Fix smart search: use EmbeddingGemma instead of Ollama mxbai

src/api/search.rs

@@ -5,6 +5,8 @@ use axum::{extract::State, http::StatusCode, response::Json, routing::post, Rout
 use serde::{Deserialize, Serialize};
 use serde_json;
 
+use crate::core::embedding::Embedder;
+
 // --- Request / Response Structures ---
 
 #[derive(Debug, Deserialize)]
@@ -51,8 +53,9 @@ pub async fn smart_search(
     let db = &state.db;
     let limit = req.limit.unwrap_or(5);
 
-    // 1. Generate Embedding using Ollama
-    let embedding = get_ollama_embedding(&req.query).await.map_err(
+    // 1. Generate Embedding using EmbeddingGemma via MOMENTRY_EMBED_URL
+    let embedder = Embedder::new("embeddinggemma-300m".to_string());
+    let embedding = embedder.embed_query(&req.query).await.map_err(
         |e| -> (StatusCode, Json<serde_json::Value>) {
             tracing::error!("Embedding failed: {}", e);
             (
@@ -155,36 +158,6 @@ pub async fn smart_search(
     Ok(Json(response))
 }
 
-// --- Helper: Ollama Embedding ---
-
-async fn get_ollama_embedding(
-    text: &str,
-) -> Result<Vec<f32>, Box<dyn std::error::Error + Send + Sync>> {
-    let client = reqwest::Client::new();
-    let payload = serde_json::json!({
-        "model": "mxbai-embed-large",
-        "prompt": text
-    });
-
-    let res = client
-        .post("http://localhost:11434/api/embeddings")
-        .json(&payload)
-        .send()
-        .await?
-        .json::<serde_json::Value>()
-        .await?;
-
-    // Parse embedding array from response
-    let embedding = res["embedding"]
-        .as_array()
-        .ok_or("No embedding found in Ollama response")?
-        .iter()
-        .map(|v| v.as_f64().unwrap_or(0.0) as f32)
-        .collect();
-
-    Ok(embedding)
-}
-
 // --- Router Setup ---
 
 pub fn search_routes() -> Router<crate::api::server::AppState> {