Skip to content

RuntimeConfig

Defined in: packages/ai.matey.types/src/model-runner.ts:151

Runtime configuration for model execution.

optional batchSize: number

Defined in: packages/ai.matey.types/src/model-runner.ts:175

Batch size for prompt processing.

512

optional contextSize: number

Defined in: packages/ai.matey.types/src/model-runner.ts:156

Context size (max tokens in context window).

2048

optional gpuLayers: number

Defined in: packages/ai.matey.types/src/model-runner.ts:163

Number of GPU layers to offload. -1 = all layers, 0 = CPU only

0

optional keepAlive: boolean

Defined in: packages/ai.matey.types/src/model-runner.ts:181

Keep model loaded in memory.

true

optional mlock: boolean

Defined in: packages/ai.matey.types/src/model-runner.ts:193

Lock model in memory (prevent swapping).

false

optional mmap: boolean

Defined in: packages/ai.matey.types/src/model-runner.ts:187

Memory map the model file.

true

optional threads: number

Defined in: packages/ai.matey.types/src/model-runner.ts:169

Number of threads to use.

(CPU cores)