Update app.py

app.py

@@ -232,58 +232,101 @@ def save_text_to_pdf(text, output_path):
     doc.save(output_path)  # Save the PDF to the specified path
     print("PDF saved successfully.")
 
-
+def get_predefined_queries(company):
+    return [
+        f"Recent earnings for {company}",
+        f"Recent News on {company}",
+        f"Recent Credit rating of {company}",
+        f"Recent conference call transcript of {company}"
+    ]
 
 
 # Integrated function to perform web scraping, formatting, and text generation
-def scrape_and_display(query, num_results, instructions, web_search=True, temperature=0.7, repetition_penalty=1.0, top_p=0.9):
+def scrape_and_display(query, num_results, earnings_instructions, news_instructions, 
+                       credit_rating_instructions, conference_call_instructions, final_instructions, 
+                       web_search=True, temperature=0.7, repetition_penalty=1.0, top_p=0.9):
     print(f"Scraping and displaying results for query: {query} with num_results: {num_results}")
+    
     if web_search:
-        search_results = google_search(query, num_results)
-        formatted_prompt = format_prompt(query, search_results, instructions)
-        generated_summary = generate_text(formatted_prompt, temperature=temperature, repetition_penalty=repetition_penalty, top_p=top_p)
+        company = query.strip()
+        predefined_queries = get_predefined_queries(company)
+        all_results = []
+        all_summaries = []
+        
+        instructions = [earnings_instructions, news_instructions, credit_rating_instructions, conference_call_instructions]
+        
+        for pq, instruction in zip(predefined_queries, instructions):
+            search_results = google_search(pq, num_results=num_results // len(predefined_queries))
+            all_results.extend(search_results)
+            
+            # Generate a summary for each predefined query
+            formatted_prompt = format_prompt(pq, search_results, instruction)
+            summary = generate_text(formatted_prompt, temperature=temperature, repetition_penalty=repetition_penalty, top_p=top_p)
+            all_summaries.append(summary)
+        
+        # Combine all summaries
+        combined_summary = "\n\n".join(all_summaries)
+        
+        # Generate final summary using the combined results and final instructions
+        final_prompt = f"{final_instructions}\n\nHere are the summaries for each aspect of {company}:\n\n{combined_summary}\n\nPlease provide a comprehensive summary based on the above information:"
+        generated_summary = generate_text(final_prompt, temperature=temperature, repetition_penalty=repetition_penalty, top_p=top_p)
     else:
-        formatted_prompt = format_prompt_with_instructions(query, instructions)
+        formatted_prompt = format_prompt_with_instructions(query, final_instructions)
         generated_summary = generate_text(formatted_prompt, temperature=temperature, repetition_penalty=repetition_penalty, top_p=top_p)
+    
     print("Scraping and display complete.")
     if generated_summary:
-        # Extract and return text starting from "Assistant:"
         assistant_index = generated_summary.find("Assistant:")
         if assistant_index != -1:
             generated_summary = generated_summary[assistant_index:]
         else:
             generated_summary = "Assistant: No response generated."
-    print(f"Generated summary: {generated_summary}")  # Debugging line
+    print(f"Generated summary: {generated_summary}")
     return generated_summary
 
+
 # Main Gradio interface function
-def gradio_interface(query, use_pdf, pdf, num_results, instructions, temperature, repetition_penalty, top_p):
+def gradio_interface(query, use_pdf, pdf, num_results, earnings_instructions, news_instructions, 
+                     credit_rating_instructions, conference_call_instructions, final_instructions, 
+                     temperature, repetition_penalty, top_p):
     if use_pdf and pdf is not None:
         pdf_text = read_pdf(pdf)
-        generated_summary = scrape_and_display(pdf_text, num_results=0, instructions=instructions, web_search=False, temperature=temperature, repetition_penalty=repetition_penalty, top_p=top_p)
+        generated_summary = scrape_and_display(pdf_text, num_results=0, instructions=final_instructions, 
+                                               web_search=False, temperature=temperature, 
+                                               repetition_penalty=repetition_penalty, top_p=top_p)
     else:
-        generated_summary = scrape_and_display(query, num_results=num_results, instructions=instructions, web_search=True, temperature=temperature, repetition_penalty=repetition_penalty, top_p=top_p)
+        generated_summary = scrape_and_display(query, num_results=num_results, 
+                                               earnings_instructions=earnings_instructions,
+                                               news_instructions=news_instructions,
+                                               credit_rating_instructions=credit_rating_instructions,
+                                               conference_call_instructions=conference_call_instructions,
+                                               final_instructions=final_instructions,
+                                               web_search=True, temperature=temperature, 
+                                               repetition_penalty=repetition_penalty, top_p=top_p)
     
-    # Save the generated summary to a PDF
     output_pdf_path = "output_summary.pdf"
     save_text_to_pdf(generated_summary, output_pdf_path)
     
     return generated_summary, output_pdf_path
 
-# Deploy Gradio Interface
+# Update the Gradio Interface
 gr.Interface(
     fn=gradio_interface,
     inputs=[
-        gr.Textbox(label="Query"),
+        gr.Textbox(label="Company Name"),
         gr.Checkbox(label="Use PDF"),
         gr.File(label="Upload PDF"),
-        gr.Slider(minimum=1, maximum=20, label="Number of Results"),  # Added Slider for num_results
-        gr.Textbox(label="Instructions"),
-        gr.Slider(minimum=0.1, maximum=1.0, label="Temperature"),
-        gr.Slider(minimum=0.1, maximum=1.0, label="Repetition Penalty"),
-        gr.Slider(minimum=0.1, maximum=1.0, label="Top p")
+        gr.Slider(minimum=4, maximum=40, step=4, value=20, label="Number of Results (total for all queries)"),
+        gr.Textbox(label="Earnings Instructions", lines=2, placeholder="Instructions for recent earnings query..."),
+        gr.Textbox(label="News Instructions", lines=2, placeholder="Instructions for recent news query..."),
+        gr.Textbox(label="Credit Rating Instructions", lines=2, placeholder="Instructions for credit rating query..."),
+        gr.Textbox(label="Conference Call Instructions", lines=2, placeholder="Instructions for conference call transcript query..."),
+        gr.Textbox(label="Final Summary Instructions", lines=2, placeholder="Instructions for the final summary..."),
+        gr.Slider(minimum=0.1, maximum=1.0, value=0.7, label="Temperature"),
+        gr.Slider(minimum=1.0, maximum=2.0, value=1.0, label="Repetition Penalty"),
+        gr.Slider(minimum=0.1, maximum=1.0, value=0.9, label="Top p")
     ],
-    outputs=["text", "file"],  # Updated to return text and a file
+    outputs=["text", "file"],
     title="Financial Analyst AI Assistant",
-    description="Enter your query about a company's financials to get valuable insights. Optionally, upload a PDF for analysis.Please instruct me for curating your output template, also for web search you can modify my search results but its advisable to restrict the same at 10. You can also adjust my parameters like Temperature, Repetition Penalty and Top_P, its adivsable to set repetition penalty at 1 and other two parameters at 0.1.",
-).launch(share=True)
+    description="Enter a company name and provide specific instructions for each query. The AI will use these instructions to gather and summarize information on recent earnings, news, credit ratings, and conference call transcripts.",
+)