Changelog

Added

Assistants API (Beta) - Full support for OpenAI Assistants v2
- Create, retrieve, and list Assistants
- Thread management (create, retrieve, delete)
- Message operations within threads
- Run operations with streaming support
- Vector Store support via tool_resources
- File Search and Code Interpreter v2 capabilities
Files API - Upload, list, retrieve, and delete files
Fine-tuning API - Create and manage fine-tuning jobs
Streaming for Assistants - Server-Sent Events (SSE) support via createRunStream
Vector Store Support - New models for ToolResources, FileSearchResources, and VectorStore
Coroutines Support - All API methods now use suspend functions
OpenAiConfig - Configuration class for API key and client settings

Changed

BREAKING: Replaced callback-based API with Kotlin Coroutines
- All methods now use suspend functions instead of callbacks
- Error handling via try-catch instead of error callbacks
BREAKING: Removed OpenAiInitializer static initialization
- Use OpenAiConfig constructor parameter instead
Chat Completions - Updated to support GPT-4 and GPT-3.5 Turbo features
- Function calling / Tools support
- JSON mode via response_format
- Added seed, system_fingerprint, and other modern parameters
Image Generation - Updated for DALL-E 3 support
- Added model, quality, and style parameters
Audio API - Enhanced transcription and translation endpoints
- Added optional parameters: prompt, response_format, temperature, language
Tool Definition - Made function nullable to support file_search and code_interpreter types
Moderation API - Updated default model to omni-moderation-latest

Deprecated

Completions API - Legacy text completions (use Chat Completions instead)
Edits API - Legacy edits endpoint (use Chat Completions instead)
file_ids in CreateAssistantRequest - Use tool_resources for Assistants v2

Updated

Gradle Wrapper to 8.6
Android Gradle Plugin to 8.3.1
Kotlin to 1.9.23
AndroidX Core KTX to 1.12.0
Material Components to 1.11.0
Retrofit to 2.9.0
OkHttp Logging Interceptor to 4.12.0
Added lifecycle-runtime-ktx 2.6.2 for Coroutines support

Migration Guide

From Callbacks to Coroutines

Before:

OpenAiInitializer.initialize("API_KEY")
val client = OpenApiClient()

client.getChatCompletion(messages) { result, error ->
    if (error != null) {
        // Handle error
    } else if (result != null) {
        // Use result
    }
}

After:

val config = OpenAiConfig(apiKey = "API_KEY")
val client = OpenApiClient(config)

lifecycleScope.launch {
    try {
        val result = client.getChatCompletion(messages)
        // Use result
    } catch (e: Exception) {
        // Handle error
    }
}

Using Vector Stores

val assistant = CreateAssistantRequest(
    model = "gpt-4",
    tools = listOf(Tool(type = "file_search")),
    tool_resources = ToolResources(
        file_search = FileSearchResources(
            vector_store_ids = listOf("vs_123")
        )
    )
)