Update app.py

app.py

@@ -179,101 +179,74 @@ def generate_output(context, query):
 #         context_data, combined_context = search_documents(query)
         
 #         document_titles = list({os.path.basename(doc["title"]) for doc in context_data})  # Get only file names
-
 #         formatted_titles = " " + "\n".join(document_titles)
 
+#         total_results = len(context_data)  # Count the total number of results
+
 #         results = {
 #             "results": [
 #                 {
 #                     "natural_language_output": generate_output(doc["relevant_text"], query),
-#                     "doc_id": doc["doc_id"],
 #                     "chunk_id": doc["chunk_id"],
+#                     "document_id": doc["doc_id"],  # Assuming doc_id is the UUID
 #                     "title": doc["title"],
 #                     "relevant_text": doc["relevant_text"],
 #                     "page_number": doc["page_number"],
 #                     "score": doc["score"],
 #                 }
 #                 for doc in context_data
-#             ]
+#             ],
+#             "total_results": total_results  # Added total_results field
 #         }
 
 #         return results, formatted_titles  # Return results and formatted document titles
 #     except Exception as e:
-#         return {"results": []}, f"Error in workflow: {str(e)}"
+#         return {"results": [], "total_results": 0}, f"Error in workflow: {str(e)}"
+
 
 def complete_workflow(query):
     try:
-        context_data, combined_context = search_documents(query)
+        # 🔹 Step 1: Perform Hybrid Search (Vector + BM25)
+        context_data, combined_context = hybrid_search_documents(query)
+
+        # 🔹 Step 2: Generate LLM-based Natural Language Output
+        llm = ChatOpenAI(model="gpt-4", openai_api_key=openai.api_key, temperature=0.7)
+        prompt_template = """
+        Use the following context to answer the question as accurately as possible:
         
-        document_titles = list({os.path.basename(doc["title"]) for doc in context_data})  # Get only file names
-        formatted_titles = " " + "\n".join(document_titles)
+        Context: {context}
+        Question: {question}
+        
+        Answer:
+        """
+        prompt = prompt_template.format(context=combined_context, question=query)
+        response = llm([HumanMessage(content=prompt)])
 
-        total_results = len(context_data)  # Count the total number of results
+        # 🔹 Step 3: Format Results
+        document_titles = list({os.path.basename(doc["title"]) for doc in context_data})  # Extract unique file names
+        formatted_titles = "\n".join(document_titles)
 
         results = {
             "results": [
                 {
-                    "natural_language_output": generate_output(doc["relevant_text"], query),
+                    "natural_language_output": response.content,
                     "chunk_id": doc["chunk_id"],
-                    "document_id": doc["doc_id"],  # Assuming doc_id is the UUID
+                    "document_id": doc["doc_id"],
                     "title": doc["title"],
                     "relevant_text": doc["relevant_text"],
                     "page_number": doc["page_number"],
                     "score": doc["score"],
+                    "method": doc["method"],  # "vector" or "bm25"
                 }
                 for doc in context_data
             ],
-            "total_results": total_results  # Added total_results field
+            "total_results": len(context_data),  # Return total number of retrieved results
         }
 
-        return results, formatted_titles  # Return results and formatted document titles
+        return results, formatted_titles  # Return both results and formatted document titles
     except Exception as e:
         return {"results": [], "total_results": 0}, f"Error in workflow: {str(e)}"
 
-
-# def complete_workflow(query):
-#     try:
-#         # 🔹 Step 1: Perform Hybrid Search (Vector + BM25)
-#         context_data, combined_context = hybrid_search_documents(query)
-
-#         # 🔹 Step 2: Generate LLM-based Natural Language Output
-#         llm = ChatOpenAI(model="gpt-4", openai_api_key=openai.api_key, temperature=0.7)
-#         prompt_template = """
-#         Use the following context to answer the question as accurately as possible:
-        
-#         Context: {context}
-#         Question: {question}
-        
-#         Answer:
-#         """
-#         prompt = prompt_template.format(context=combined_context, question=query)
-#         response = llm([HumanMessage(content=prompt)])
-
-#         # 🔹 Step 3: Format Results
-#         document_titles = list({os.path.basename(doc["title"]) for doc in context_data})  # Extract unique file names
-#         formatted_titles = "\n".join(document_titles)
-
-#         results = {
-#             "results": [
-#                 {
-#                     "natural_language_output": response.content,
-#                     "chunk_id": doc["chunk_id"],
-#                     "document_id": doc["doc_id"],
-#                     "title": doc["title"],
-#                     "relevant_text": doc["relevant_text"],
-#                     "page_number": doc["page_number"],
-#                     "score": doc["score"],
-#                     "method": doc["method"],  # "vector" or "bm25"
-#                 }
-#                 for doc in context_data
-#             ],
-#             "total_results": len(context_data),  # Return total number of retrieved results
-#         }
-
-#         return results, formatted_titles  # Return both results and formatted document titles
-#     except Exception as e:
-#         return {"results": [], "total_results": 0}, f"Error in workflow: {str(e)}"
-
 def gradio_app():
     with gr.Blocks(css=".result-output {width: 150%; font-size: 16px; padding: 10px;}") as app:
         gr.Markdown("### Intelligent Document Search Prototype-v0.1.2 ")