Polish search and embedding UX

- add semantic text search with toolbar mode switching and sqlite-vec query support
- improve embedding progress visibility, failure recovery, and similar-image affordances
- add search clearing and keyboard controls for filename vs semantic search modes
- refine background task interactions and gallery/lightbox embedding states

Refs: #4

src-tauri/Cargo.lock

@@ -1369,6 +1369,9 @@ name = "esaxx-rs"
 version = "0.1.10"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d817e038c30374a4bcb22f94d0a8a0e216958d4c3dcde369b1439fec4bdda6e6"
+dependencies = [
+ "cc",
+]
 
 [[package]]
 name = "event-listener"
@@ -3962,6 +3965,7 @@ dependencies = [
  "tauri-plugin-dialog",
  "tauri-plugin-fs",
  "tauri-plugin-opener",
+ "tokenizers",
  "tokio",
  "uuid",
  "walkdir",
@@ -5944,6 +5948,7 @@ dependencies = [
  "derive_builder",
  "esaxx-rs",
  "getrandom 0.3.4",
+ "indicatif",
  "itertools",
  "log",
  "macro_rules_attribute",

src-tauri/Cargo.toml

@@ -39,3 +39,4 @@ candle-core = { version = "0.10.2", features = ["cuda"] }
 candle-nn = { version = "0.10.2", features = ["cuda"] }
 candle-transformers = { version = "0.10.2", features = ["cuda"] }
 hf-hub = { version = "0.5.0", default-features = false, features = ["ureq", "native-tls"] }
+tokenizers = "0.22.1"

src-tauri/src/commands.rs

@@ -1,4 +1,5 @@
 use crate::db::{self, DbPool, Folder, FolderJobProgress, ImageRecord};
+use crate::embedder::ClipImageEmbedder;
 use crate::indexer;
 use crate::vector;
 use serde::{Deserialize, Serialize};
@@ -44,6 +45,15 @@ pub struct RetryFailedEmbeddingsParams {
     pub folder_id: i64,
 }
 
+#[derive(Deserialize)]
+pub struct SemanticSearchParams {
+    pub query: String,
+    pub folder_id: Option<i64>,
+    pub media_kind: Option<String>,
+    pub favorites_only: Option<bool>,
+    pub limit: Option<usize>,
+}
+
 #[tauri::command]
 pub async fn add_folder(
     app: AppHandle,
@@ -188,3 +198,52 @@ pub async fn retry_failed_embeddings(
     let conn = db.get().map_err(|e| e.to_string())?;
     db::retry_failed_embedding_jobs(&conn, params.folder_id).map_err(|e| e.to_string())
 }
+
+#[tauri::command]
+pub async fn semantic_search_images(
+    db: State<'_, DbState>,
+    params: SemanticSearchParams,
+) -> Result<Vec<ImageRecord>, String> {
+    let embedder = ClipImageEmbedder::new().map_err(|e| e.to_string())?;
+    let embedding = embedder.embed_text(&params.query).map_err(|e| e.to_string())?;
+
+    let conn = db.get().map_err(|e| e.to_string())?;
+    let limit = params.limit.unwrap_or(64);
+    let ids = vector::search_image_ids_by_embedding(&conn, &embedding, limit).map_err(|e| e.to_string())?;
+    let mut images = db::get_images_by_ids(&conn, &ids).map_err(|e| e.to_string())?;
+
+    if let Some(folder_id) = params.folder_id {
+        images.retain(|image| image.folder_id == folder_id);
+    }
+    if let Some(media_kind) = params.media_kind.as_deref() {
+        images.retain(|image| image.media_kind == media_kind);
+    }
+    if params.favorites_only.unwrap_or(false) {
+        images.retain(|image| image.favorite);
+    }
+
+    Ok(images)
+}
+
+#[derive(Serialize)]
+pub struct WorkerStates {
+    pub thumbnail_paused: bool,
+    pub metadata_paused: bool,
+    pub embedding_paused: bool,
+}
+
+#[tauri::command]
+pub async fn set_worker_paused(worker: String, paused: bool) -> Result<(), String> {
+    indexer::set_worker_paused(&worker, paused);
+    Ok(())
+}
+
+#[tauri::command]
+pub async fn get_worker_states() -> Result<WorkerStates, String> {
+    let states = indexer::get_worker_paused_states();
+    Ok(WorkerStates {
+        thumbnail_paused: states[0],
+        metadata_paused: states[1],
+        embedding_paused: states[2],
+    })
+}

src-tauri/src/db.rs

@@ -813,6 +813,8 @@ pub fn get_images(
         "date_desc" => "modified_at DESC NULLS LAST",
         "size_asc" => "file_size ASC",
         "size_desc" => "file_size DESC",
+        "duration_asc" => "duration_ms ASC NULLS LAST",
+        "duration_desc" => "duration_ms DESC NULLS LAST",
         _ => "modified_at DESC NULLS LAST",
     };
 

