Struct NumCtxInputs 

pub struct NumCtxInputs {
    pub model_max: Option<usize>,
    pub dims: Option<ModelDims>,
    pub model_weight_bytes: Option<u64>,
    pub per_model_override: Option<u32>,
    pub global_num_ctx: Option<i32>,
    pub allow_ram_offload: bool,
    pub vram_bytes: Option<u64>,
    pub system_ram_bytes: Option<u64>,
    pub max_auto_cap: Option<usize>,
    pub is_cloud: bool,
}

Everything resolve_ollama_num_ctx needs. Built identically at both call sites so the effective window stays consistent.

Fields

model_max: Option<usize>

The model’s architectural max window (from /api/show). Without it, auto can’t run and the resolver returns None (caller omits num_ctx).

dims: Option<ModelDims>

Architecture dimensions for the KV-cache estimate.

model_weight_bytes: Option<u64>

Model weight bytes (from /api/show size), subtracted from the budget.

per_model_override: Option<u32>

Per-model override (/context <n>); highest precedence.

global_num_ctx: Option<i32>

Global [ollama] num_ctx; beats auto, loses to the override.

allow_ram_offload: bool

When true, auto-fit budgets against system RAM (allows the slow GPU/CPU split); when false (default), budgets against VRAM so the model stays on the GPU.

vram_bytes: Option<u64>

Total VRAM in bytes (best-effort), used as the budget when offload is off.

system_ram_bytes: Option<u64>

Total system RAM in bytes, used as the budget when offload is on.

max_auto_cap: Option<usize>

Optional hard cap on the auto value ([ollama] max_auto_num_ctx).

is_cloud: bool

True for a cloud-served (:cloud) model. These run on Ollama’s servers, not the local GPU, so they bypass the VRAM auto-fit and use the model’s full advertised window.

Trait Implementations

impl Clone for NumCtxInputs

fn clone(&self) -> NumCtxInputs

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for NumCtxInputs

impl Debug for NumCtxInputs

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl Default for NumCtxInputs

fn default() -> NumCtxInputs

Returns the “default value” for a type.