Enum CompletionTokensParam 

pub enum CompletionTokensParam {
    MaxTokens,
    MaxCompletionTokens,
}

Which token-limit field to use in the API request body.

OpenAI introduced max_completion_tokens for reasoning (o*) models while keeping max_tokens for legacy ones. The OpenAiProvider infers the correct field from the model name via a prefix table, but that table cannot cover every fine-tuned or newly-released variant. Set this override when the inference would be wrong for your model.

Examples

use zeph_llm::openai::{CompletionTokensParam, OpenAiConfig, OpenAiProvider};

// Force max_completion_tokens for a fine-tuned reasoning model
// whose name does not start with "o" + digit.
let cfg = OpenAiConfig {
    api_key: "sk-...".into(),
    base_url: "https://api.openai.com/v1".into(),
    model: "my-ft-reasoner-v1".into(),
    max_tokens: 8192,
    embedding_model: None,
    reasoning_effort: None,
    context_window: None,
    completion_tokens_param: Some(CompletionTokensParam::MaxCompletionTokens),
    vision: None,
};
let provider = OpenAiProvider::new(cfg);

Variants

MaxTokens

Use the legacy max_tokens field (standard chat models).

MaxCompletionTokens

Use max_completion_tokens (required by o* reasoning models and gpt-5).

Trait Implementations

impl Clone for CompletionTokensParam

fn clone(&self) -> CompletionTokensParam

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for CompletionTokensParam

impl Debug for CompletionTokensParam

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

Formats the value using the given formatter.

impl Eq for CompletionTokensParam

impl PartialEq for CompletionTokensParam

fn eq(&self, other: &CompletionTokensParam) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for CompletionTokensParam