How sweta3304/resolvyn works
Resolvyn is a multichannel AI customer-support product: it understands a request, recalls relevant policy and prior cases, chooses a resolution, takes and verifies actions, and replies only with supported facts. Its team console exposes ticket progress and human controls; staff can teach fixes that become reusable memory. The README documents the product workflow and capabilities; sampled backend code confirms key memory, decision, ticket-state, context, and model layers, but not every route or reply-guard implementation.
Main entry points: base_agent.py Department Agents · decision_engine.py Resolution Decision · api Channel APIs · memory_engine.py Memory Retrieval
Components 22
Customer and Team Experience
- Support Console
frontend/app/ops/overview/page.tsx - Channel APIs
backend/app/api - Voice Interface
backend/app/voice - Email Service
backend/app/services/email_service.py - Analytics Views
frontend/app/analytics/page.tsx
Understand and Decide
- Language Perception
backend/app/perception/perception_service.py - Jev Judgment
backend/app/judgment/jev.py - Resolution Decision
backend/app/decision_engine/decision_engine.py
Resolve and Verify
- Department Agents
backend/app/agents/base_agent.py - Support Tools
backend/app/tools/tool_service.py - Human Actions
backend/app/human_intelligence/human_service.py - Ticket State
backend/app/services/ticket_engine.py - Truth Guard
backend/app/services/conversation.py - Live Case Context
backend/app/context_engine/context_engine.py
Memory and Learning
- Memory Retrieval
backend/app/memory/memory_engine.py - Knowledge Ingestion
backend/app/memory/ingest.py - Learning Service
backend/app/learning/learning_service.py
Runtime Platform
- Realtime Updates
backend/app/realtime.py - Application Database
backend/app/database.py - Model Routing
backend/app/llm/router.py - Local Model Tiers
engine/backend/tiers/model_manager.py
Other components
- Customer
Connections 27
- Customer → Channel APIs: contacts
- Channel APIs → Language Perception: submits request
- Language Perception → Jev Judgment: extracts intent
- Jev Judgment → Memory Retrieval: writes query
backend/app/memory/memory_engine.py - Memory Retrieval → Resolution Decision: supplies matches
backend/app/memory/memory_engine.py - Resolution Decision → Department Agents: chooses action
backend/app/decision_engine/decision_engine.py - Resolution Decision → Human Actions: requests human gate
- Department Agents → Support Tools: runs actions
backend/app/agents/base_agent.py - Support Tools → Department Agents: returns confirmation
backend/app/agents/base_agent.py - Human Actions → Department Agents: approves action
- Department Agents → Truth Guard: provides verified facts
- Truth Guard → Customer: sends verified reply
- Channel APIs → Voice Interface: handles calls
- Customer → Email Service: emails support
- Email Service → Customer: sends summary
- Ticket State → Application Database: persists state
backend/app/services/ticket_engine.py - Ticket State → Realtime Updates: broadcasts changes
backend/app/services/ticket_engine.py - Ticket State → Live Case Context: refreshes case notes
backend/app/context_engine/context_engine.py - Live Case Context → Realtime Updates: pushes summaries
backend/app/context_engine/context_engine.py - Knowledge Ingestion → Memory Retrieval: indexes knowledge
- Human Actions → Learning Service: teaches fix
backend/app/human_intelligence/human_service.py - Learning Service → Memory Retrieval: adds reusable memory
- Support Console → Channel APIs: operates support
- Support Console → Analytics Views: views metrics
- Model Routing → Local Model Tiers: uses local tiers
- Resolution Decision → Model Routing: checks grounding
backend/app/decision_engine/decision_engine.py - Customer → Support Console: team observes progress