SLIIT Support Assistant

Students currently need to go through multiple pages, links, and menus in the SLIIT support portal to find specific details. This project solves that by providing a single assistant interface where users can ask questions directly and get relevant answers based on support content.

Project Structure

• webapp/ - React + Vite frontend
• ballerina-integration/ - Ballerina API service (/api/chat)
• docs/ - Markdown knowledge source files
• ingest.py - Ingestion script to embed and upsert knowledge into Pinecone

Prerequisites

• Node.js 20+
• pnpm
• Python 3.10+
• Ballerina 2201.13.1+
• OpenAI API key
• Pinecone API key and index endpoint

Setup

1) Install frontend dependencies

cd webapp
pnpm install

2) Install Python dependencies for ingestion

pip install openai pinecone python-dotenv pyyaml PyMuPDF

3) Configure ingestion environment

Create a .env file in the project root:

OPENAI_API_KEY=your_openai_api_key
PINECONE_API_KEY=your_pinecone_api_key
PINECONE_INDEX_NAME=sliit-support
ASGARDEO_ISSUER="https://api.asgardeo.io/t/your-org/oauth2/token"
ASGARDEO_AUDIENCE="your-client-id"
ASGARDEO_JWKS_URL="https://api.asgardeo.io/t/your-org/oauth2/jwks"

4) Configure Ballerina service

Create ballerina-integration/Config.toml:

OPENAI_TOKEN="your_openai_api_key"
PINECONE_API_KEY="your_pinecone_api_key"
PINECONE_URL="your_pinecone_index_url"
bucketName="sliit-support"

Run

1) Ingest knowledge into Pinecone

python ingest.py

2) Start Ballerina API

cd ballerina-integration
bal run

3) Start web app

cd webapp
pnpm dev

Open the app in your browser and start chatting.