src-tauri/src/embedder.rs

@@ -4,9 +4,11 @@ use candle_nn::VarBuilder;
 use candle_transformers::models::clip::{self, ClipModel};
 use hf_hub::{api::sync::Api, Repo, RepoType};
 use std::path::{Path, PathBuf};
+use tokenizers::Tokenizer;
 
 pub struct ClipImageEmbedder {
     model: ClipModel,
+    tokenizer: Tokenizer,
     device: Device,
     image_size: usize,
 }
@@ -21,6 +23,11 @@ impl ClipImageEmbedder {
         ));
         println!("Resolving CLIP model weights from Hugging Face cache...");
         let model_path = repo.get("model.safetensors")?;
+        let tokenizer_repo = api.repo(Repo::new(
+            "openai/clip-vit-base-patch32".to_string(),
+            RepoType::Model,
+        ));
+        let tokenizer_path = tokenizer_repo.get("tokenizer.json")?;
 
         let config = clip::ClipConfig::vit_base_patch32();
         let device = resolve_device()?;
@@ -32,10 +39,12 @@ impl ClipImageEmbedder {
             )?
         };
         let model = ClipModel::new(vb, &config)?;
+        let tokenizer = Tokenizer::from_file(tokenizer_path).map_err(anyhow::Error::msg)?;
         println!("CLIP image embedder ready.");
 
         Ok(Self {
             model,
+            tokenizer,
             device,
             image_size: config.image_size,
         })
@@ -56,6 +65,22 @@ impl ClipImageEmbedder {
         }
         Ok(embeddings)
     }
+
+    pub fn embed_text(&self, query: &str) -> Result<Vec<f32>> {
+        let encoding = self
+            .tokenizer
+            .encode(query, true)
+            .map_err(anyhow::Error::msg)?;
+        let token_ids = encoding
+            .get_ids()
+            .iter()
+            .map(|token| *token as u32)
+            .collect::<Vec<_>>();
+        let input_ids = Tensor::new(vec![token_ids], &self.device)?;
+        let features = self.model.get_text_features(&input_ids)?;
+        let normalized = clip::div_l2_norm(&features)?;
+        Ok(normalized.flatten_all()?.to_vec1::<f32>()?)
+    }
 }
 
 fn resolve_device() -> Result<Device> {

src-tauri/src/indexer.rs

@@ -9,6 +9,7 @@ use rayon::prelude::*;
 use serde::Serialize;
 use std::collections::{HashMap, HashSet};
 use std::path::{Path, PathBuf};
+use std::sync::atomic::{AtomicBool, Ordering};
 use std::sync::{Mutex, OnceLock};
 use std::time::{Duration, Instant};
 use tauri::{AppHandle, Emitter};
@@ -24,6 +25,27 @@ const JOB_PROGRESS_EMIT_INTERVAL: Duration = Duration::from_millis(750);
 
 static LAST_JOB_PROGRESS_EMIT: OnceLock<Mutex<HashMap<i64, Instant>>> = OnceLock::new();
 static ACTIVE_INDEXING_FOLDERS: OnceLock<Mutex<HashSet<i64>>> = OnceLock::new();
+
+static THUMBNAIL_WORKER_PAUSED: AtomicBool = AtomicBool::new(false);
+static METADATA_WORKER_PAUSED: AtomicBool = AtomicBool::new(false);
+static EMBEDDING_WORKER_PAUSED: AtomicBool = AtomicBool::new(false);
+
+pub fn set_worker_paused(worker: &str, paused: bool) {
+    match worker {
+        "thumbnail" => THUMBNAIL_WORKER_PAUSED.store(paused, Ordering::Relaxed),
+        "metadata" => METADATA_WORKER_PAUSED.store(paused, Ordering::Relaxed),
+        "embedding" => EMBEDDING_WORKER_PAUSED.store(paused, Ordering::Relaxed),
+        _ => {}
+    }
+}
+
+pub fn get_worker_paused_states() -> [bool; 3] {
+    [
+        THUMBNAIL_WORKER_PAUSED.load(Ordering::Relaxed),
+        METADATA_WORKER_PAUSED.load(Ordering::Relaxed),
+        EMBEDDING_WORKER_PAUSED.load(Ordering::Relaxed),
+    ]
+}
 static FOLDER_STORAGE_PROFILES: OnceLock<Mutex<HashMap<i64, RuntimeAdaptiveProfile>>> =
     OnceLock::new();
 static DB_WRITE_LOCK: OnceLock<Mutex<()>> = OnceLock::new();
@@ -73,20 +95,26 @@ pub fn start_thumbnail_worker(
     cache_dir: PathBuf,
 ) {
     std::thread::spawn(move || loop {
+        if THUMBNAIL_WORKER_PAUSED.load(Ordering::Relaxed) {
+            std::thread::sleep(std::time::Duration::from_millis(500));
+            continue;
+        }
         if let Err(error) = process_thumbnail_batch(&app, &pool, &media_tools, &cache_dir) {
             eprintln!("Thumbnail worker error: {}", error);
         }
-
         std::thread::sleep(std::time::Duration::from_millis(250));
     });
 }
 
