prompt-executor-dashscope-client
A client implementation for executing prompts using Alibaba Cloud's DashScope API with Qwen models. Provides OpenAI-compatible interface for seamless integration.
Overview
This module provides a client implementation for the DashScope API, allowing you to execute prompts using Alibaba's Qwen models. It leverages OpenAI-compatible endpoints and handles authentication, request formatting, response parsing, and tool integration specific to DashScope's API requirements.
Model-Specific Parameters Support
DashScope Parameters
The client supports DashScope-specific parameters through DashscopeParams class:
val dashscopeParams = DashscopeParams(
temperature = 0.7, // Controls randomness (0.0 to 2.0)
maxTokens = 1000, // Maximum tokens to generate
toolChoice = LLMParams.ToolChoice.AUTO // Tool usage control
)API Endpoints
The client connects to DashScope using OpenAI-compatible endpoints:
China mainland:
https://dashscope.aliyuncs.com/compatible-mode/v1International:
https://dashscope-intl.aliyuncs.com/compatible-mode/v1
Using in your project
Add the dependency to your project:
dependencies {
implementation("ai.koog.prompt:prompt-executor-dashscope-client:$version")
}Configure the client with your API key:
val dashscopeClient = DashscopeLLMClient(
apiKey = "your-dashscope-api-key",
)Configuration Options
// Default configuration (China mainland)
val client = DashscopeLLMClient(
apiKey = "your-api-key"
)
// International configuration
val internationalClient = DashscopeLLMClient(
apiKey = "your-api-key",
settings = DashscopeClientSettings(
baseUrl = "https://dashscope-intl.aliyuncs.com/",
timeoutConfig = ConnectionTimeoutConfig(
requestTimeoutMillis = 60_000,
connectTimeoutMillis = 30_000,
socketTimeoutMillis = 60_000
)
)
)Example of usage
suspend fun main() {
val client = DashscopeLLMClient(
apiKey = System.getenv("DASHSCOPE_API_KEY"),
)
// Basic text completion
val response = client.execute(
prompt = prompt {
system("你是一个有用的AI助手")
user("你好，请介绍一下自己")
},
model = DashscopeModels.QWEN_PLUS,
params = LLMParams(
temperature = 0.7,
maxTokens = 1000
)
)
println(response)
}Tool Usage Examples
// Using tools with DashScope
val responseWithTools = client.execute(
prompt = prompt {
system("You are a helpful assistant with access to tools")
user("What's the weather like in Beijing?")
},
model = DashscopeModels.QWEN_PLUS,
params = LLMParams(
temperature = 0.5,
toolChoice = LLMParams.ToolChoice.AUTO
),
tools = listOf(weatherTool) // Your defined tools
)
// Streaming response
val streamingResponse = client.executeStream(
prompt = prompt {
user("Write a short story about a robot learning to paint")
},
model = DashscopeModels.QWEN_FLASH,
params = LLMParams(
temperature = 0.8,
maxTokens = 2000
)
)
streamingResponse.collect { chunk ->
print(chunk)
}
// Using speculation for complex reasoning
val speculationResponse = client.execute(
prompt = prompt {
system("Analyze the following problem step by step")
user("How can we optimize database query performance for a social media platform?")
},
model = DashscopeModels.QWEN_PLUS, // Supports speculation capability
params = LLMParams(
temperature = 0.3,
maxTokens = 3000
)
)Advanced Configuration
// Custom timeout configuration
val customClient = DashscopeLLMClient(
apiKey = "your-api-key",
settings = DashscopeClientSettings(
baseUrl = "https://dashscope.aliyuncs.com/",
chatCompletionsPath = "compatible-mode/v1/chat/completions",
timeoutConfig = ConnectionTimeoutConfig(
requestTimeoutMillis = 120_000, // 2 minutes
connectTimeoutMillis = 30_000, // 30 seconds
socketTimeoutMillis = 120_000 // 2 minutes
)
),
baseClient = HttpClient {
// Custom HTTP client configuration
}
)