Struct LlmConfig 

pub struct LlmConfig {

    pub model: String,
    pub resolved_provider_id: Option<String>,
    pub host_provider_kind: Option<String>,
    pub host_supplied_connection: bool,
    pub reasoning_effort: Option<ReasoningEffort>,
    pub api_key: Option<String>,
    pub api_key_env: Option<String>,
    pub base_url: Option<String>,
    pub protocol: String,
    pub parallel_tool_calls: Option<bool>,
    pub timeout_secs: u64,
    pub max_retries: u32,
    pub cache: Option<LlmCacheConfig>,
    pub budget: Option<LlmBudgetConfig>,
    pub rate_limit: Option<LlmRateLimitConfig>,
    pub cost_tracking: bool,
    pub tracing: bool,
    pub cooldown_secs: Option<u64>,
    pub health_check_secs: Option<u64>,
    pub thinking_budget: Option<u32>,
    pub embedding: EmbeddingBackendConfig,
    pub provider_keys: HashMap<String, String>,
}

Unified LLM configuration for the liter-llm client.

Settings are loaded from all layers via load_llm_config: compiled defaults → config.yaml llm: section → UAR_LLM__* env vars → LLM_* legacy env vars → CLI arguments.

Fields

model: String

Default model in provider/model format (e.g., "openai/gpt-4o").

resolved_provider_id: Option<String>

Authoritative provider identity attached by runtime provider resolution.

This is runtime metadata rather than a user-configured field. It keeps provider identity available when an explicit base URL requires a bare model name on the wire.

host_provider_kind: Option<String>

Wire protocol selected by a host-supplied run credential. The provider id remains the host’s stable identity and may differ from this protocol.

host_supplied_connection: bool

Marks a connection assembled from request-scoped secret material. Its endpoint is redacted from Debug and shared health state is not attached.

reasoning_effort: Option<ReasoningEffort>

Request-scoped reasoning override inherited by local child runs.

api_key: Option<String>

API key for the default provider.

api_key_env: Option<String>

Environment variable name that contains the API key for the default provider.

This lets deployments keep provider-specific secrets in .env without copying the secret value into config.yaml.

base_url: Option<String>

Base URL override (bypasses provider auto-detection).

protocol: String

Protocol hint: "auto" | "chat" | "responses".

parallel_tool_calls: Option<bool>

Whether to enable parallel tool calls.

timeout_secs: u64

Request timeout in seconds.

max_retries: u32

Maximum retries on 429/5xx.

cache: Option<LlmCacheConfig>

Cache configuration for Tower middleware.

budget: Option<LlmBudgetConfig>

Budget enforcement configuration.

rate_limit: Option<LlmRateLimitConfig>

Rate limiting configuration.

cost_tracking: bool

Enable per-request cost tracking.

tracing: bool

Enable OpenTelemetry tracing spans.

cooldown_secs: Option<u64>

Cooldown after transient errors (seconds).

health_check_secs: Option<u64>

Health check interval (seconds).

thinking_budget: Option<u32>

Extended thinking budget (tokens). When set, enables thinking for Anthropic models with the specified token budget.

embedding: EmbeddingBackendConfig

Embedding backend configuration for RAG and memory.

provider_keys: HashMap<String, String>

Per-provider API keys loaded from env shortcuts (OPENAI_API_KEY, etc.).

Keyed by provider id (e.g. "openai", "anthropic"). Populated at config-load time from well-known env vars. Used by the provider registry to compute configured status (key present → truly configured) and to pass the right key when a non-default provider is used.

NOT read from config.yaml — always sourced from env at boot.

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 Default for LlmConfig

fn default() -> Self

Returns the “default value” for a type.

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 JsonSchema for LlmConfig

fn schema_name() -> Cow<'static, str>

The name of the generated JSON Schema.

fn schema_id() -> Cow<'static, str>

Returns a string that uniquely identifies the schema produced by this type.

fn json_schema(generator: &mut SchemaGenerator) -> Schema

Generates a JSON Schema for this type.

fn inline_schema() -> bool

Whether JSON Schemas generated for this type should be included directly in parent schemas, rather than being re-used where possible using the $ref keyword.

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.