@anvia/azure

Connect Anvia agents to Azure OpenAI and Azure AI Foundry endpoints. Azure configuration lives here; request and response mapping is shared with @anvia/openai.

Install

pnpm add @anvia/azure @anvia/core

Quickstart

import { Agent } from "@anvia/core";
import { AzureOpenAIClient } from "@anvia/azure";

const client = new AzureOpenAIClient({
  endpoint: process.env.AZURE_OPENAI_ENDPOINT!,
  apiKey: process.env.AZURE_OPENAI_API_KEY!,
});

const agent = new Agent({
  id: "assistant",
  model: client.completionModel({
    modelId: process.env.AZURE_OPENAI_DEPLOYMENT!,
    api: "responses",
  }),
});

const result = await agent.generate({ prompt: "Hello!" });
if (result.type === "response") console.log(result.output);

endpoint is the resource origin, such as https://example.openai.azure.com. Use baseUrl instead for a full API URL, including Foundry project endpoints. modelId is your Azure deployment name. Chat Completions is the default API.

Supports API keys, Microsoft Entra token providers, injected SDK clients, streaming, tools, structured output, embeddings, and media models where available on your deployment.

Learn more

• Usage and migration guide
• Anvia overview