ChariotAI - AI Powered Student Support Chatbot

ChariotAI is an enterprise-grade, RAG-powered (Retrieval-Augmented Generation) student support chatbot designed specifically for the UK Universities. It provides 24/7, highly accurate, and legally compliant answers to student queries ranging from admissions and academic integrity to wellbeing support.

🌟 Live Demo

• Live Frontend: (https://delightful-coast-0e2e5d103.6.azurestaticapps.net/)
• Production API: https://chariotai-api-prod.azurewebsites.net

---

🏗️ Architecture

This project strictly adheres to Cloud-Native and Infrastructure-as-Code (IaC) methodologies using a unified Monorepo structure.

graph TD
    A[Student] --> B[React Frontend]
    B -->|Deployed via GitHub Actions| C(Azure Static Web Apps)
    C -->|REST API POST /chat| D[FastAPI Backend]
    D -->|Deployed via GitHub Actions| E(Azure App Service - Linux B1)
    E -->|Embed Query| F[Azure OpenAI]
    F -->|GPT-4o-mini / ada-002| E
    E -->|Vector Search| G[Azure AI Search]
    G -->|Retrieve Official Kent Docs| E

Loading

The Tech Stack

• Frontend: React.js + Vite (Tailwind/CSS glassmorphism UI, Markdown rendering support).
• Backend: Python FastAPI (High-performance async REST framework).
• AI Orchestrator: LangChain (Manages conversation memory and RAG execution).
• Vector Database: Azure AI Search (Stores vector embeddings of the University's rulebooks).
• LLM: Azure OpenAI (gpt-4o-mini and text-embedding-ada-002).
• CI/CD: GitHub Actions (Independent pipelines for frontend and backend pushes).
• Infrastructure: HashiCorp Terraform (Provisions the entire Azure ecosystem).

---

🛡️ Enterprise Safety & Guardrails

A university-facing AI must never hallucinate policies or mishandle student mental health crises.

1. Zero Hallucination Policy: The LLM is restricted (temperature=0.0) and grounded strictly in documents retrieved from Azure AI Search.
2. Crisis Escalation Guardrail: The FastAPI backend intercepts severe keywords (e.g., crisis, overwhelmed, depressed). If triggered, it instantly bypasses the AI generation and returns a hard-coded response directing the student to the Student Support & Wellbeing (SSW) emergency contact numbers and the Samaritans.

---

🚀 Deployment & CI/CD

This monorepo utilizes GitHub Actions for chirurgically precise deployments:

• Frontend Pipeline (.github/workflows/frontend-deploy.yml): Triggers only when React code is changed. Compiles the Vite build and pushes directly to Azure Static Web Apps.
• Backend Pipeline (.github/workflows/backend-deploy.yml): Triggers only on Python code changes. Packages the FastAPI application and uses ZipDeploy to update the Azure App Service container.

Local Development

1. Clone the repository.
2. Navigate to /backend and run pip install -r requirements.txt.
3. Add your .env file with your Azure credentials.
4. Run the backend: uvicorn main:app --reload.
5. Open a new terminal, navigate to /frontend, run npm install and npm run dev.