This project demonstrates how to build an AI-enhanced weather service using Genkit, TypeScript, OpenWeatherAPI and Github Models. The application showcases modern Node.js patterns and AI integration techniques.
The core AI setup is initialized with Genkit and GitHub plugin integration. In this case we are going to use the OpenAI o3-mini model:
const ai = genkit({
plugins: [
github({ githubToken: process.env.GITHUB_TOKEN }),
],
model: openAIO3Mini,
});The application defines a custom weather tool using Zod schema validation:
const getWeather = ai.defineTool(
{
name: 'getWeather',
description: 'Gets the current weather in a given location',
inputSchema: weatherToolInputSchema,
outputSchema: z.string(),
},
async (input) => {
const weather = new OpenWeatherAPI({
key: process.env.OPENWEATHER_API_KEY,
units: "metric"
})
const data = await weather.getCurrent({locationName: input.location});
return `The current weather in ${input.location} is: ${data.weather.temp.cur} Degrees in Celsius`;
}
);The service exposes an AI flow that processes weather requests:
const helloFlow = ai.defineFlow(
{
name: 'helloFlow',
inputSchema: z.object({ location: z.string() }),
outputSchema: z.string(),
},
async (input) => {
const response = await ai.generate({
tools: [getWeather],
prompt: `What's the weather in ${input.location}?`
});
return response.text;
}
);The application uses the Genkit Express plugin to create an API server:
const app = express({
flows: [helloFlow],
});The full code for the weather service is as follows:
/* eslint-disable @typescript-eslint/no-explicit-any */
import { genkit, z } from 'genkit';
import { startFlowServer } from '@genkit-ai/express';
import { openAIO3Mini, github } from 'genkitx-github';
import {OpenWeatherAPI } from 'openweather-api-node';
import dotenv from 'dotenv';
dotenv.config();
const ai = genkit({
plugins: [
github({ githubToken: process.env.GITHUB_TOKEN }),
],
model: openAIO3Mini,
});
const weatherToolInputSchema = z.object({
location: z.string().describe('The location to get the current weather for')
});
const getWeather = ai.defineTool(
{
name: 'getWeather',
description: 'Gets the current weather in a given location',
inputSchema: weatherToolInputSchema,
outputSchema: z.string(),
},
async (input) => {
const weather = new OpenWeatherAPI({
key: process.env.OPENWEATHER_API_KEY,
units: "metric"
})
const data = await weather.getCurrent({locationName: input.location});
return `The current weather in ${input.location} is: ${data.weather.temp.cur} Degrees in Celsius`;
}
);
const helloFlow = ai.defineFlow(
{
name: 'helloFlow',
inputSchema: z.object({ location: z.string() }),
outputSchema: z.string(),
},
async (input) => {
const response = await ai.generate({
tools: [getWeather],
prompt: `What's the weather in ${input.location}?`
});
return response.text;
}
);
startFlowServer({
flows: [helloFlow]
});- Install dependencies:
npm install- Configure environment variables:
GITHUB_TOKEN=your_token
OPENWEATHER_API_KEY=your_key- Start development server:
npm run genkit:start- To run the project in debug mode and set breakpoints, you can run:
npm run genkit:start:debugAnd then launch the debugger in your IDE. See the .vscode/launch.json file for the configuration.
- If you want to build the project, you can run:
npm run build- Run the project in production mode:
npm run start:productiongenkit: ^1.0.5@genkit-ai/express: ^1.0.5openweather-api-node: ^3.1.5genkitx-github: ^1.13.1dotenv: ^16.4.7
tsx: ^4.19.2typescript: ^5.7.2
- Uses ES Modules (
"type": "module") - TypeScript with
NodeNextmodule resolution - Output directory: lib
- Full TypeScript support with type definitions
Apache 2.0