How victoromoboye/ai-powered-customer-care-chatbot works
Project Type and Purpose
• VicaBot is a lightweight, Streamlit-based full-stack chatbot application.
• Purpose: Receive user messages via a web UI, apply rule-based logic, or defer to an external NLP model (BlenderBot via HuggingFace), then display responses and maintain conversation context.
File Structure
• Top-level files only:
• README.md – Project description and methodology
• VicBot.py – Contains all code: Streamlit UI, rule engine, model API calls, and conversation state
• No separate folders for “frontend” or “backend,” but conceptually it’s a single-module monolith.
README for Key Components
• Streamlit – Web UI layer for chat interface
• Rule-Based Intent Engine – Python logic in VicBot.py
• BlenderBot 400M (via HuggingFace) – External AI service for fallback replies
• In-memory Conversation Log – Stored in Python runtime as session state
Main Components
a. Frontend (Streamlit App)
• Chat Window Display
b. Backend Logic (VicBot.py)
• Intent Recognition Module (rule-based keyword matching)
• Conversation State Manager (stores history in session state)
• AI Router (decides rule-based vs. NLP fallback)
c. External NLP Service
• HuggingFace API endpoint hosting BlenderBot 400M
d. (Optional future) Data Store / Authentication Service
Relationships & Data Flows
• If match → Rule Response → Conversation State → UI Response
• If no match → API Request → BlenderBot Service → API Response → Conversation State → UI Response
Architectural Patterns
• Monolithic single-process but logically layered (UI / Business Logic / External Service)
• “Router” pattern deciding between rule engine and AI model
• Session-based state management for context
Relevant Technologies
• Python (VicBot.py)
• Streamlit (UI Layer)
• HuggingFace Transformers API (BlenderBot)
• (Future) SQL/NoSQL Database, Auth0/OIDC, Speech-to-Text & TTS