Struct CollectionContract 

pub struct CollectionContract {

    pub name: String,
    pub declared_model: CollectionModel,
    pub schema_mode: SchemaMode,
    pub origin: ContractOrigin,
    pub version: u32,
    pub created_at_unix_ms: u128,
    pub updated_at_unix_ms: u128,
    pub default_ttl_ms: Option<u64>,
    pub vector_dimension: Option<usize>,
    pub vector_metric: Option<DistanceMetric>,
    pub context_index_fields: Vec<String>,
    pub declared_columns: Vec<DeclaredColumnContract>,
    pub table_def: Option<TableDef>,
    pub timestamps_enabled: bool,
    pub context_index_enabled: bool,
    pub metrics_raw_retention_ms: Option<u64>,
    pub metrics_rollup_policies: Vec<String>,
    pub metrics_tenant_identity: Option<String>,
    pub metrics_namespace: Option<String>,
    pub append_only: bool,
    pub subscriptions: Vec<SubscriptionDescriptor>,
}

Fields

name: String

declared_model: CollectionModel

schema_mode: SchemaMode

origin: ContractOrigin

version: u32

created_at_unix_ms: u128

updated_at_unix_ms: u128

default_ttl_ms: Option<u64>

vector_dimension: Option<usize>

vector_metric: Option<DistanceMetric>

context_index_fields: Vec<String>

declared_columns: Vec<DeclaredColumnContract>

table_def: Option<TableDef>

timestamps_enabled: bool

Enabled by CREATE TABLE ... WITH timestamps = true. When true, the runtime auto-populates two user-visible columns created_at + updated_at (BIGINT unix-ms) sourced from the UnifiedEntity::created_at/updated_at fields. created_at is immutable after insert; updated_at is bumped on every mutation.

context_index_enabled: bool

Enabled by CREATE TABLE ... WITH context_index = true (or by naming specific context_index_fields). When true, every INSERT tokenises the row’s text fields and populates the global context index that backs SEARCH CONTEXT / SEARCH SIMILAR TEXT / ASK (RAG). When false (default), inserts skip the tokenisation + 3-way RwLock write storm entirely — ~800 ns faster per insert, and SEARCH returns empty for this collection.

Opt-in by design: pure OLTP tables (accounts, orders, events) pay zero indexing tax; search-oriented tables (articles, docs) flip the switch at CREATE time.

metrics_raw_retention_ms: Option<u64>

Metrics collections are backed by time-series storage but carry a metrics-specific raw sample retention contract.

metrics_rollup_policies: Vec<String>

Metrics rollup tiers declared by CREATE METRICS ... DOWNSAMPLE.

metrics_tenant_identity: Option<String>

Metrics tenant identity source. Defaults to current tenant context and can be declared as a stable identity path for future ingestion slices.

metrics_namespace: Option<String>

Metrics namespace identity. v0 starts with a default namespace so series identity is namespace-aware before Prometheus ingestion exists.

append_only: bool

Enabled by CREATE TABLE ... APPEND ONLY or WITH (append_only = true). When true, the runtime rejects UPDATE and DELETE against this collection at parse time with a clear error — the operator’s immutability intent becomes a first-class catalog fact rather than an RLS-shaped approximation. Default false so legacy DDL keeps its mutable semantics.

subscriptions: Vec<SubscriptionDescriptor>

Declarative subscriptions created by WITH EVENTS. This is metadata only in #291; event emission is wired by the outbox slice.

Trait Implementations

impl Clone for CollectionContract

fn clone(&self) -> CollectionContract

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for CollectionContract

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

Formats the value using the given formatter.