prompt-executor-openai-client-base

Abstract base client and data models for OpenAI-compatible LLM client implementations.

Overview

This module provides the foundational components for building OpenAI-compatible LLM clients. It includes:

  • AbstractOpenAILLMClient: Abstract base class that implements common OpenAI API functionality

  • OpenAI Data Models: Complete set of serializable data classes for OpenAI API requests and responses

  • Multimodal Content Support: Handles text, images, audio, and file attachments

  • Tool Integration: Full support for function calling and tool execution

  • Streaming Support: Built-in streaming capabilities with server-sent events

This module serves as the foundation for concrete OpenAI client implementations and can be extended to support OpenAI-compatible APIs like Azure OpenAI, OpenRouter, and other providers that follow the OpenAI API format.

Key Components

AbstractOpenAILLMClient

  • Base implementation for OpenAI-compatible clients

  • Handles HTTP communication, authentication, and request/response processing

  • Supports both synchronous and streaming execution modes

  • Includes multimodal content processing (text, images, audio, files)

  • Provides tool calling functionality with automatic serialization

Data Models

  • OpenAIChatCompletionRequest/Response: Complete request/response models

  • OpenAIMessage: Message types (System, User, Assistant, Tool, Developer)

  • Content & ContentPart: Multimodal content support

  • OpenAITool: Function calling definitions and responses

  • OpenAIAudio: Audio configuration and response handling

Using in your project

This module is typically not used directly but as a dependency for concrete OpenAI client implementations:

dependencies {
    implementation("ai.koog.prompt:prompt-executor-openai-model:$version")
}
Content copied to clipboard

Creating a Custom OpenAI-Compatible Client

class CustomOpenAIClient(
    apiKey: String,
    baseUrl: String = "https://api.openai.com/v1"
) : AbstractOpenAILLMClient(
    apiKey = apiKey,
    settings = object : OpenAIBasedSettings(
        baseUrl = baseUrl,
        chatCompletionsPath = "/chat/completions"
    ) {}
)

// Usage
suspend fun main() {
    val client = CustomOpenAIClient(
        apiKey = System.getenv("API_KEY"),
        baseUrl = "https://custom-openai-api.com/v1"
    )
    
    val response = client.execute(
        prompt = prompt {
            system("You are a helpful assistant")
            user("Hello!")
        },
        model = MyCustomModel("gpt-4o")
    )
    
    println(response.first().content)
}
Content copied to clipboard

Supported Features

FeatureSupportNotes
Text GenerationComplete text generation support
StreamingServer-sent events with Flow API
Function CallingFull tool integration with serialization
Multimodal InputImages, audio, documents via attachments
Audio OutputAudio generation for compatible models
Multiple ChoicesSupport for n 1 completions
Temperature ControlTemperature and other sampling parameters
Token UsageDetailed token usage statistics

Extension Points

When extending AbstractOpenAILLMClient:

  1. Custom Settings: Override OpenAIBasedSettings for provider-specific configuration

  2. Authentication: Customize HTTP client setup for different auth methods

  3. Error Handling: Override error processing for provider-specific error formats

  4. Model Capabilities: Define model capabilities and validation

  5. Request Modification: Customize request building for API variations

This module ensures consistency across different OpenAI-compatible providers while allowing flexibility for provider-specific customizations.

Packages

common
common