Struct LlmConfig 

pub struct LlmConfig {
    pub provider: String,
    pub model: String,
    pub base_url: String,
    pub temperature: f64,
    pub max_tokens: u32,
    pub context_window: usize,
    pub api_key: String,
    pub api_key_file: Option<PathBuf>,
    pub providers: Vec<ProviderEntry>,
}

Fields

provider: String

👎Deprecated:

Set llm.providers[] instead. Single-provider mode is still functional but no longer the recommended shape.

Legacy single-provider selector. Superseded by providers[], which supports multi-provider failover and runtime health checks. Still honoured as the implicit single entry when providers[] is empty, and Embedder::from_config reads it to pick the embedding transport — so it can’t be removed yet. New configs should leave this set to a reasonable default and drive everything from providers[] instead.

model: String

👎Deprecated:

Set llm.providers[].model (and optionally preferred_models) instead.

Legacy single-provider model name. Superseded by per-entry llm.providers[].model + preferred_models[]. Still consulted when providers[] is empty.

base_url: String

👎Deprecated:

Set llm.providers[].base_url instead. Embedder transport selection still reads this field as a fallback.

Legacy single-provider endpoint. Superseded by per-entry llm.providers[].base_url. Still consulted when providers[] is empty and by Embedder::from_config to pick the embedding transport.

temperature: f64

max_tokens: u32

context_window: usize

The active model’s input context window, in tokens. Drives the prompt assembler’s TokenBudget so a large-window model (e.g. 128k) reads far more file/attachment + memory content instead of being clipped to the conservative 8k default. Set this to your model’s real context size. Defaults to 8192 (safe for most models) when omitted, preserving the historical budget.

api_key: String

👎Deprecated:

Move credentials to llm.providers[].api_key_file (or api_key_file here) — the YAML field gets backed up and replicated.

API key for the LLM provider (required for OpenAI, OpenRouter, etc.). Can also be set via BRAIN_LLM__API_KEY environment variable. Prefer api_key_file (chmod-0600) for secrets that shouldn’t live in YAML, or move credentials to llm.providers[].api_key_file.

api_key_file: Option<PathBuf>

Issue 125: path to a chmod-0600 file holding the API key. Preferred over api_key because the YAML config typically gets backed up, version-controlled, and replicated; a sibling file with restricted perms keeps the secret out of those flows. When both are set, api_key_file wins.

providers: Vec<ProviderEntry>

Optional multi-provider entries. When non-empty, startup probes each entry’s /models endpoint and selects the first reachable one whose preferred_models are live. When empty, the legacy single-provider fields above are used as-is.

Trait Implementations

impl Clone for LlmConfig

fn clone(&self) -> LlmConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LlmConfig

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for LlmConfig

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for LlmConfig

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

where __S: Serializer,

Serialize this value into the given Serde serializer.