refactoring: continuing to break stuff out

src/retrieval/vector.rs

@@ -0,0 +1,45 @@
+use surrealdb::{engine::remote::ws::Client, Surreal};
+
+use crate::{error::ProcessingError, utils::embedding::generate_embedding};
+
+/// Compares vectors and retrieves a number of items from the specified table.
+///
+/// This function generates embeddings for the input text, constructs a query to find the closest matches in the database,
+/// and then deserializes the results into the specified type `T`.
+///
+/// # Arguments
+///
+/// * `take`: The number of items to retrieve from the database.
+/// * `input_text`: The text to generate embeddings for.
+/// * `db_client`: The SurrealDB client to use for querying the database.
+/// * `table`: The table to query in the database.
+/// * `openai_client`: The OpenAI client to use for generating embeddings.
+///
+/// # Returns
+///
+/// A vector of type `T` containing the closest matches to the input text. Returns a `ProcessingError` if an error occurs.
+///
+/// # Type Parameters
+///
+/// * `T`: The type to deserialize the query results into. Must implement `serde::Deserialize`.
+pub async fn find_items_by_vector_similarity<T>(
+    take: u8,
+    input_text: String,
+    db_client: &Surreal<Client>,
+    table: String,
+    openai_client: &async_openai::Client<async_openai::config::OpenAIConfig>,
+) -> Result<Vec<T>, ProcessingError>
+where
+    T: for<'de> serde::Deserialize<'de>,
+{
+    // Generate embeddings
+    let input_embedding = generate_embedding(openai_client, input_text).await?;
+
+    // Construct the query
+    let closest_query = format!("SELECT *, vector::distance::knn() AS distance FROM {} WHERE embedding <|{},40|> {:?} ORDER BY distance", table, take, input_embedding);
+
+    // Perform query and deserialize to struct
+    let closest_entities: Vec<T> = db_client.query(closest_query).await?.take(0)?;
+
+    Ok(closest_entities)
+}