Struct CandleConfig 

pub struct CandleConfig {
    pub source: String,
    pub local_path: String,
    pub filename: Option<String>,
    pub chat_template: String,
    pub device: String,
    pub embedding_repo: Option<String>,
    pub hf_token: Option<String>,
    pub generation: GenerationParams,
    pub inference_timeout_secs: u64,
}

Fields

source: String

local_path: String

filename: Option<String>

chat_template: String

device: String

embedding_repo: Option<String>

hf_token: Option<String>

Resolved HuggingFace Hub API token for authenticated model downloads.

Must be the token value — resolved by the caller before constructing this config.

generation: GenerationParams

inference_timeout_secs: u64

Maximum seconds to wait for each half of a single inference request.

The timeout is applied twice per chat() call: once for the channel send (waiting for a free slot) and once for the oneshot reply (waiting for the worker to finish). The effective maximum wall-clock wait per request is therefore 2 × inference_timeout_secs. CPU inference can be slow; 120s is a conservative default for large models, giving up to 240s total before an error is returned. Values of 0 are silently promoted to 1 at bootstrap.

Trait Implementations

impl Debug for CandleConfig

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for CandleConfig

fn deserialize<__D>( __deserializer: __D, ) -> Result<CandleConfig, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for CandleConfig

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

where __S: Serializer,

Serialize this value into the given Serde serializer.