New changes

.gitignore

@@ -0,0 +1 @@
+.env

app.py

@@ -22,6 +22,11 @@ import traceback
 import shutil
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from tqdm import tqdm
+from dotenv import load_dotenv
+import os
+
+load_dotenv()
+GROQ_API_KEY = os.getenv("GROQ_API_KEY")
 
 embedder = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
 
@@ -29,13 +34,11 @@ def build_index_and_dataset(domain, subsets, embedder_type="sentence-transformer
     dataset_path = f"{domain}_dataset"
     index_path = f"{domain}_index/faiss.index"
 
-    # ❌ Always remove previous
-    if os.path.exists(dataset_path):
-        shutil.rmtree(dataset_path)
-    if os.path.exists(index_path):
-        os.remove(index_path)
+    if os.path.exists(dataset_path) and os.path.exists(index_path):
+        print(f"✅ Using cached dataset and index for domain: {domain}")
+        return Dataset.load_from_disk(dataset_path), faiss.read_index(index_path)
 
-    print(f"🚀 Rebuilding dataset and index for domain: {domain}")
+    print(f"🚀 Building dataset and index for domain: {domain}")
 
     all_docs = []
     for subset in subsets:
@@ -107,6 +110,8 @@ gk_dataset = load_dataset("rungalileo/ragbench", "hotpotqa", split="test")
 cs_dataset = load_dataset("rungalileo/ragbench", "emanual", split="test")
 fin_dataset = load_dataset("rungalileo/ragbench", "finqa", split="test")
 
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
 # Load BGE reranker
 reranker = CrossEncoder("BAAI/bge-reranker-base", max_length=512)
 
@@ -173,7 +178,7 @@ def retrieve_top_c(query, domain, embedder,  k=5):
 
 
 client = Groq(
-    api_key= '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',#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',
+    api_key= 'GROQ_API_KEY',
 )
 
 
@@ -584,7 +589,7 @@ def evaluate_rag_pipeline(domain, q_indices):
         result["AUC-ROC (Adherence)"] = round(roc_auc_score(gt_adherence, pred_adherence), 4)
     else:
         result["Adherence"] = compute_rmse(gt_adherence, pred_adherence)
-        result["AUC-ROC (Adherence)"] = "N/A - one class only"
+        #result["AUC-ROC (Adherence)"] = "N/A - one class only"
 
     return result
 
@@ -627,4 +632,4 @@ iface = gr.Interface(
 )
 
 # Launch app
-iface.launch(server_name="0.0.0.0", server_port=7860, debug=True)
+iface.launch(server_name="0.0.0.0", server_port=7860, debug=True)

requirements.txt

@@ -7,4 +7,5 @@ datasets
 scikit-learn
 groq
 langchain
-tqdm
+tqdm
+python-dotenv