Azure
Azure hosts two surfaces Supercompat supports: Azure OpenAI deployments (models like
gpt-4.1 deployed under an Azure resource) and Azure AI Foundry Agents (stateful agents with threads, runs, and tool use).Azure OpenAI
Use
azureOpenaiClientAdapter with the AzureOpenAI class from the OpenAI SDK.import { AzureOpenAI } from 'openai'
import {
supercompat,
azureOpenaiClientAdapter,
azureResponsesRunAdapter,
memoryStorageAdapter,
} from 'supercompat/openai'
const azureOpenai = new AzureOpenAI({
endpoint: process.env.AZURE_OPENAI_ENDPOINT!,
apiKey: process.env.AZURE_OPENAI_API_KEY,
apiVersion: '2024-10-01-preview',
})
const client = supercompat({
clientAdapter: azureOpenaiClientAdapter({ azureOpenai }),
storageAdapter: memoryStorageAdapter(),
runAdapter: azureResponsesRunAdapter(),
})
const response = await client.responses.create({
model: 'my-gpt-4-1-deployment',
input: 'Summarize the solar system.',
})
For Azure-managed conversation state, use
azureResponsesStorageAdapter:import { azureResponsesStorageAdapter } from 'supercompat/openai'
supercompat({
clientAdapter: azureOpenaiClientAdapter({ azureOpenai }),
storageAdapter: azureResponsesStorageAdapter(),
runAdapter: azureResponsesRunAdapter(),
})
Azure AI Foundry Agents
Use
azureAiProjectClientAdapter with an AIProjectClient, and pair it with azureAgentsRunAdapter and azureAgentsStorageAdapter.npm install supercompat openai @azure/ai-projects @azure/identity @prisma/client
import { AIProjectClient } from '@azure/ai-projects'
import { DefaultAzureCredential } from '@azure/identity'
import { PrismaClient } from '@prisma/client'
import {
supercompat,
azureAiProjectClientAdapter,
azureAgentsRunAdapter,
azureAgentsStorageAdapter,
} from 'supercompat/openai'
const prisma = new PrismaClient()
const azureAiProject = new AIProjectClient(
process.env.AZURE_AI_PROJECT_ENDPOINT!,
new DefaultAzureCredential(),
)
const client = supercompat({
clientAdapter: azureAiProjectClientAdapter({ azureAiProject }),
storageAdapter: azureAgentsStorageAdapter({ azureAiProject, prisma }),
runAdapter: azureAgentsRunAdapter({ azureAiProject }),
})
const assistant = await client.beta.assistants.create({
model: 'my-agent-deployment',
instructions: 'You are a concise assistant.',
})
const thread = await client.beta.threads.create()
await client.beta.threads.messages.create(thread.id, {
role: 'user',
content: 'Hello.',
})
const run = await client.beta.threads.runs.createAndPoll(thread.id, {
assistant_id: assistant.id,
})
Azure owns the assistants and threads; Prisma covers durability and run-step caching.
Hybrid
If your Azure deployment exposes both Agents and Responses,
azureAgentsResponsesRunAdapter handles both surfaces on a single client. See the source for the full option list.