Struct MongoDbVectorIndex 

pub struct MongoDbVectorIndex<C>
where
    C: Send + Sync,
{ /* private fields */ }

Vector index over a MongoDB collection.

Queries are embedded with the same model that populated the collection, so results are meaningless under another model.

Example

use rig_mongodb::{MongoDbVectorIndex, SearchParams};
use rig_core::{providers::openai::{self, OpenAI}, vector_store::{VectorStoreIndex, VectorSearchRequest}};

#[derive(serde::Deserialize, serde::Serialize, Debug)]
struct WordDefinition {
    #[serde(rename = "_id")]
    id: String,
    definition: String,
    embedding: Vec<f64>,
}

let mongodb_client = mongodb::Client::with_uri_str("mongodb://localhost:27017").await?; // <-- replace with your mongodb uri.
let openai = OpenAI::from_env()?;

let collection = mongodb_client.database("db").collection::<WordDefinition>(""); // <-- replace with your mongodb collection.

// <-- replace with your embedding model.
let model = openai.embedding(openai::TEXT_EMBEDDING_ADA_002, None);
let index = MongoDbVectorIndex::new(
    collection,
    model,
    "vector_index", // <-- replace with the name of the index in your mongodb collection.
    SearchParams::new(), // <-- field name in `Document` that contains the embeddings.
)
.await?;

let req = VectorSearchRequest::builder()
    .query("My boss says I zindle too much, what does that mean?")
    .samples(1)
    .build();

// Query the index
let definitions = index
    .top_n::<WordDefinition>(req)
    .await?;

Implementations

impl<C> MongoDbVectorIndex<C>

where C: Send + Sync,

pub async fn new( collection: Collection<C>, model: impl Into<DynModel<Embedding>>, index_name: &str, search_params: SearchParams, ) -> Result<Self, VectorStoreError>

Creates an index handle after confirming the named search index exists and is queryable. The embedded field is taken from the index’s first defined field. Errors when the index is missing, not queryable, or defines no fields. See the MongoDB documentation on creating vector indexes.

Trait Implementations

impl<C> InsertDocuments for MongoDbVectorIndex<C>

where C: Send + Sync,

async fn insert_documents<Doc: Serialize + Embed + WasmCompatSend>( &self, documents: Vec<(Doc, Vec<Embedding>)>, ) -> Result<(), VectorStoreError>

Insert precomputed embeddings for each document.

impl<C> VectorStoreIndex for MongoDbVectorIndex<C>

where C: Sync + Send,

async fn top_n<T: DeserializeOwned + WasmCompatSend>( &self, req: VectorSearchRequest<MongoDbSearchFilter>, ) -> Result<Vec<VectorSearchResult<T>>, VectorStoreError>

Returns matches with the search score. The embedding field is projected out, so T must not require it.

async fn top_n_ids( &self, req: VectorSearchRequest<MongoDbSearchFilter>, ) -> Result<Vec<VectorSearchIdResult>, VectorStoreError>

Like top_n but projects only the id and score.

type Filter = MongoDbSearchFilter

The filter type for this backend.