OpenAI Provider

Setup

Set your OpenAI API key:

export OPENAI_API_KEY=sk-...

llmist will automatically discover and use OpenAI.

Available Models

Text Models

| Model | Alias | Best For | |-------|-------|----------| | gpt-5 | gpt5 | Complex reasoning, coding | | gpt-5-mini | gpt5-mini | Fast, cost-effective tasks | | gpt-4-turbo | gpt4-turbo | Balanced performance | | gpt-4o | gpt4o | Multimodal tasks | | o3 | o3 | Advanced reasoning Reasoning | | o4-mini | o4-mini | Fast reasoning Reasoning | | o3-mini | o3-mini | Compact reasoning |

Image Models

| Model | Description | |-------|-------------| | dall-e-3 | High-quality image generation | | dall-e-2 | Faster, lower cost |

Speech Models

| Model | Description | |-------|-------------| | tts-1 | Text-to-speech, standard quality | | tts-1-hd | Text-to-speech, high quality |

Usage Examples

Text Generation

import { LLMist } from 'llmist';

const answer = await LLMist.createAgent()
  .withModel('gpt5')
  .askAndCollect('Explain quantum computing');

Image Generation

import { LLMist } from 'llmist';

const client = new LLMist();
const result = await client.image.generate({
  prompt: 'A sunset over mountains',
  model: 'dall-e-3',
  size: '1024x1024',
});

console.log(result.url);

Speech Generation

import { LLMist } from 'llmist';
import { writeFileSync } from 'fs';

const client = new LLMist();
const result = await client.speech.generate({
  input: 'Hello, world!',
  model: 'tts-1',
  voice: 'nova',
});

writeFileSync('speech.mp3', result.audio);

Reasoning Models

OpenAI's o-series models (o3, o4-mini) and GPT-5 family support reasoning. llmist maps ReasoningEffort to OpenAI's native reasoning.effort parameter:

| Effort | OpenAI Value | |--------|-------------| | "none" | "none" | | "low" | "low" | | "medium" | "medium" | | "high" | "high" | | "maximum" | "xhigh" |

const answer = await LLMist.createAgent()
  .withModel('o3')
  .withReasoning('high')
  .askAndCollect('Prove that √2 is irrational.');

Tip

Models with features.reasoning: true (like o3) auto-enable reasoning at "medium" effort. Use .withoutReasoning() to opt out.

See the Reasoning Models guide for full details.

Vision (Image Input)

GPT-4o and GPT-5 models support image input:

import { LLMist, imageFromUrl } from 'llmist';

const answer = await LLMist.createAgent()
  .withModel('gpt4o')
  .askWithImage(
    'What is in this image?',
    imageFromUrl('https://example.com/photo.jpg')
  )
  .askAndCollect();

Configuration Options

import { LLMist, OpenAIChatProvider } from 'llmist';

const client = new LLMist({
  autoDiscoverProviders: false,
  adapters: [
    new OpenAIChatProvider({
      apiKey: process.env.OPENAI_API_KEY,
      baseUrl: 'https://api.openai.com/v1', // Custom endpoint
      organization: 'org-xxx',              // Optional org ID
    }),
  ],
});

Azure OpenAI

For Azure OpenAI, set a custom baseUrl pointing to your Azure endpoint and use the deployment name as the model.

Cost Tracking

llmist automatically tracks token usage and costs:

for await (const event of agent.run()) {
  if (event.type === 'llm_call_complete') {
    console.log('Tokens:', event.usage);
    console.log('Cost:', event.cost);
  }
}

Rate Limiting

OpenAI has rate limits. Use retry configuration for resilience:

const agent = LLMist.createAgent()
  .withModel('gpt5')
  .withRetry({
    maxAttempts: 3,
    retryOn: ['rate_limit', 'server_error'],
  });