DeepSeek Kotlin SDK/org.oremif.deepseek.models/FIMCompletionParams

FIMCompletionParams

class FIMCompletionParams : DeepSeekParams

Parameters for configuring Fill-in-the-Middle (FIM) completion requests to DeepSeek models.

FIM allows the model to complete text where you provide both a prefix and a suffix, making it useful for code completion, text infilling, and template-based generation.

Example:

val params = fimCompletionParams {
    temperature = 0.7
    maxTokens = 200
    suffix = "}"  // The code should end with a closing bracket
}

val completion = client.fim(params, "function calculateArea(radius) {")
println(completion.choices.first().text)

Types

Builder

class Builder

Builder for creating FIMCompletionParams with standard (non-streaming) configuration.

StreamBuilder

class StreamBuilder

Builder for creating FIMCompletionParams specifically configured for streaming responses.

Properties

echo

val echo: Boolean?

Whether to include the prompt in the returned completion. Cannot be combined with suffix or logprobs — the API answers 400 echo should not be used with suffix / echo should not be used with logprobs

frequencyPenalty

val ~~frequencyPenalty~~: Double?

Sent as frequency_penalty, ignored by the API.

logprobs

val logprobs: Int?

Maximum number of log probabilities to return (up to 20)

maxTokens

val maxTokens: Int?

Maximum number of tokens to generate in the response. At least 1, and otherwise bounded by the model's context length.

model

val model: ChatModel

The DeepSeek model to use. The reference documents only ChatModel.DEEPSEEK_V4_PRO, which is the default, but the endpoint also serves ChatModel.DEEPSEEK_FLASH

presencePenalty

val ~~presencePenalty~~: Double?

Sent as presence_penalty, ignored by the API.

stop

val stop: StopReason?

Custom stop sequences that cause the model to stop generating further tokens.

stream

val stream: Boolean?

Whether to stream the response back piece by piece

streamOptions

val streamOptions: StreamOptions?

Configuration options for streaming responses

suffix

val suffix: String?

Text that the model should complete towards (the ending part in FIM)

temperature

val temperature: Double?

Controls randomness in output generation (higher = more random). Expected range: 0.0..2.0.

topP

val topP: Double?

Controls diversity via nucleus sampling (consider only tokens with top_p probability mass). Expected range: 0.0..1.0.

Functions

chat

fun chat(block: ChatCompletionParams.Builder.() -> Unit): ChatCompletionParams

Shortcut to chatCompletionParams, available on any existing DeepSeekParams.

chatStream

fun chatStream(block: ChatCompletionParams.StreamBuilder.() -> Unit): ChatCompletionParams

Shortcut to chatCompletionStreamParams, available on any existing DeepSeekParams.

copy

fun copy(model: ChatModel = this.model, echo: Boolean? = this.echo, frequencyPenalty: Double? = this.frequencyPenalty, logprobs: Int? = this.logprobs, maxTokens: Int? = this.maxTokens, presencePenalty: Double? = this.presencePenalty, stop: StopReason? = this.stop, stream: Boolean? = this.stream, streamOptions: StreamOptions? = this.streamOptions, suffix: String? = this.suffix, temperature: Double? = this.temperature, topP: Double? = this.topP): FIMCompletionParams

Creates a copy of these parameters with optional changes to specific properties.

createRequest

fun createRequest(prompt: String): FIMCompletionRequest

Creates a FIMCompletionRequest from these parameters and the provided prompt.

equals

open operator override fun equals(other: Any?): Boolean

fim

fun fim(block: FIMCompletionParams.Builder.() -> Unit): FIMCompletionParams

Shortcut to fimCompletionParams, available on any existing DeepSeekParams.

fimStream

fun fimStream(block: FIMCompletionParams.StreamBuilder.() -> Unit): FIMCompletionParams

Shortcut to fimCompletionStreamParams, available on any existing DeepSeekParams.

hashCode

open override fun hashCode(): Int

toString

open override fun toString(): String