added endpoint_embeddings_style

src/caps.rs

@@ -48,6 +48,8 @@ pub struct CodeAssistantCaps {
     pub default_embeddings_model: String,
     #[serde(default)]
     pub endpoint_embeddings_template: String,
+    #[serde(default)]
+    pub endpoint_embeddings_style: String,
     pub running_models: Vec<String>,
     #[serde(default)]
     pub caps_version: i64,  // need to reload if it increases on server, that happens when server configuration changes
@@ -279,6 +281,7 @@ const HF_DEFAULT_CAPS: &str = r#"
 
     "default_embeddings_model": "BAAI/bge-small-en-v1.5",
     "endpoint_embeddings_template": "https://api-inference.huggingface.co/models/$MODEL",
+    "endpoint_embeddings_style": "hf",
 
     "tokenizer_path_template": "https://huggingface.co/$MODEL/resolve/main/tokenizer.json",
     "tokenizer_rewrite_path": {

src/fetch_embedding.rs

@@ -7,20 +7,18 @@ use crate::forward_to_openai_endpoint::get_embedding_openai_style;
 
 
 pub async fn get_embedding(
-    address_url: &String,
+    endpoint_embeddings_style: &String,
     model_name: &String,
     endpoint_template: &String,
     text: String,
     api_key: &String,
 ) -> Result<Vec<f32>, String> {
-    match address_url.to_lowercase().as_str() {
+    match endpoint_embeddings_style.to_lowercase().as_str() {
         "hf" => Ok(get_embedding_hf_style(text, endpoint_template, model_name, api_key).await?),
-        url if url == "refact" || url.starts_with("http") => {
-            Ok(get_embedding_openai_style(text, endpoint_template, model_name, api_key).await?)
-        }
+        "openai" => Ok(get_embedding_openai_style(text, endpoint_template, model_name, api_key).await?),
         _ => {
-            error!("Invalid address_url: {}", address_url);
-            Err("Invalid address_url".to_string())
+            error!("Invalid endpoint_embeddings_style: {}", endpoint_embeddings_style);
+            Err("Invalid endpoint_embeddings_style".to_string())
         }
     }
 }

src/lsp.rs

@@ -171,6 +171,7 @@ impl Backend {
 #[tower_lsp::async_trait]
 impl LanguageServer for Backend {
     async fn initialize(&self, params: InitializeParams) -> Result<InitializeResult> {
+        // TODO: vecdb can be initialized after LSP, so retrieve_files_by_proj_folders won't start
         info!("LSP client_info {:?}", params.client_info);
         {
             let gcx_locked = self.gcx.write().await;

src/vecdb/vecdb.rs

@@ -69,12 +69,13 @@ pub async fn vecdb_background_reload(
             continue;
         }
 
-        let (default_embeddings_model, endpoint_embeddings_template) = {
+        let (default_embeddings_model, endpoint_embeddings_template, endpoint_embeddings_style) = {
             let caps = caps_mb.unwrap();
             let caps_locked = caps.read().unwrap();
             (
                 caps_locked.default_embeddings_model.clone(),
-                caps_locked.endpoint_embeddings_template.clone()
+                caps_locked.endpoint_embeddings_template.clone(),
+                caps_locked.endpoint_embeddings_style.clone(),
             )
         };
 
@@ -84,7 +85,11 @@ pub async fn vecdb_background_reload(
         }
 
         let vecdb_mb = create_vecdb_if_caps_present(
-            default_embeddings_model, endpoint_embeddings_template, cmdline, cache_dir
+            default_embeddings_model,
+            endpoint_embeddings_template,
+            endpoint_embeddings_style,
+            cmdline,
+            cache_dir
         ).await;
 
         if vecdb_mb.is_none() {
@@ -106,6 +111,7 @@ pub async fn vecdb_background_reload(
 pub async fn create_vecdb_if_caps_present(
     default_embeddings_model: String,
     endpoint_embeddings_template: String,
+    endpoint_embeddings_style: String,
 
     cmdline: &CommandLine,
     cache_dir: &PathBuf,
@@ -115,6 +121,7 @@ pub async fn create_vecdb_if_caps_present(
         384, 60, 512, 1024,
         default_embeddings_model.clone(),
         endpoint_embeddings_template.clone(),
+        endpoint_embeddings_style.clone(),
     ).await {
         Ok(res) => Some(res),
         Err(err) => {
@@ -143,6 +150,7 @@ impl VecDb {
 
         model_name: String,
         endpoint_template: String,
+        endpoint_embeddings_style: String,
     ) -> Result<VecDb, String> {
         let handler = match VecDBHandler::init(cache_dir, embedding_size).await {
             Ok(res) => res,
@@ -157,7 +165,7 @@ impl VecDb {
 
             model_name.clone(),
             cmdline.api_key.clone(),
-            cmdline.address_url.clone(),
+            endpoint_embeddings_style.clone(),
             endpoint_template.clone(),
         ).await));
 

src/vecdb/vectorizer_service.rs

@@ -24,7 +24,7 @@ pub struct FileVectorizerService {
 
     model_name: String,
     api_key: String,
-    address_url: String,
+    endpoint_embeddings_style: String,
     endpoint_template: String,
 }
 
@@ -73,7 +73,7 @@ async fn vectorize_thread(
 
     model_name: String,
     api_key: String,
-    address_url: String,
+    endpoint_embeddings_style: String,
     endpoint_template: String,
 
     max_concurrent_tasks: usize,
@@ -120,7 +120,7 @@ async fn vectorize_thread(
         let join_handles: Vec<_> = split_data_filtered.into_iter().map(|x| {
             let model_name_clone = model_name.clone();
             let api_key_clone = api_key.clone();
-            let address_url_clone = address_url.clone();
+            let endpoint_embeddings_style_clone = endpoint_embeddings_style.clone();
             let endpoint_template_clone = endpoint_template.clone();
 
             let semaphore_clone = Arc::clone(&semaphore);
@@ -133,7 +133,7 @@ async fn vectorize_thread(
                 };
 
                 let result = get_embedding(
-                    &address_url_clone,
+                    &endpoint_embeddings_style_clone,
                     &model_name_clone,
                     &endpoint_template_clone,
                     x.window_text.clone(),
@@ -203,7 +203,7 @@ impl FileVectorizerService {
 
         model_name: String,
         api_key: String,
-        address_url: String,
+        endpoint_embeddings_style: String,
         endpoint_template: String,
     ) -> Self {
         let update_request_queue = Arc::new(Mutex::new(VecDeque::new()));
@@ -227,7 +227,7 @@ impl FileVectorizerService {
 
             model_name,
             api_key,
-            address_url,
+            endpoint_embeddings_style,
             endpoint_template,
         }
     }
@@ -252,7 +252,7 @@ impl FileVectorizerService {
 
                 self.model_name.clone(),
                 self.api_key.clone(),
-                self.address_url.clone(),
+                self.endpoint_embeddings_style.clone(),
                 self.endpoint_template.clone(),
 
                 4,