openaiResponsesRunAdapter
Executes runs against OpenAI's native
/responses endpoint. Use this when you want OpenAI-native features: server-side web_search, code_interpreter, file_search, and computer_use_preview.Signature
openaiResponsesRunAdapter({
getOpenaiAssistant?: (args?: { select?: { id?: false } }) =>
Promise<Assistant | Pick<Assistant, 'id'>>,
waitUntil?: <T>(p: Promise<T>) => void | Promise<void>,
})
Basic use
import OpenAI from 'openai'
import {
supercompat,
openaiClientAdapter,
openaiResponsesRunAdapter,
memoryStorageAdapter,
} from 'supercompat/openai'
const client = supercompat({
clientAdapter: openaiClientAdapter({ openai: new OpenAI() }),
storageAdapter: memoryStorageAdapter(),
runAdapter: openaiResponsesRunAdapter(),
})
const response = await client.responses.create({
model: 'gpt-4.1-mini',
input: 'Hello.',
})
With OpenAI-managed state
import {
openaiClientAdapter,
openaiResponsesRunAdapter,
openaiResponsesStorageAdapter,
} from 'supercompat/openai'
const client = supercompat({
clientAdapter: openaiClientAdapter({ openai }),
storageAdapter: openaiResponsesStorageAdapter(),
runAdapter: openaiResponsesRunAdapter(),
})
No database, no Prisma — OpenAI owns the conversation state.
Built-in tools
await client.responses.create({
model: 'gpt-4.1',
input: 'What happened in open-source AI this week?',
tools: [{ type: 'web_search' }],
})
await client.responses.create({
model: 'gpt-4.1',
input: 'Factor 420.',
tools: [{ type: 'code_interpreter' }],
})
await client.responses.create({
model: 'gpt-4.1',
input: 'Summarize my uploaded document.',
tools: [{ type: 'file_search', vector_store_ids: [vectorStoreId] }],
})
getOpenaiAssistant
Attach a stored OpenAI Assistant to the run. Useful when instructions/model/tools are managed as an Assistant record.
openaiResponsesRunAdapter({
getOpenaiAssistant: async () => openai.beta.assistants.retrieve('asst_...'),
})
waitUntil (edge runtimes)
On Vercel Edge or Cloudflare Workers, pass the runtime's
waitUntil so post-stream work (writing the response to storage) completes.import { openaiResponsesRunAdapter } from 'supercompat/openai'
openaiResponsesRunAdapter({
waitUntil: (p) => context.waitUntil(p),
})