Upload 2 files

src/apps/utils/main.py

@@ -6,10 +6,25 @@ import pickle
 import os
 
 BASE_DIR = os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))))
-pkl_path = os.path.join(BASE_DIR, "volumes", "metadata", "new_pdfs_corpus_data.pkl")
+def get_path(folder, filename):
+    # Try the standard 'volumes/' path (local structure)
+    path1 = os.path.join(BASE_DIR, "volumes", folder, filename)
+    if os.path.exists(path1):
+        return path1
+    # Try the root-level path (Hugging Face structure)
+    path2 = os.path.join(BASE_DIR, folder, filename)
+    if os.path.exists(path2):
+        return path2
+    # Default to path1 if neither exists, so it raises the error on the first expected path
+    return path1
 
-with open(pkl_path, "rb") as p:
-    metadata = pickle.load(p)
+pkl_path = get_path("metadata", "new_pdfs_corpus_data.pkl")
+
+if not os.path.exists(pkl_path):
+    print(f"⚠️ ERROR: Metadata file not found at {pkl_path}")
+else:
+    with open(pkl_path, "rb") as p:
+        metadata = pickle.load(p)
 
 # ids = list(metadata.keys())
 

src/apps/utils/retriever.py

@@ -1,19 +1,35 @@
-import os
-os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"  # Temporary fix
-os.environ["FAISS_NO_OPENMP"] = "1"  # Prevent FAISS from using OpenMP
-
-import faiss
-from numpy.linalg import norm
-
-BASE_DIR = os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))))
-index_path = os.path.join(BASE_DIR, "volumes", "indexes", "law_corpus_index2.bin")
-if not os.path.exists(index_path):
-    index_path = os.path.join(BASE_DIR, "indexes", "law_corpus_index2.bin")
-index = faiss.read_index(index_path)
-print("Index loaded successfully!")
-print("Number of vectors in the index:", index.ntotal)
-
-def vector_db_retriever(query_embeddings, top_k=10):
-    query_embeddings = query_embeddings / norm(query_embeddings[0])
-    distances, indices = index.search(query_embeddings, top_k)
-    return indices, distances
+import os
+os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"  # Temporary fix
+os.environ["FAISS_NO_OPENMP"] = "1"  # Prevent FAISS from using OpenMP
+
+import faiss
+from numpy.linalg import norm
+
+BASE_DIR = os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))))
+def get_path(folder, filename):
+    # Try the standard 'volumes/' path
+    path1 = os.path.join(BASE_DIR, "volumes", folder, filename)
+    if os.path.exists(path1):
+        return path1
+    # Try the root-level path
+    path2 = os.path.join(BASE_DIR, folder, filename)
+    if os.path.exists(path2):
+        return path2
+    return path1
+
+index_path = get_path("indexes", "law_corpus_index2.bin")
+
+if not os.path.exists(index_path):
+    print(f"⚠️ ERROR: Index file not found at {index_path}")
+    # Fail gracefully or provide a placeholder for the index if needed
+    # (Leaving it to crash on faiss.read_index if it truly doesn't exist)
+    pass
+
+index = faiss.read_index(index_path)
+print("Index loaded successfully!")
+print("Number of vectors in the index:", index.ntotal)
+
+def vector_db_retriever(query_embeddings, top_k=10):
+    query_embeddings = query_embeddings / norm(query_embeddings[0])
+    distances, indices = index.search(query_embeddings, top_k)
+    return indices, distances