How ys-pundir/agentic-support-system works
This repository is a CLI e-commerce support system: a customer message enters the support-agent workflow, which can use policy retrieval, order-data queries, delivery-day calculations, feedback logging, or human escalation before returning a response. The graph preserves these documented branches and the configuration/LLM layer. Source excerpts support the sampled tool and storage connections; broader conversation policies and routing details are documented in the README rather than independently verified here.
Main entry points: workflow.py Support Workflow · config.py Agent Configuration · rag_tool.py Policy Retrieval · sql_tool.py Order Query Agent
Components 17
Conversation
- CLI Entry
src/main.py - Support Workflow
src/agent/workflow.py - Conversation History
Storage/support_agent_conversation.json - Support Response
Policy Answers
- Policy Retrieval
src/tools/rag_tool.py - Policy Vector Store
Database/policy_docs
Order Support
- Order Query Agent
src/tools/sql_tool.py - Delivery Day Count
src/tools/daycount_tool.py - E-commerce Database
Database/ecomm.db
Support Outcomes
- Feedback Logger
src/tools/feedback_tool.py - Feedback Log
Storage/agent_feedbacks.csv - Human Escalation
src/tools/deffered_tool.py - Deferred Cases
Storage/deffered_cases.csv
Agent Foundation
- Agent Configuration
src/config.py - LLM Clients
src/sources/llm.py
Other components
- Customer
- Groq Inference
Connections 21
- Customer → CLI Entry: enters message
- CLI Entry → Support Workflow: starts
- Support Workflow → Customer: returns response
- Support Workflow → Conversation History: appends conversation
- Support Workflow → Policy Retrieval: dispatches
- Support Workflow → Order Query Agent: dispatches
- Support Workflow → Delivery Day Count: dispatches
- Support Workflow → Feedback Logger: dispatches
- Support Workflow → Human Escalation: dispatches
- Policy Retrieval → Policy Vector Store: retrieves context
- Policy Retrieval → Groq Inference: requests completion
- Order Query Agent → E-commerce Database: queries
- Order Query Agent → LLM Clients: uses model
- Support Workflow → LLM Clients: uses model
- LLM Clients → Groq Inference: calls
- Feedback Logger → Feedback Log: appends entries
- Human Escalation → Deferred Cases: writes case
- Agent Configuration → LLM Clients: supplies settings
- Agent Configuration → Support Workflow: supplies prompt
- Agent Configuration → Policy Retrieval: supplies settings
- Agent Configuration → Order Query Agent: supplies settings