Edu SDKModels & providers
How to pass a model to edu-sdk with the AI SDK.
Every generator takes a model option. The SDK forwards it to the AI SDK generateText call — it does not ship a private model registry.
What model accepts
| Value | When to use |
|---|---|
string | A model ID your AI SDK setup can resolve (for example a gateway ID) |
LanguageModel | A provider instance from @ai-sdk/openai, @ai-sdk/anthropic, @ai-sdk/google, and so on |
import { createQuiz } from "edu-sdk";
import { openai } from "@ai-sdk/openai";
// String ID (requires a configured AI SDK gateway / default provider)
const quiz = await createQuiz({
model: "google/gemini-3.6-flash",
content,
count: 10,
});
// Explicit provider instance
const quizWithProvider = await createQuiz({
model: openai("gpt-4.1-mini"),
content,
count: 10,
});Install and configure whichever AI SDK provider packages your app needs. edu-sdk depends on ai and uses it for generation; your app still owns API keys and provider wiring.
Environment
Typical setups use an API key in the environment (for example OPENAI_API_KEY, AI_GATEWAY_API_KEY, or your provider’s equivalent). Follow the provider’s AI SDK docs for the exact variable names and client setup.
Generation runs on the server (or any Node-capable runtime). Do not call generators from the browser with a secret key.
What the SDK does not wrap
Validation and file extraction errors throw EduSDKError subclasses. Model, network, rate-limit, and generation failures come from the AI SDK and are not wrapped. Catch those separately from InvalidInputError. See Error handling.
Artifact metadata
The returned Artifact.metadata.model is a string label: the ID you passed, or model.modelId from a LanguageModel, or "language-model" as a fallback.