Struct TierConfig 

pub struct TierConfig {
    pub tier: FeatureTier,
    pub embedding_model: Option<EmbeddingModel>,
    pub llm_model: Option<String>,
    pub cross_encoder: bool,
    pub max_memory_mb: usize,
}

Runtime configuration derived from a FeatureTier.

Fields

tier: FeatureTier

embedding_model: Option<EmbeddingModel>

llm_model: Option<String>

Default backend LLM model tag for this tier, or None for tiers that use no LLM (keyword / semantic). The value is the provider-agnostic compiled default (default_tier_llm_model); the operator-resolved backend/model is carried by ResolvedLlm via AppConfig::resolve_llm and can be ANY backend. Treated as an on/off gate at the call sites.

cross_encoder: bool

max_memory_mb: usize

Implementations

impl TierConfig

pub fn capabilities(&self) -> Capabilities

Produce a Capabilities (schema v2) report suitable for JSON serialisation. The MCP / HTTP handle_capabilities_with_conn wrapper overlays live runtime state (recall mode, reranker mode, embedder-loaded flag) and live DB counts (active rules, hook registrations, pending approvals) before the report goes on the wire.

v2 honesty patch (P1, v0.6.3.1): recall_mode_active and reranker_active start at conservative defaults (disabled / off); the wrapper updates them based on the runtime embedder

• reranker handles, not the configured tier values.

#1168 back-compat shim. Delegates to Self::capabilities_with_resolved with a ResolvedModels::from_tier_preset triple so the pre-#1168 wire shape is byte-equal for callers (legacy tests, migrate-tool diagnostics) that don’t load an operator AppConfig. Production wrappers MUST call Self::capabilities_with_resolved directly with AppConfig::resolve_models output — otherwise memory_capabilities.models.* drifts from the live LLM / embedder / reranker wiring.

pub fn capabilities_with_resolved( &self, models: &ResolvedModels, ) -> Capabilities

v0.7.x (issue #1168) — resolver-aware capabilities builder.

Identical to Self::capabilities except models.embedding / models.llm / models.cross_encoder come from the operator-resolved models triple (built via AppConfig::resolve_models) instead of the compiled tier preset. This is the production entry point used by every handle_capabilities_with_conn[_v3] wrapper post-#1168.

The display logic mirrors the boot banner (src/cli/boot.rs BootManifest::build): Ollama-backend LLM emits the bare model id (legacy banner shape); other backends emit backend:model. Embedder + reranker respect the tier-preset disable flag so the keyword tier still reports embedding="none" even if an operator left a stale [embeddings] block in their config.

Trait Implementations

impl Clone for TierConfig

fn clone(&self) -> TierConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TierConfig

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for TierConfig

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for TierConfig

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

where __S: Serializer,

Serialize this value into the given Serde serializer.