@obsku/provider-ai-sdk

Vercel AI SDK provider adapter for @obsku/framework. Use any AI SDK-compatible model as an obsku LLMProvider.

Installation

npm install @obsku/provider-ai-sdk

Quick Start

Provider Factories (Recommended)

import { agent } from "@obsku/framework";
import { openai, anthropic, google, groq } from "@obsku/provider-ai-sdk";

// OpenAI
const provider = openai({ model: "gpt-4o" });

// Anthropic
const provider = anthropic({ model: "claude-sonnet-4-20250514" });

// Anthropic with Extended Thinking
const provider = anthropic({
  model: "claude-opus-4",
  thinkingBudgetTokens: 10000,
});

// Google
const provider = google({ model: "gemini-2.0-flash" });

// Groq
const provider = groq({ model: "llama-3.3-70b" });

const assistant = agent({
  name: "assistant",
  prompt: "You are a helpful assistant.",
});

const result = await assistant.run("Hello!", provider);

Generic Adapter

Use fromAiSdk() to wrap any LanguageModelV1:

import { fromAiSdk } from "@obsku/provider-ai-sdk";
import { createOpenAI } from "@ai-sdk/openai";

const openai = createOpenAI({ apiKey: "..." });
const provider = fromAiSdk(openai("gpt-4o"), {
  contextWindowSize: 128000,
  maxOutputTokens: 16384,
});

API Reference

Factories

| Function | Config | Description | |----------|--------|-------------| | openai(config) | BaseProviderConfig | OpenAI models via @ai-sdk/openai | | anthropic(config) | AnthropicProviderConfig | Anthropic models with optional Extended Thinking | | google(config) | BaseProviderConfig | Google Gemini models via @ai-sdk/google | | groq(config) | BaseProviderConfig | Groq models via @ai-sdk/groq |

fromAiSdk(model, config?)

Wraps any AI SDK LanguageModelV1 as an obsku LLMProvider.

Config Types

interface BaseProviderConfig {
  model: string;
  contextWindowSize?: number;
  maxOutputTokens?: number;
}

interface AnthropicProviderConfig extends BaseProviderConfig {
  thinkingBudgetTokens?: number;  // Enables Extended Thinking
}

interface AdapterConfig {
  contextWindowSize?: number;     // Default: auto-resolved from model registry
  maxOutputTokens?: number;
  providerOptions?: Record<string, unknown>;
}

Model Resolution

Context window sizes and max output tokens are resolved via the framework's ModelRegistry (LiteLLM-backed). Pass explicit values to override:

const provider = await openai({
  model: "gpt-4o",
  contextWindowSize: 128000,
  maxOutputTokens: 16384,
});

Environment Variables

API keys are managed by the AI SDK packages:

• OPENAI_API_KEY — OpenAI
• ANTHROPIC_API_KEY — Anthropic
• GOOGLE_GENERATIVE_AI_API_KEY — Google
• GROQ_API_KEY — Groq

License

MIT