google-genai

Sample Google GenAI Application

This application demonstrates using the agent to instrument @google/genai and record spans for chat completions and embeddings.

This app supports both Gemini AI and Vertex AI. Specify which one you want with environment variable GOOGLE_GENAI_USE_VERTEXAI; acceptable values are 'true' for true and blank (not defined) for false.

Getting started

Note: This application requires the use of Node.js v20+.

1. Make a Google Cloud project or a Gemini API Key.

2. Clone or fork this repository.

3. Install dependencies and run application.

   npm i
   cp env.sample .env
   # Fill out `NEW_RELIC_LICENSE_KEY`, `GEMINI_API_KEY` or `GOOGLE_CLOUD_PROJECT` and `GOOGLE_CLOUD_LOCATION` in .env and save 
   npm start

4. Make requests to application.

   # Basic generate content example
   curl http://localhost:3000/
   
   # Generate content stream example
   curl http://localhost:3000/stream
   
   # Embed content example
   curl http://localhost:3000/embed

Inspecting AI Responses

After sending a few requests, navigate to your application in APM & Services. Select AI Monitoring > AI Responses:

If you click the details of a response you will see metrics, trace information and LLM specific information:

Click the metadata tab to see more information about the raw events: