@crewai-ts/openai

OpenAI and OpenAI-compatible native providers for CrewAI TypeScript.

Provides OpenAICompletion and OpenAICompatibleCompletion with support for chat completions, responses API, built-in tools (web search, file search, code interpreter, computer use), reasoning summaries, and structured outputs.

Install

npm install @crewai-ts/openai

Requirements:

• Node.js 22 or later
• @crewai-ts/core 0.2.0 or later

Quick Start

import { Agent } from "@crewai-ts/core";
import { OpenAICompletion, registerOpenAIProvider } from "@crewai-ts/openai";

registerOpenAIProvider();

const agent = new Agent({
  role: "Assistant",
  goal: "Answer questions",
  backstory: "A helpful assistant.",
  llm: new OpenAICompletion({ model: "gpt-4o" }),
});

Or use the registered provider name:

const agent = new Agent({
  role: "Assistant",
  goal: "Answer questions",
  backstory: "A helpful assistant.",
  llm: "openai/gpt-4o",
});

OpenAI-Compatible Providers

OpenAICompatibleCompletion extends OpenAICompletion and is configured for a specific upstream provider via the provider option. The list of known provider keys is exported as OPENAI_COMPATIBLE_PROVIDERS.

import { OpenAICompatibleCompletion } from "@crewai-ts/openai";

// Ollama — local llama.cpp / Ollama server
const ollama = new OpenAICompatibleCompletion({
  provider: "ollama", // or "ollama_chat"
  model: "llama3.1",
  baseUrl: "http://localhost:11434/v1",
  apiKey: "ollama", // any non-empty value; Ollama ignores the key
});

// OpenRouter — hosted OpenAI-compatible routing
const openrouter = new OpenAICompatibleCompletion({
  provider: "openrouter",
  model: "anthropic/claude-3.5-sonnet",
  apiKey: process.env.OPENROUTER_API_KEY,
});

Omit provider only when you want the default openrouter config. Passing an unknown provider throws at construction time.

Built-in Tools

Enable OpenAI built-in tools:

const llm = new OpenAICompletion({
  model: "gpt-4o",
  builtinTools: ["web_search", "file_search", "code_interpreter"],
});

Responses API

Use the OpenAI Responses API:

const llm = new OpenAICompletion({
  model: "gpt-4o",
  api: "responses",
  store: true,
});

Exports

• OpenAICompletion — main OpenAI provider class
• OpenAICompatibleCompletion — OpenAI-compatible provider class
• ResponsesAPIResult — responses API result wrapper
• registerOpenAIProvider — register the provider with the core runtime
• Result types: WebSearchResult, FileSearchResult, CodeInterpreterResult, ComputerUseResult, ReasoningSummary

License

MIT