feat(migration): implement shared cache migration with SHA1 dedup

- Add migrate_to_shared_caches() with hard link preference
- SHA1-based deduplication across all instances
- Copy fallback for cross-filesystem scenarios
- Auto-enable use_shared_caches after successful migration
- UI shows statistics: moved files, hardlinks/copies, MB saved

2 files changed, 275 insertions, 1 deletions

diff --git a/src-tauri/src/core/instance.rs b/src-tauri/src/core/instance.rs
index 738dbd8..183e1cc 100644
--- a/src-tauri/src/core/instance.rs
+++ b/src-tauri/src/core/instance.rs
@@ -6,6 +6,7 @@
 //! - Support for instance switching and isolation
 
 use serde::{Deserialize, Serialize};
+use std::collections::HashMap;
 use std::fs;
 use std::path::{Path, PathBuf};
 use std::sync::Mutex;
@@ -344,3 +345,226 @@ pub fn migrate_legacy_data(
 
     Ok(())
 }
+
+/// Migrate instance caches to shared global caches
+///
+/// This function deduplicates versions, libraries, and assets from all instances
+/// into a global shared cache. It prefers hard links (instant, zero-copy) and
+/// falls back to copying if hard links are not supported.
+///
+/// # Arguments
+/// * `app_handle` - Tauri app handle
+/// * `instance_state` - Instance state management
+///
+/// # Returns
+/// * `Ok((moved_count, hardlink_count, copy_count, saved_bytes))` on success
+/// * `Err(String)` on failure
+pub fn migrate_to_shared_caches(
+    app_handle: &AppHandle,
+    instance_state: &InstanceState,
+) -> Result<(usize, usize, usize, u64), String> {
+    let app_dir = app_handle.path().app_data_dir().unwrap();
+
+    // Global shared cache directories
+    let global_versions = app_dir.join("versions");
+    let global_libraries = app_dir.join("libraries");
+    let global_assets = app_dir.join("assets");
+
+    // Create global cache directories
+    std::fs::create_dir_all(&global_versions).map_err(|e| e.to_string())?;
+    std::fs::create_dir_all(&global_libraries).map_err(|e| e.to_string())?;
+    std::fs::create_dir_all(&global_assets).map_err(|e| e.to_string())?;
+
+    let mut total_moved = 0;
+    let mut hardlink_count = 0;
+    let mut copy_count = 0;
+    let mut saved_bytes = 0u64;
+
+    // Get all instances
+    let instances = instance_state.list_instances();
+
+    for instance in instances {
+        let instance_versions = instance.game_dir.join("versions");
+        let instance_libraries = instance.game_dir.join("libraries");
+        let instance_assets = instance.game_dir.join("assets");
+
+        // Migrate versions
+        if instance_versions.exists() {
+            let (moved, hardlinks, copies, bytes) =
+                deduplicate_directory(&instance_versions, &global_versions)?;
+            total_moved += moved;
+            hardlink_count += hardlinks;
+            copy_count += copies;
+            saved_bytes += bytes;
+        }
+
+        // Migrate libraries
+        if instance_libraries.exists() {
+            let (moved, hardlinks, copies, bytes) =
+                deduplicate_directory(&instance_libraries, &global_libraries)?;
+            total_moved += moved;
+            hardlink_count += hardlinks;
+            copy_count += copies;
+            saved_bytes += bytes;
+        }
+
+        // Migrate assets
+        if instance_assets.exists() {
+            let (moved, hardlinks, copies, bytes) =
+                deduplicate_directory(&instance_assets, &global_assets)?;
+            total_moved += moved;
+            hardlink_count += hardlinks;
+            copy_count += copies;
+            saved_bytes += bytes;
+        }
+    }
+
+    Ok((total_moved, hardlink_count, copy_count, saved_bytes))
+}
+
+/// Deduplicate a directory tree into a global cache
+///
+/// Recursively processes all files, checking SHA1 hashes for deduplication.
+/// Returns (total_moved, hardlink_count, copy_count, saved_bytes)
+fn deduplicate_directory(
+    source_dir: &Path,
+    dest_dir: &Path,
+) -> Result<(usize, usize, usize, u64), String> {
+    let mut moved = 0;
+    let mut hardlinks = 0;
+    let mut copies = 0;
+    let mut saved_bytes = 0u64;
+
+    // Build a hash map of existing files in dest (hash -> path)
+    let mut dest_hashes: HashMap<String, PathBuf> = HashMap::new();
+    if dest_dir.exists() {
+        index_directory_hashes(dest_dir, dest_dir, &mut dest_hashes)?;
+    }
+
+    // Process source directory
+    process_directory_for_migration(
+        source_dir,
+        source_dir,
+        dest_dir,
+        &dest_hashes,
+        &mut moved,
+        &mut hardlinks,
+        &mut copies,
+        &mut saved_bytes,
+    )?;
+
+    Ok((moved, hardlinks, copies, saved_bytes))
+}
+
+/// Index all files in a directory by their SHA1 hash
+fn index_directory_hashes(
+    dir: &Path,
+    base: &Path,
+    hashes: &mut HashMap<String, PathBuf>,
+) -> Result<(), String> {
+    if !dir.is_dir() {
+        return Ok(());
+    }
+
+    for entry in std::fs::read_dir(dir).map_err(|e| e.to_string())? {
+        let entry = entry.map_err(|e| e.to_string())?;
+        let path = entry.path();
+
+        if path.is_dir() {
+            index_directory_hashes(&path, base, hashes)?;
+        } else if path.is_file() {
+            let hash = compute_file_sha1(&path)?;
+            hashes.insert(hash, path);
+        }
+    }
+
+    Ok(())
+}
+
+/// Process directory for migration (recursive)
+fn process_directory_for_migration(
+    current: &Path,
+    source_base: &Path,
+    dest_base: &Path,
+    dest_hashes: &HashMap<String, PathBuf>,
+    moved: &mut usize,
+    hardlinks: &mut usize,
+    copies: &mut usize,
+    saved_bytes: &mut u64,
+) -> Result<(), String> {
+    if !current.is_dir() {
+        return Ok(());
+    }
+
+    for entry in std::fs::read_dir(current).map_err(|e| e.to_string())? {
+        let entry = entry.map_err(|e| e.to_string())?;
+        let source_path = entry.path();
+
+        // Compute relative path
+        let rel_path = source_path
+            .strip_prefix(source_base)
+            .map_err(|e| e.to_string())?;
+        let dest_path = dest_base.join(rel_path);
+
+        if source_path.is_dir() {
+            // Recurse into subdirectory
+            process_directory_for_migration(
+                &source_path,
+                source_base,
+                dest_base,
+                dest_hashes,
+                moved,
+                hardlinks,
+                copies,
+                saved_bytes,
+            )?;
+        } else if source_path.is_file() {
+            let file_size = std::fs::metadata(&source_path)
+                .map(|m| m.len())
+                .unwrap_or(0);
+
+            // Compute file hash
+            let source_hash = compute_file_sha1(&source_path)?;
+
+            // Check if file already exists in dest with same hash
+            if let Some(_existing) = dest_hashes.get(&source_hash) {
+                // File exists, delete source (already deduplicated)
+                std::fs::remove_file(&source_path).map_err(|e| e.to_string())?;
+                *saved_bytes += file_size;
+                *moved += 1;
+            } else {
+                // File doesn't exist, move it
+                // Create parent directory in dest
+                if let Some(parent) = dest_path.parent() {
+                    std::fs::create_dir_all(parent).map_err(|e| e.to_string())?;
+                }
+
+                // Try hard link first
+                if std::fs::hard_link(&source_path, &dest_path).is_ok() {
+                    // Hard link succeeded, remove source
+                    std::fs::remove_file(&source_path).map_err(|e| e.to_string())?;
+                    *hardlinks += 1;
+                    *moved += 1;
+                } else {
+                    // Hard link failed (different filesystem?), copy instead
+                    std::fs::copy(&source_path, &dest_path).map_err(|e| e.to_string())?;
+                    std::fs::remove_file(&source_path).map_err(|e| e.to_string())?;
+                    *copies += 1;
+                    *moved += 1;
+                }
+            }
+        }
+    }
+
+    Ok(())
+}
+
+/// Compute SHA1 hash of a file
+fn compute_file_sha1(path: &Path) -> Result<String, String> {
+    use sha1::{Digest, Sha1};
+
+    let data = std::fs::read(path).map_err(|e| e.to_string())?;
+    let mut hasher = Sha1::new();
+    hasher.update(&data);
+    Ok(hex::encode(hasher.finalize()))
+}
diff --git a/src-tauri/src/main.rs b/src-tauri/src/main.rs
index 6a230c9..a506713 100644
--- a/src-tauri/src/main.rs
+++ b/src-tauri/src/main.rs
@@ -2373,6 +2373,55 @@ async fn assistant_chat_stream(
         .await
 }
 
