prompt-executor-deepseek-client
A client implementation for executing prompts using DeepSeek's GPT models with support for custom parameters.
Overview
This module provides a client implementation for the DeepSeek API, allowing you to execute prompts using DeepSeek models with full control over generation parameters, reasoning, and structured outputs.
Supported Models
| Name | Speed | Price | Input | Output |
|---|---|---|---|---|
| DeepSeekChat | Fast | $0.27-$ 1.1 | Text, Tools | Text, Tools |
| DeepSeekReasoner | Medium | $0.55-$ 2.19 | Text, Tools | Text, Tools |
Model-Specific Parameters Support
The client supports DeepSeek-specific parameters through DeepSeekParams class:
val deepSeekParams = DeepSeekParams(
temperature = 0.7,
maxTokens = 1000,
frequencyPenalty = 0.5,
presencePenalty = 0.5,
topP = 0.9,
topK = 40,
stop = listOf("\n", "END"),
logprobs = true,
topLogprobs = 5,
includeThoughts = true,
thinkingBudget = 2000
)Key Parameters:
temperature (0.0-2.0): Controls randomness in generation
topP (0.0-1.0): Nucleus sampling parameter
frequencyPenalty (-2.0-2.0): Reduces repetition of frequent tokens
presencePenalty (-2.0-2.0): Encourages new topics/tokens
includeThoughts: Request model reasoning traces (reasoning models)
thinkingBudget: Token limit for internal reasoning
logprobs: Include log probabilities for generated tokens
stop: Stop sequences to halt generation
Using in your project
Add the dependency to your project:
dependencies {
implementation("ai.koog.prompt:prompt-executor-deepseek-client:$version")
}Configure the client with your API key:
val deepseekClient = DeepSeekLLMClient(
apiKey = "your-deepseek-api-key",
)Example of usage
suspend fun main() {
val client = DeepSeekLLMClient(
apiKey = System.getenv("DEEPSEEK_API_KEY"),
)
// Basic example
val response = client.execute(
prompt = prompt {
system("You are helpful assistant")
user("What time is it now?")
},
model = DeepSeekModels.DeepSeekChat,
)
// Advanced example with custom parameters
val advancedResponse = client.execute(
prompt = prompt {
system("You are a helpful coding assistant")
user("Write a Python function to calculate factorial")
},
model = DeepSeekModels.DeepSeekReasoner,
params = DeepSeekParams(
temperature = 0.3,
maxTokens = 2000,
includeThoughts = true,
thinkingBudget = 1000,
frequencyPenalty = 0.1,
topP = 0.95,
stop = listOf("```\n\n")
)
)
println(response)
println(advancedResponse)
}Additional Examples
// Structured output with JSON schema
val structuredResponse = client.execute(
prompt = prompt {
system("Extract key information as JSON")
user("John Doe, age 30, works as software engineer at TechCorp")
},
model = DeepSeekModels.DeepSeekChat,
params = DeepSeekParams(
temperature = 0.1,
schema = jsonSchema {
object {
property("name", string())
property("age", integer())
property("occupation", string())
property("company", string())
}
}
)
)
// Reasoning with thinking budget
val reasoningResponse = client.execute(
prompt = prompt {
system("Solve this step by step")
user("If a train travels 120 km in 2 hours, what's its average speed?")
},
model = DeepSeekModels.DeepSeekReasoner,
params = DeepSeekParams(
temperature = 0.5,
includeThoughts = true,
thinkingBudget = 500,
topLogprobs = 3,
logprobs = true
)
)