Quick Methods

Simple APIs for basic LLM interactions without agent setup.

Overview

For simple prompts without tools, use quick methods:

import { LLMist } from 'llmist';

// One-shot completion
const answer = await LLMist.complete('What is 2+2?');

// Streaming
for await (const chunk of LLMist.stream('Tell me a story')) {
  process.stdout.write(chunk);
}

Static Methods

`LLMist.complete()`

Get a complete response as a string:

// Basic
const answer = await LLMist.complete('Explain quantum computing');

// With options
const answer = await LLMist.complete('Write a haiku', {
  model: 'sonnet',
  temperature: 0.9,
  systemPrompt: 'You are a poet',
  maxTokens: 100,
});

`LLMist.stream()`

Stream text chunks in real-time:

// Basic
for await (const chunk of LLMist.stream('Tell me a story')) {
  process.stdout.write(chunk);
}

// With options
for await (const chunk of LLMist.stream('Write code', {
  model: 'gpt4o',
  systemPrompt: 'You are a coding assistant',
})) {
  process.stdout.write(chunk);
}

Instance Methods

Use with a configured client:

const client = new LLMist({
  defaultProvider: 'anthropic',
});

// Complete
const answer = await client.complete('Hello');

// Stream
for await (const chunk of client.streamText('Hello')) {
  process.stdout.write(chunk);
}

Options

interface TextGenerationOptions {
  model?: string;        // Model name or alias (default: 'gpt-5-mini')
  temperature?: number;  // 0-1 (default: provider default)
  systemPrompt?: string; // System prompt (default: none)
  maxTokens?: number;    // Max tokens (default: provider default)
}

Model Shortcuts

Works with all model shortcuts:

await LLMist.complete('Hello', { model: 'haiku' });
await LLMist.complete('Hello', { model: 'sonnet' });
await LLMist.complete('Hello', { model: 'gpt4o' });
await LLMist.complete('Hello', { model: 'flash' });

When to Use Quick Methods

Use quick methods when:

Simple prompts without tools
No conversation history needed
No need for event handling
Just want text output

Use agents when:

Need tools (gadgets)
Want streaming events
Need conversation history
Want lifecycle hooks

Comparison

// Quick method (simple)
const answer = await LLMist.complete('What is 2+2?');

// Agent (same result, more verbose)
const answer = await LLMist.createAgent()
  .withModel('gpt-5-mini')
  .askAndCollect('What is 2+2?');