Update app.py

app.py

@@ -5,8 +5,6 @@ import fitz  # PyMuPDF for reading PDF files
 import pytesseract
 from PIL import Image
 import docx  # for reading .docx files
-from ragchecker import RAGResults, RAGChecker
-from ragchecker.metrics import all_metrics
 from llama_index.core import VectorStoreIndex, Document
 from llama_index.embeddings.openai import OpenAIEmbedding
 from llama_index.llms.openai import OpenAI
@@ -24,7 +22,7 @@ load_dotenv()
 
 # Initialize global variables
 vector_index = None
-query_log = []  # Store queries and results for RAGChecker
+query_log = []  # Store queries and results for logging purposes
 
 # Function to handle PDF and OCR for scanned PDFs
 def load_pdf_manually(pdf_path):
@@ -102,7 +100,10 @@ def query_app(query, model_name, use_rag_checker, openai_api_key):
         return "Please provide a valid OpenAI API Key.", None
 
     # Initialize the LLM with the selected model and provided API key
-    llm = OpenAI(model=model_name, api_key=openai_api_key)
+    try:
+        llm = OpenAI(model=model_name, api_key=openai_api_key)
+    except Exception as e:
+        return f"Error initializing the OpenAI model: {e}", None
 
     # Create a query engine and query the indexed documents
     response_synthesizer = get_response_synthesizer(llm=llm)
@@ -123,39 +124,8 @@ def query_app(query, model_name, use_rag_checker, openai_api_key):
         "retrieved_context": [{"text": doc.text} for doc in response.source_nodes]
     })
 
-    # Initialize metrics dictionary
-    metrics = {}
-
-    # Calculate BERTScore if RAGChecker is selected
-    if use_rag_checker:
-        try:
-            rag_results = RAGResults.from_dict({"results": query_log})
-            evaluator = RAGChecker(
-                extractor_name="openai/gpt-4o-mini",
-                checker_name="openai/gpt-4o-mini",
-                batch_size_extractor=32,   # Removed any instances where max_workers might be passed
-                batch_size_checker=32
-            )
-            evaluator.evaluate(rag_results, all_metrics)
-            metrics = rag_results.metrics
-
-            # Calculate BERTScore as an additional metric
-            gt_answer = ["Placeholder ground truth answer"]  # Replace with actual ground truth
-            candidate = [generated_response]
-
-            P, R, F1 = bert_score(candidate, gt_answer, lang="en", verbose=False)
-            metrics['bertscore'] = {
-                "precision": P.mean().item() * 100,
-                "recall": R.mean().item() * 100,
-                "f1": F1.mean().item() * 100
-            }
-        except Exception as e:
-            metrics['error'] = f"Error calculating metrics: {e}"
-
-    if use_rag_checker:
-        return generated_response, metrics
-    else:
-        return generated_response, None
+    # Skip RAG Checker process entirely since use_rag_checker is disabled
+    return generated_response, None
 
 # Define the Gradio interface
 def main():
@@ -185,7 +155,7 @@ def main():
                     value="gpt-4o",
                     label="Select Model"
                 )
-                rag_checkbox = gr.Checkbox(label="Use RAG Checker", value=True)
+                rag_checkbox = gr.Checkbox(label="Use RAG Checker", value=False)  # Set RAG Checker default to False
                 query_button = gr.Button("Ask")
             with gr.Column():
                 answer_output = gr.Textbox(label="Answer", interactive=False)