Enums and Constants

The LLM sections on this page reflect ZeticMLange iOS 1.5.11.

This page documents the enums and configuration types most commonly used from the iOS SDK.

ModelMode

Controls the inference strategy for general, non-LLM models.

import ZeticMLange

Value	Description
`RUN_AUTO`	Default. Balanced speed and accuracy.
`RUN_SPEED`	Prioritizes lower latency.
`RUN_ACCURACY`	Prioritizes higher precision.

Controls automatic target selection for ZeticMLangeLLMModel.

Value	Description
`.RUN_AUTO`	Default. Uses metadata-based selection to balance practical speed and quality.
`.RUN_SPEED`	Prioritizes lower latency.
`.RUN_ACCURACY`	Prioritizes higher score or lower loss.

let model = try ZeticMLangeLLMModel(
    personalKey: PERSONAL_KEY,
    name: MODEL_NAME,
    modelMode: .RUN_SPEED
)

Optional dataset hint for .RUN_ACCURACY.

Value	Description
`.MMLU`	Massive Multitask Language Understanding
`.TRUTHFULQA`	Truthfulness and informativeness benchmark
`.CNN_DAILYMAIL`	Summarization benchmark
`.GSM8K`	Grade-school math reasoning benchmark

let model = try ZeticMLangeLLMModel(
    personalKey: PERSONAL_KEY,
    name: MODEL_NAME,
    modelMode: .RUN_ACCURACY,
    dataSetType: .MMLU
)

Controls how the conversation KV cache is handled when it reaches capacity.

Value	Description
`.CLEAN_UP_ON_FULL`	Clears the conversation context when the KV cache is full. Default.
`.DO_NOT_CLEAN_UP`	Keeps the context as-is. You must call `cleanUp()` before the next conversation.

Configures LLM runtime initialization.

public struct LLMInitOption {
    public let kvCacheCleanupPolicy: LLMKVCacheCleanupPolicy
    public let nCtx: Int
}

Field	Type	Description
`kvCacheCleanupPolicy`	`LLMKVCacheCleanupPolicy`	KV-cache cleanup behavior during generation.
`nCtx`	`Int`	Requested context length.

nCtx is not guaranteed to be applied exactly as provided. The runtime can adjust it internally depending on the model or backend.

Manual runtime selection target for ZeticMLangeLLMModel.

LLMTarget.LLAMA_CPP

Value	Description
`.LLAMA_CPP`	Use this value for explicit GGUF selection.

GGUF quantization types used with .LLAMA_CPP.

Processor type used with explicit runtime selection.

Device / runtime	Supported values
Apple + `.LLAMA_CPP`	`.CPU`, `.GPU`

Apple LLaMA.cpp does not support .NPU.

Managed artifact cache policy for downloaded model files.

Value	Description
`.REMOVE_OVERLAPPING`	Default. Removes overlapping cached aliases for the same model entry.
`.KEEP_EXISTING`	Keeps older cached aliases and artifacts.

Detailed semantics for ZeticMLangeCacheHandlingPolicy are currently TBD. See Cache Management.