CompletionConfig
data class CompletionConfig(val temperature: Double = DEFAULT_TEMPERATURE, val numCtx: Int = DEFAULT_NUM_CTX, val maxTokens: Int = DEFAULT_MAX_TOKENS, val numPredict: Int = DEFAULT_NUM_PREDICT, val reasoningEffort: ReasoningEffort? = null, val maxToolIterations: Int? = DEFAULT_MAX_TOOL_ITERATIONS, val responseFormat: ResponseFormat? = null)
Configuration for LLM completion requests.
Provides a unified, immutable configuration value across providers. Mirrors mojentic.llm.completion_config.CompletionConfig in the Python reference.
Constructors
Link copied to clipboard
constructor(temperature: Double = DEFAULT_TEMPERATURE, numCtx: Int = DEFAULT_NUM_CTX, maxTokens: Int = DEFAULT_MAX_TOKENS, numPredict: Int = DEFAULT_NUM_PREDICT, reasoningEffort: ReasoningEffort? = null, maxToolIterations: Int? = DEFAULT_MAX_TOOL_ITERATIONS, responseFormat: ResponseFormat? = null)
Properties
Link copied to clipboard
Optional ceiling on broker tool-recursion depth; null is unlimited.
Link copied to clipboard
Tokens to predict, -1 for no limit. Ollama-specific knob.
Link copied to clipboard
Optional reasoning effort for thinking-capable models.
Link copied to clipboard
Optional output format forwarded by gateways in streaming and non-streaming requests. Null keeps the provider default. See ResponseFormat.
Link copied to clipboard
Sampling temperature. Higher values produce more random output.