Skip to content

Rajeshdevandla/support-gpt

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 

Repository files navigation

SupportGPT

AI customer support assistant — FAQ matching, sentiment-aware escalation, and LLM-generated responses. Built with Amazon Bedrock (Claude) and FastAPI.

Python 3.11+ Amazon Bedrock FastAPI Streamlit License: MIT Docker


Live Demo

Portfolio: rajeshdevandla.github.io

View the full project overview, architecture, and demo walkthrough on the portfolio site.

To run locally, follow the Quick Start below.


What Problem This Solves

Most customer support workloads are repetitive: the same 20 questions account for 80% of tickets. SupportGPT handles those automatically via a FAQ layer, uses sentiment analysis to detect frustrated customers before they escalate themselves, and falls back to Claude for everything else. Human agents only see the hard cases.


Demo

Customer: "How do I return an item?"
SupportGPT: [FAQ match] "Returns are accepted within 30 days with original receipt..."
handled_by: faq | sentiment_score: 0.72 | escalate: false


Customer: "I've been waiting 3 weeks and NOBODY is helping me this is unacceptable"
SupportGPT: [Escalating to human agent — frustrated customer detected]
handled_by: escalation | sentiment_score: 0.18 | escalate: true


Customer: "Can you explain your enterprise pricing options?"
SupportGPT: [Bedrock] "Our enterprise plans are customized based on team size and usage..."
handled_by: llm | sentiment_score: 0.65 | escalate: false

Ticket summary (auto-generated at end of session):

Session: 3 messages | Duration: 4 minutes
Topics: returns policy, shipping delay, enterprise pricing
Escalated: Yes (message 2 — high frustration detected)
Recommended action: Follow up on shipping delay

About

AI support chatbot with sentiment-aware escalation (80%+ FAQ auto-resolution), auto-ticket summarization, and frustrated-customer detection. Built on Amazon Bedrock.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages