Struct VectorStore 

pub struct VectorStore { /* private fields */ }

In-memory vector store for retrieval-augmented generation.

Uses the Ollama /api/embed endpoint to compute embeddings and cosine similarity for nearest-neighbour search.

use std::sync::Arc;
use oxide_agent::rag::VectorStore;
use oxide_agent::client::HttpOllamaClient;

let client = Arc::new(HttpOllamaClient::new("http://localhost:11434"));
let mut store = VectorStore::new(client, "nomic-embed-text");

store.add_text("Rust ownership means one owner at a time.", Default::default()).await?;
let results = store.query("Who owns memory in Rust?", 3).await?;
println!("{}", results[0].content);

Implementations

impl VectorStore

pub fn new<C: OllamaClient + 'static>( client: Arc<C>, embed_model: impl Into<String>, ) -> Self

pub async fn add_text( &mut self, text: impl Into<String>, metadata: HashMap<String, String>, ) -> Result<(), OxideError>

Embed a single string and add it to the store.

pub async fn add_file(&mut self, path: &Path) -> Result<usize, OxideError>

Read a UTF-8 text file and add each non-empty line as a separate document.

pub async fn query( &self, query: impl Into<String>, top_k: usize, ) -> Result<Vec<SearchResult>, OxideError>

Return the top_k most similar documents to query, ranked by cosine similarity (highest first).

pub fn len(&self) -> usize

Number of documents in the store.

pub fn is_empty(&self) -> bool