cache utils

app.py

@@ -52,16 +52,8 @@ def _escape_srcdoc(text: str) -> str:
 
 
 def run(repo_url: str, threshold: float, multimodal: bool, sim_method: str):
-    # Check if we can use cached embeddings for embedding similarity
-    embeddings_cache = Path("embeddings_cache.npz")
-    
-    if sim_method == "embedding" and embeddings_cache.exists():
-        print("🚀 Using cached embeddings - skipping repo download")
-        # Use a dummy path since we won't need the actual repo
-        repo_path = Path("/tmp/dummy")
-    else:
-        print("📥 Downloading/updating repository")
-        repo_path = clone_or_cache(repo_url)
+    # Always download repo for now - let build_graph_json decide if it needs it
+    repo_path = clone_or_cache(repo_url)
 
     graph = build_graph_json(
         transformers_dir=repo_path,

modular_graph_and_candidates.py

@@ -317,10 +317,14 @@ def build_graph_json(
     
     # Check if we can use cached embeddings only
     embeddings_cache = Path("embeddings_cache.npz")
+    print(f"🔍 Cache file exists: {embeddings_cache.exists()}, sim_method: {sim_method}")
+    
     if sim_method == "embedding" and embeddings_cache.exists():
         try:
             # Try to compute from cache without accessing repo
             cached_sims = compute_similarities_from_cache(threshold)
+            print(f"🔍 Got {len(cached_sims)} cached similarities")
+            
             if cached_sims:
                 # Create minimal graph with cached data
                 cached_data = np.load(embeddings_cache, allow_pickle=True)