+/// Migrate instance caches to shared global caches
+#[derive(Serialize)]
+struct MigrationResult {
+    moved_files: usize,
+    hardlinks: usize,
+    copies: usize,
+    saved_bytes: u64,
+    saved_mb: f64,
+}
+
+#[tauri::command]
+async fn migrate_shared_caches(
+    window: Window,
+    instance_state: State<'_, core::instance::InstanceState>,
+    config_state: State<'_, core::config::ConfigState>,
+) -> Result<MigrationResult, String> {
+    emit_log!(window, "Starting migration to shared caches...".to_string());
+
+    let app_handle = window.app_handle();
+    let (moved, hardlinks, copies, saved_bytes) =
+        core::instance::migrate_to_shared_caches(app_handle, &instance_state)?;
+
+    let saved_mb = saved_bytes as f64 / (1024.0 * 1024.0);
+
+    emit_log!(
+        window,
+        format!(
+            "Migration complete: {} files moved ({} hardlinks, {} copies), {:.2} MB saved",
+            moved, hardlinks, copies, saved_mb
+        )
+    );
+
+    // Automatically enable shared caches config
+    let mut config = config_state.config.lock().unwrap().clone();
+    config.use_shared_caches = true;
+    drop(config);
+    *config_state.config.lock().unwrap() = config_state.config.lock().unwrap().clone();
+    config_state.config.lock().unwrap().use_shared_caches = true;
+    config_state.save()?;
+
+    Ok(MigrationResult {
+        moved_files: moved,
+        hardlinks,
+        copies,
+        saved_bytes,
+        saved_mb,
+    })
+}
+
 fn main() {
     tauri::Builder::default()
         .plugin(tauri_plugin_fs::init())
@@ -2479,7 +2528,8 @@ fn main() {
             get_instance,
             set_active_instance,
             get_active_instance,
-            duplicate_instance
+            duplicate_instance,
+            migrate_shared_caches
         ])
         .run(tauri::generate_context!())
         .expect("error while running tauri application");