@tepa/provider-openai

OpenAI LLM provider for the Tepa agent pipeline. Uses the Responses API.

Install

npm install @tepa/provider-openai

Setup

Set your OpenAI API key as an environment variable:

export OPENAI_API_KEY=sk-...

Or use a .env file in your project root.

Usage

import { Tepa } from "@tepa/core";
import { OpenAIProvider } from "@tepa/provider-openai";

const tepa = new Tepa({
  tools: [
    /* ... */
  ],
  provider: new OpenAIProvider(),
});

Provider Options

const provider = new OpenAIProvider({
  apiKey: "sk-...", // Defaults to OPENAI_API_KEY env var
  maxRetries: 3, // Default: 3
  retryBaseDelayMs: 1000, // Base delay for exponential backoff
});

Logging

Every LLM call is automatically logged to a JSONL file in .tepa/logs/. You can disable the default file logger, add custom log listeners, or send logs to external services like Prometheus, NewRelic, or Datadog using the onLog() method:

const provider = new OpenAIProvider({ defaultLog: false });

provider.onLog((entry) => {
  externalLogger.send(entry);
});

See @tepa/provider-core for full logging documentation.

Native Tool Use

This provider supports native tool calling via the OpenAI Responses API. When tool schemas are passed via options.tools, they are forwarded as function definitions. The LLM returns structured function_call blocks with pre-parsed parameters, eliminating text-based JSON parsing errors.

Implementing Custom Providers

Any object implementing the LLMProvider interface from @tepa/types can be used:

import type { LLMProvider } from "@tepa/types";

class MyProvider implements LLMProvider {
  async complete(messages, options) {
    // Your implementation
  }
}