Expand description
Derive macros for lvv, turning Rust structs into
vector points.
VectorDatabaseItemmakes a record type produce one point: the text to embed, a category, and the payload stored next to the vector.VectorDatabasemakes a container struct produce the points of every item it holds.
Both implement traits from
lvv::points.
The generated code refers to ::lvv, so the crate using the derives must
depend on lvv under that name. Enable lvv’s derive feature instead of
depending on this crate directly; it re-exports both macros next to the
traits they implement:
[dependencies]
lvv = { version = "0.5", features = ["derive"] }
serde = { version = "1", features = ["derive"] }§Example
ⓘ
use lvv::points::{VectorDatabase, VectorDatabaseItem};
use serde::{Deserialize, Serialize};
#[derive(Serialize, Deserialize, VectorDatabaseItem)]
struct Project {
#[lvv(description)]
name: String,
#[lvv(description)]
summary: String,
#[lvv(rename = "stack")]
technologies: Vec<String>,
#[lvv(skip)]
private_notes: String,
}
#[derive(VectorDatabase)]
struct Portfolio {
projects: Vec<Project>,
featured: Option<Project>,
#[lvv(skip)]
owner: String,
}
for draft in portfolio.point_drafts()? {
// Embed `draft.description`, store `draft.payload` next to the vector.
println!("{}: {}", draft.category, draft.description);
}For a complete lvv pipeline built on these derives, from reading files to
writing Qdrant collections, see lvv’s
examples/derive.rs.
Derive Macros§
- Vector
Database - Derives
VectorDatabasefor a struct whose fields holdVectorDatabaseItems. - Vector
Database Item - Derives
VectorDatabaseItemfor a struct, so each value becomes one vector point.