Démarrage Rapide

Construisez votre première application IA avec Orka AI en moins de 5 minutes.

Méthodes d'Import

Import standard :

import { createOrka, OpenAIAdapter, MemoryVectorAdapter } from 'orkajs';

Import tree-shakeable (recommandé pour production) :

import { createOrka } from 'orkajs/core';
import { OpenAIAdapter } from 'orkajs/adapters/openai';
import { MemoryVectorAdapter } from 'orkajs/adapters/memory';

Créer l'instance Orka

const orka = createOrka({
  llm: new OpenAIAdapter({ apiKey: process.env.OPENAI_API_KEY! }),
  vectorDB: new MemoryVectorAdapter(),
});

Indexer vos documents

await orka.knowledge.create({
  name: 'company-docs',
  source: [
    'Orka AI is a TypeScript framework for building LLM applications.',
    'It supports OpenAI, Anthropic, Mistral, and Ollama as LLM providers.',
    'Vector databases supported include Pinecone, Qdrant, Chroma, and in-memory.',
  ],
});

Poser une question

const result = await orka.ask({
  knowledge: 'company-docs',
  question: 'What LLM providers does Orka AI support?',
  includeContext: true,
});
 
console.log(result.answer);
// → "Orka AI supports OpenAI, Anthropic, Mistral, and Ollama."
 
console.log(`Latency: ${result.latencyMs}ms`);
console.log(`Tokens: ${result.usage.totalTokens}`);

Exemple Complet

complete-example.ts

// Standard import
import { createOrka, OpenAIAdapter, MemoryVectorAdapter } from 'orkajs';
 
// OR Tree-shakeable import (recommended)
// import { createOrka } from 'orkajs';
// import { OpenAIAdapter } from 'orkajs/adapters/openai';
// import { MemoryVectorAdapter } from 'orkajs/adapters/memory';
 
async function main() {
  const orka = createOrka({
    llm: new OpenAIAdapter({ apiKey: process.env.OPENAI_API_KEY! }),
    vectorDB: new MemoryVectorAdapter(),
  });
 
  const { chunkCount } = await orka.knowledge.create({
    name: 'docs',
    source: [
      'Orka AI is a TypeScript framework for building LLM applications.',
      'It supports RAG, agents, workflows, evaluation, and multi-model orchestration.',
    ],
  });
  console.log(`Indexed ${chunkCount} chunks`);
 
  const result = await orka.ask({
    knowledge: 'docs',
    question: 'What is Orka AI?',
  });
  console.log(result.answer);
}
 
main().catch(console.error);

npx tsx quick-start.ts

Exécutez l'exemple

Et Ensuite ?

Knowledge & RAG →Workflows →Agents →