49371b44cb
This introduces semantic indexing in Zed based on chunking text from files in the developer's workspace and creating vector embeddings using an embedding model. As part of this, we've created an embeddings provider trait that allows us to work with OpenAI, a local Ollama model, or a Zed hosted embedding. The semantic index is built by breaking down text for known (programming) languages into manageable chunks that are smaller than the max token size. Each chunk is then fed to a language model to create a high dimensional vector which is then normalized to a unit vector to allow fast comparison with other vectors with a simple dot product. Alongside the vector, we store the path of the file and the range within the document where the vector was sourced from. Zed will soon grok contextual similarity across different text snippets, allowing for natural language search beyond keyword matching. This is being put together both for human-based search as well as providing results to Large Language Models to allow them to refine how they help developers. Remaining todo: * [x] Change `provider` to `model` within the zed hosted embeddings database (as its currently a combo of the provider and the model in one name) Release Notes: - N/A --------- Co-authored-by: Nathan Sobo <nathan@zed.dev> Co-authored-by: Antonio Scandurra <me@as-cii.com> Co-authored-by: Conrad Irwin <conrad@zed.dev> Co-authored-by: Marshall Bowers <elliott.codes@gmail.com> Co-authored-by: Antonio <antonio@zed.dev>
94 lines
3.2 KiB
Rust
94 lines
3.2 KiB
Rust
use super::*;
|
|
use time::Duration;
|
|
use time::OffsetDateTime;
|
|
|
|
impl Database {
|
|
pub async fn get_embeddings(
|
|
&self,
|
|
model: &str,
|
|
digests: &[Vec<u8>],
|
|
) -> Result<HashMap<Vec<u8>, Vec<f32>>> {
|
|
self.weak_transaction(|tx| async move {
|
|
let embeddings = {
|
|
let mut db_embeddings = embedding::Entity::find()
|
|
.filter(
|
|
embedding::Column::Model.eq(model).and(
|
|
embedding::Column::Digest
|
|
.is_in(digests.iter().map(|digest| digest.as_slice())),
|
|
),
|
|
)
|
|
.stream(&*tx)
|
|
.await?;
|
|
|
|
let mut embeddings = HashMap::default();
|
|
while let Some(db_embedding) = db_embeddings.next().await {
|
|
let db_embedding = db_embedding?;
|
|
embeddings.insert(db_embedding.digest, db_embedding.dimensions);
|
|
}
|
|
embeddings
|
|
};
|
|
|
|
if !embeddings.is_empty() {
|
|
let now = OffsetDateTime::now_utc();
|
|
let retrieved_at = PrimitiveDateTime::new(now.date(), now.time());
|
|
|
|
embedding::Entity::update_many()
|
|
.filter(
|
|
embedding::Column::Digest
|
|
.is_in(embeddings.keys().map(|digest| digest.as_slice())),
|
|
)
|
|
.col_expr(embedding::Column::RetrievedAt, Expr::value(retrieved_at))
|
|
.exec(&*tx)
|
|
.await?;
|
|
}
|
|
|
|
Ok(embeddings)
|
|
})
|
|
.await
|
|
}
|
|
|
|
pub async fn save_embeddings(
|
|
&self,
|
|
model: &str,
|
|
embeddings: &HashMap<Vec<u8>, Vec<f32>>,
|
|
) -> Result<()> {
|
|
self.weak_transaction(|tx| async move {
|
|
embedding::Entity::insert_many(embeddings.iter().map(|(digest, dimensions)| {
|
|
let now_offset_datetime = OffsetDateTime::now_utc();
|
|
let retrieved_at =
|
|
PrimitiveDateTime::new(now_offset_datetime.date(), now_offset_datetime.time());
|
|
|
|
embedding::ActiveModel {
|
|
model: ActiveValue::set(model.to_string()),
|
|
digest: ActiveValue::set(digest.clone()),
|
|
dimensions: ActiveValue::set(dimensions.clone()),
|
|
retrieved_at: ActiveValue::set(retrieved_at),
|
|
}
|
|
}))
|
|
.on_conflict(
|
|
OnConflict::columns([embedding::Column::Model, embedding::Column::Digest])
|
|
.do_nothing()
|
|
.to_owned(),
|
|
)
|
|
.exec_without_returning(&*tx)
|
|
.await?;
|
|
Ok(())
|
|
})
|
|
.await
|
|
}
|
|
|
|
pub async fn purge_old_embeddings(&self) -> Result<()> {
|
|
self.weak_transaction(|tx| async move {
|
|
embedding::Entity::delete_many()
|
|
.filter(
|
|
embedding::Column::RetrievedAt
|
|
.lte(OffsetDateTime::now_utc() - Duration::days(60)),
|
|
)
|
|
.exec(&*tx)
|
|
.await?;
|
|
|
|
Ok(())
|
|
})
|
|
.await
|
|
}
|
|
}
|