Enum QueryStep 

pub enum QueryStep {
    Search {
        query: String,
        limit: usize,
    },
    VectorSearch {
        query: String,
        limit: usize,
    },
    SemanticSearch {
        text: String,
        limit: usize,
    },
    RawVectorSearch {
        vec: Vec<f32>,
        limit: usize,
    },
    TextSearch {
        query: TextQuery,
        limit: usize,
    },
    Traverse {
        direction: TraverseDirection,
        label: String,
        max_depth: usize,
        filter: Option<Predicate>,
    },
    Filter(Predicate),
}

A single step in the query pipeline.

Eq is intentionally NOT derived — see QueryAst for the rationale.

Variants

Search

Unified adaptive retrieval entry step consumed by the Phase 12 retrieval planner.

Carries the caller’s raw query string (not a parsed TextQuery): the planner decides how to interpret and route it across the text strict, text relaxed, and (future) vector branches. See crate::compile_retrieval_plan for the planner entry point.

Fields

query: String

The raw caller-supplied query string.

limit: usize

Maximum number of candidate rows requested by the caller.

VectorSearch

Nearest-neighbor search over vector embeddings.

Fields

query: String

The search query text (to be embedded by the caller).

limit: usize

Maximum number of candidate rows from the vector index.

SemanticSearch

Pack F1 semantic-search step: a natural-language query that the engine embeds at query time using the db-wide active profile embedder, then runs KNN against the per-kind vec_<kind> table.

Unlike QueryStep::VectorSearch, the text is NEVER a JSON float-array literal — the caller supplies natural language and the engine handles embedding internally.

Fields

text: String

Natural-language query string to embed at query time.

limit: usize

Maximum number of candidate rows from the vector KNN scan.

RawVectorSearch

Pack F1 raw-vector-search step: a caller-supplied dense vector that the engine binds directly to the per-kind vec_<kind> KNN scan with no embedder call. The vector’s dimension must match the active embedding profile’s dimension or the coordinator returns super::CompileError-free with a hard DimensionMismatch error at plan-time.

Fields

vec: Vec<f32>

Caller-supplied dense vector.

limit: usize

Maximum number of candidate rows from the vector KNN scan.

TextSearch

Full-text search over indexed chunk content using FathomDB’s supported safe text-query subset.

Fields

query: TextQuery

Parsed text-search intent to be lowered into safe FTS5 syntax.

limit: usize

Maximum number of candidate rows from the FTS index.

Traverse

Graph traversal following edges of the given label.

Fields

direction: TraverseDirection

Direction to traverse.

label: String

Edge kind to follow.

max_depth: usize

Maximum hops from each candidate.

filter: Option<Predicate>

Optional predicate to filter traversal results. None is exactly equivalent to the pre-Pack-2 behavior. Some(_) is not yet implemented; see Pack 3.

Filter(Predicate)

Row-level filter predicate.

Trait Implementations

impl Clone for QueryStep

fn clone(&self) -> QueryStep

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QueryStep

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

Formats the value using the given formatter.

impl PartialEq for QueryStep

fn eq(&self, other: &QueryStep) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for QueryStep