Serverless Chatbot with OpenAI Vector Store

A serverless chatbot that leverages OpenAI's Response API and Vector Store for similarity search and real-time streaming responses.

Features

• Serverless architecture using AWS Lambda
• Real-time streaming responses (SSE)
• Similarity search using OpenAI Vector Store (embedding-based retrieval)
• Simple integration with OpenAI's Response API
• Easy deployment with AWS SAM CLI

How it works

The chatbot uses OpenAI's Vector Store to find and reference similar documents based on user queries via embeddings, and streams responses in real time.

Local Development

1. npm run build (if using TypeScript)
2. sam build
3. sam local start-api for testing

Testing

Tests are integration tests that call the actual OpenAI API. You need valid API credentials to run them.

Setup

npm install

Environment Variables

Create a .env file for testing:

OPENAI_API_KEY=your-api-key           # Required
OPENAI_VECTOR_STORE_ID=vs_xxxxx       # Required for file search tests
PROMPT_TEMPLATE_ID=pmpt_xxxxx         # Optional

Run Tests

# Run all tests
npm test

# Run specific test file
npm test stream_response
npm test chat_handler

# Watch mode
npm run test:watch

# With coverage
npm run test:coverage

Test Structure

test/
├── chat_handler.test.ts    # Request validation tests
└── stream_response.test.ts # OpenAI API integration tests

Deploy

Run sam deploy (use --guided for the first time)

CI/CD

SonarCloud Integration

On push to main, GitHub Actions runs tests with coverage and sends results to SonarCloud.

Required GitHub Secrets:

Secret	Description
SONAR_TOKEN	SonarCloud token (for CI IT Test)
OPENAI_API_KEY	For running integration tests
VECTOR_STORE_ID	For file search tests
PROMPT_TEMPLATE_ID	Optional

GitHub Actions Integration

You can integrate this repository from another GitHub Actions workflow using the actions/checkout step with the repository option:

- name: Checkout ServerlessEmbedAI
  uses: actions/checkout@v4
  with:
    repository: auditive-tokyo/serverlessembeddai
    ref: main