Struct OllamaGenerationParams 

pub struct OllamaGenerationParams {
    pub num_predict: Option<u32>,
    pub temperature: Option<f32>,
    pub top_p: Option<f32>,
    pub top_k: Option<u32>,
    pub stop: Option<Vec<String>>,
    pub repeat_penalty: Option<f32>,
    pub num_ctx: Option<u32>,
    pub keep_alive: Option<String>,
    pub context: Option<Vec<i64>>,
}

Generation parameters for Ollama requests

Fields

num_predict: Option<u32>

Maximum tokens to generate

temperature: Option<f32>

Temperature for sampling (0.0 - 1.0)

top_p: Option<f32>

Top-p nucleus sampling threshold

top_k: Option<u32>

Top-k sampling

stop: Option<Vec<String>>

Stop sequences

repeat_penalty: Option<f32>

Repeat penalty

num_ctx: Option<u32>

Context window size in tokens.

Critical for long documents: Ollama silently truncates prompts that exceed the default context size (often 2k-8k tokens). Set this to accommodate the full document + chunk + instructions when using Contextual Enrichment.

For KV Cache efficiency, calculate as: tokens(instructions) + tokens(document) + tokens(max_chunk) + output_tokens + 5% margin

keep_alive: Option<String>

How long to keep the model loaded in memory after the request (e.g. “1h”, “30m”, “0”).

Critical for KV Cache: Without this, Ollama may unload the model between consecutive requests, destroying the KV cache and forcing a full re-evaluation of the static document prefix for every chunk. Set to “1h” when processing multiple chunks from the same document.

This is a top-level Ollama API field, not an option — serialized separately.

context: Option<Vec<i64>>

KV cache context from a previous /api/generate response.

When set, the model continues from this token state instead of re-evaluating the entire prompt. Use this for the two-step KV cache pattern:

1. Prime: send the full document, get context back (loads doc into KV cache)
2. Per chunk: send only the chunk text with the priming context → Ollama skips document re-evaluation, only evaluates ~128 chunk tokens

This is a top-level Ollama API field — serialized separately.

Trait Implementations

impl Clone for OllamaGenerationParams

fn clone(&self) -> OllamaGenerationParams

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for OllamaGenerationParams

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

Formats the value using the given formatter.

impl Default for OllamaGenerationParams

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for OllamaGenerationParams

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for OllamaGenerationParams

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.