resolvyn works
Resolvyn is an AI customer-support system that handles voice, chat, and email, while giving operators a live view and explicit human controls. Its main workflow is to receive a customer issue, perceive and judge it, recall relevant knowledge, choose a resolution, route work to a specialist, perform and verify actions, respond, and learn from the outcome. The repository also includes an Epsilon model-serving engine. The README documents the overall workflow and product capabilities; sampled source confirms selected decision, memory, persistence, and model boundaries, but not every runtime connection.
Main entry points: graph.py Ticket orchestrator · ticket_engine.py Ticket state machine · telephony.py Voice channel · human_service.py Human intelligence
Components 24
Customer channels
- Chat interface
frontend/app/talk/page.tsx - Voice channel
backend/app/api/telephony.py - Email support
backend/app/services/email_service.py - Speech services
backend/app/voice/speech.py
Resolution workflow
- Ticket orchestrator
agents/graph.py - Issue perception
backend/app/perception/perception_service.py - Jev judgment
backend/app/judgment/jev.py - Knowledge and memory
backend/app/memory/memory_engine.py - Resolution decision
backend/app/decision_engine/decision_engine.py - Department routing
agents/swarm_router.py - Specialist agents
backend/app/agents/base_agent.py - Support tools
backend/app/tools/tool_service.py - Verified response
backend/app/services/conversation.py
Human operations
- Human intelligence
backend/app/human_intelligence/human_service.py - Ticket state machine
backend/app/services/ticket_engine.py - Live operations console
frontend/app/ops/overview/page.tsx - Ticket context
backend/app/context_engine/context_engine.py - Learning signals
backend/app/learning/learning_service.py
Runtime platform
- API entry points
backend/app/api/router.py - Application database
backend/app/database.py - Model interface
backend/app/llm/router.py - Tiered model engine
engine/backend/tiers/model_manager.py
Other components
- Customer
- Support operator
Connections 27
- Customer → Chat interface: starts chat
- Customer → Voice channel: calls support
- Customer → Email support: emails support
- Chat interface → API entry points: sends requests
- Voice channel → API entry points: enters service
backend/app/main.py - Email support → Ticket orchestrator: shares workflow
- API entry points → Ticket orchestrator: starts workflow
- Ticket orchestrator → Issue perception: perceives issue
- Ticket orchestrator → Jev judgment: judges issue
- Ticket orchestrator → Knowledge and memory: recalls knowledge
- Ticket orchestrator → Resolution decision: chooses resolution
- Ticket orchestrator → Department routing: routes ticket
- Department routing → Specialist agents: selects specialist
- Specialist agents → Support tools: runs actions
backend/app/agents/base_agent.py - Ticket orchestrator → Human intelligence: pauses for input
- Human intelligence → Ticket state machine: updates ticket
backend/app/human_intelligence/human_service.py - Ticket orchestrator → Ticket state machine: records progress
- Ticket state machine → Application database: persists state
backend/app/services/ticket_engine.py - Ticket state machine → Live operations console: broadcasts updates
backend/app/services/ticket_engine.py - Ticket orchestrator → Verified response: verifies claims
- Verified response → Customer: sends reply
- Ticket orchestrator → Ticket context: refreshes notes
- Ticket context → Live operations console: pushes summaries
backend/app/context_engine/context_engine.py - Human intelligence → Learning signals: teaches system
backend/app/human_intelligence/human_service.py - Ticket orchestrator → Learning signals: learns outcomes
- Voice channel → Speech services: handles speech
- Model interface → Tiered model engine: uses local engine