Struct AskConfig 

pub struct AskConfig {

    pub k_summary: u32,
    pub k_raw: u32,
    pub escalation_threshold: f64,
    pub mmr_threshold: f64,
    pub max_context_tokens: u32,
    pub response_tokens: u32,
    pub timeout_secs: u32,
    pub min_score: f64,
    pub continue_history_turns: u32,
    pub rewriter_timeout_secs: u32,
    pub compress_hits_enabled: bool,
    pub summarize_hits_enabled: bool,
    pub summarize_model: Option<String>,
    pub backend: Option<BackendConfig>,
    pub rewriter_backend: Option<BackendConfig>,
}

Fields

k_summary: u32

k_raw: u32

escalation_threshold: f64

mmr_threshold: f64

max_context_tokens: u32

response_tokens: u32

timeout_secs: u32

min_score: f64

continue_history_turns: u32

rewriter_timeout_secs: u32

Separate, shorter timeout for the rewriter LLM call (Phase 3.3). Rewriter output is small (~80 tokens) and falling back to the raw question on failure is non-fatal, so we don’t want to burn the full timeout_secs budget waiting on a slow/unreachable Ollama before the user sees any response.

compress_hits_enabled: bool

summarize_hits_enabled: bool

summarize_model: Option<String>

backend: Option<BackendConfig>

Per-stage backend override for the answer-generation model. None = inherit the smart slot (config.llm).

rewriter_backend: Option<BackendConfig>

Per-stage backend override for the query rewriter. None = inherit the answer stage’s backend (self.backend), falling through to the smart slot (config.llm) only if that is also unset.

Implementations

impl AskConfig

pub fn effective_backend(&self, llm: &LlmConfig) -> BackendConfig

Effective backend for answer generation. An explicit per-stage override wins; otherwise the stage inherits the smart slot (config.llm) with this stage’s own timeout baked in, so a slow backend cannot silently fall back to the factory’s 120s default.

pub fn effective_rewriter_backend(&self, llm: &LlmConfig) -> BackendConfig

Effective backend for the query rewriter. An explicit rewriter_backend override wins outright. Otherwise the rewriter follows the answer stage’s backend (self.backend), and only falls through to the smart slot (llm) when the answer stage has no override either — the rewriter is not an independent pinning point, only an independent timeout. Whichever source it resolves from, it always keeps its own much tighter rewriter_timeout_secs budget: the rewriter’s output is small and falling back to the raw question on timeout is non-fatal.

Trait Implementations

impl Clone for AskConfig

fn clone(&self) -> AskConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for AskConfig

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

Formats the value using the given formatter.

impl Default for AskConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for AskConfig

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for AskConfig

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

where __S: Serializer,

Serialize this value into the given Serde serializer.