 pub fn start_metadata_worker(app: AppHandle, pool: DbPool, media_tools: MediaTools) {
     std::thread::spawn(move || loop {
+        if METADATA_WORKER_PAUSED.load(Ordering::Relaxed) {
+            std::thread::sleep(std::time::Duration::from_millis(500));
+            continue;
+        }
         if let Err(error) = process_metadata_batch(&app, &pool, &media_tools) {
             eprintln!("Metadata worker error: {}", error);
         }
-
         std::thread::sleep(std::time::Duration::from_millis(250));
     });
 }
@@ -96,10 +124,13 @@ pub fn start_embedding_worker(app: AppHandle, pool: DbPool) {
         let mut embedder: Option<ClipImageEmbedder> = None;
         println!("Embedding worker started.");
         loop {
+            if EMBEDDING_WORKER_PAUSED.load(Ordering::Relaxed) {
+                std::thread::sleep(std::time::Duration::from_millis(500));
+                continue;
+            }
             if let Err(error) = process_embedding_batch(&app, &pool, &mut embedder) {
                 eprintln!("Embedding worker error: {}", error);
             }
-
             std::thread::sleep(std::time::Duration::from_millis(500));
         }
     });
@@ -319,7 +350,7 @@ fn process_thumbnail_batch(
         return Ok(());
     }
 
-    println!("Embedding batch claimed: {} items", jobs.len());
+    println!("Thumbnail batch claimed: {} items", jobs.len());
 
     let (image_jobs, video_jobs): (Vec<_>, Vec<_>) =
         jobs.into_iter().partition(|job| job.media_kind == "image");

src-tauri/src/lib.rs

@@ -74,6 +74,9 @@ pub fn run() {
             commands::update_image_details,
             commands::find_similar_images,
             commands::retry_failed_embeddings,
+            commands::semantic_search_images,
+            commands::set_worker_paused,
+            commands::get_worker_states,
         ])
         .run(tauri::generate_context!())
         .expect("error while running tauri application");

src-tauri/src/media.rs

@@ -6,6 +6,13 @@ use serde::Deserialize;
 use std::path::{Path, PathBuf};
 use std::process::Command;
 
+// On Windows, GUI apps spawn subprocesses with a visible console window by default.
+// CREATE_NO_WINDOW suppresses that for every ffmpeg/ffprobe invocation.
+#[cfg(target_os = "windows")]
+use std::os::windows::process::CommandExt;
+#[cfg(target_os = "windows")]
+const CREATE_NO_WINDOW: u32 = 0x08000000;
+
 #[derive(Debug, Clone)]
 pub struct MediaTools {
     ffmpeg_path: PathBuf,
@@ -30,6 +37,10 @@ impl MediaTools {
     }
 
     pub fn ensure_installed() -> Result<()> {
+        // Skip download entirely if both binaries are already present.
+        if ffmpeg_path().exists() && ffprobe_path().exists() {
+            return Ok(());
+        }
         auto_download_with_progress(|event| match event {
             FfmpegDownloadProgressEvent::Starting => {
                 println!("Downloading bundled FFmpeg...");
@@ -53,11 +64,17 @@ impl MediaTools {
     }
 
     pub fn ffmpeg_command(&self) -> Command {
-        Command::new(&self.ffmpeg_path)
+        let mut cmd = Command::new(&self.ffmpeg_path);
+        #[cfg(target_os = "windows")]
+        cmd.creation_flags(CREATE_NO_WINDOW);
+        cmd
     }
 
     pub fn ffprobe_command(&self) -> Command {
-        Command::new(&self.ffprobe_path)
+        let mut cmd = Command::new(&self.ffprobe_path);
+        #[cfg(target_os = "windows")]
+        cmd.creation_flags(CREATE_NO_WINDOW);
+        cmd
     }
 }
 

src-tauri/src/vector.rs

@@ -78,6 +78,38 @@ pub fn find_similar_image_ids(conn: &Connection, image_id: i64, limit: usize) ->
     Ok(ids)
 }
 
+pub fn search_image_ids_by_embedding(
+    conn: &Connection,
+    embedding: &[f32],
+    limit: usize,
+) -> Result<Vec<i64>> {
+    if embedding.len() != CLIP_VECTOR_DIM {
+        return Err(anyhow!(
+            "expected {}-dimensional embedding, got {}",
+            CLIP_VECTOR_DIM,
+            embedding.len()
+        ));
+    }
+
+    let packed = pack_f32(embedding);
+    let mut stmt = conn.prepare(
+        "SELECT image_id
+         FROM image_vec
+         WHERE embedding MATCH vec_f32(?1)
+           AND k = ?2",
+    )?;
+    let rows = stmt.query_map((&packed, limit as i64), |row| row.get::<_, i64>(0))?;
+
+    let mut ids = Vec::new();
+    for row in rows {
+        ids.push(row?);
+        if ids.len() >= limit {
+            break;
+        }
+    }
+    Ok(ids)
+}
+
 #[allow(dead_code)]
 fn pack_f32(values: &[f32]) -> Vec<u8> {
     let mut out = Vec::with_capacity(values.len() * std::mem::size_of::<f32>());