v0.0.29

v0.0.29 Changelog

OpenResponses Provider

New responses() provider implementing the OpenResponses specification - an open-source standard for multi-provider, interoperable LLM interfaces.

Features

• Configurable host URL for any OpenResponses-compatible server
• Full multimodal support: text, images, documents, video, audio
• Streaming with semantic SSE events
• Function/tool calling with parallel execution
• Structured output via JSON schema
• Reasoning summaries for gpt-5/o-series models
• Multi-turn conversation context preservation

Usage

import { llm } from '@providerprotocol/ai';
import { responses } from '@providerprotocol/ai/responses';

// Using with OpenAI
const model = llm({
  model: responses('gpt-5.2', {
    host: 'https://api.openai.com/v1',
    apiKeyEnv: 'OPENAI_API_KEY'
  }),
  params: { max_output_tokens: 1000 }
});

const turn = await model.generate('Hello!');

// Using with OpenRouter
const routerModel = llm({
  model: responses('openai/gpt-4o', {
    host: 'https://openrouter.ai/api/v1',
    apiKeyEnv: 'OPENROUTER_API_KEY'
  })
});

// Using with self-hosted server
const localModel = llm({
  model: responses('llama-3.3-70b', {
    host: 'http://localhost:8080/v1'
  })
});

Provider Options

interface ResponsesProviderOptions {
  host: string;      // Required: Base URL (e.g., 'https://api.openai.com/v1')
  apiKeyEnv?: string; // Optional: Env var name (default: 'OPENRESPONSES_API_KEY')
}

Parameters

interface ResponsesParams {
  max_output_tokens?: number;
  temperature?: number;
  top_p?: number;
  presence_penalty?: number;
  frequency_penalty?: number;
  reasoning?: { effort?: 'none' | 'low' | 'medium' | 'high' | 'xhigh'; summary?: 'auto' | 'concise' | 'detailed' };
  truncation?: 'auto' | 'disabled';
  parallel_tool_calls?: boolean;
  // ... and more
}

New Exports

From @providerprotocol/ai/responses:

• responses - Provider factory
• ResponsesProviderOptions
• ResponsesParams
• ResponsesRequest
• ResponsesResponse
• ResponsesStreamEvent
• ResponsesUsage
• ResponsesInputItem
• ResponsesOutputItem
• ResponsesContentPart
• ResponsesFunctionTool
• ResponsesBuiltInTool
• ResponsesToolUnion
• ResponsesHeaders

New Files

• src/providers/responses/types.ts - Type definitions
• src/providers/responses/transform.ts - Request/response transformers
• src/providers/responses/llm.ts - LLM handler
• src/providers/responses/index.ts - Provider factory
• src/responses/index.ts - Re-export entry point

Tests

• tests/unit/providers/responses.transform.test.ts - 15 unit tests for transforms
• tests/live/responses.live.test.ts - 13 live tests against OpenAI backend