Refining UI

app.py

@@ -116,7 +116,8 @@ def sanitize_outputs(outputs):
 def diagnose(gpt_key, llama_key, gemini_key, top_p, temperature, symptoms):
 
     if symptoms:
-        prompt = "Given the next set of symptoms, classify the diagnosis as one of the following: Psoriasis, Arthritis, Bronchial asthma, Cervical spondylosis. Please only output the classified diagnosis and nothing after that."
+        prompt = "Given the next set of symptoms, classify the diagnosis as one of the following: "
+        prompt += "Psoriasis, Arthritis, Bronchial asthma, Cervical spondylosis. Please only output the classified diagnosis and nothing after that."
         prompt += "Choose only one among the words Psoriasis, Arthritis, Bronchial asthma or Cervical spondylosis"
         prompt += "Do not list the symptoms again in the response. Do not add any additional text. Do not attempt to explain your answer."
         prompt += symptoms
@@ -129,12 +130,34 @@ def diagnose(gpt_key, llama_key, gemini_key, top_p, temperature, symptoms):
         outputs = sanitize_outputs(outputs)
         output_counts = Counter(outputs)
         majority_output, majority_count = output_counts.most_common(1)[0]
-        confidence = (majority_count / len(outputs)) * 100
+        confidence = int((majority_count / len(outputs)) * 100)
+
+        if majority_output > 0:
+            if confidence < 50:
+                message = "LLMs do not seem to agree on one output." + "\n"
+                message += " GPT output: " + gpt_message + "\n"
+                message += " Llama output: " + llama_message + "\n"
+                message += " Gemini output: " + gemini_message + "\n"
+                message += " Confidence score: " + confidence + "%"
+            elif confidence > 50:
+                message = "LLMs seem to agree on one output." + "\n"
+                message += " GPT output: " + gpt_message + "\n"
+                message += " Llama output: " + llama_message + "\n"
+                message += " Gemini output: " + gemini_message + "\n"
+                message += " Confidence score: " + confidence + "%"
+            elif confidence >= 100:
+                message = "All LLMs agree on the same output." + "\n"
+                message += " GPT output: " + gpt_message + "\n"
+                message += " Llama output: " + llama_message + "\n"
+                message += " Gemini output: " + gemini_message + "\n"
+                message += " Confidence score: " + confidence + "%"
+        else:
+            message = "Majority output is not a valid value, please check what is wrong."
+        return message
     else:
-        majority_output = "Please add the symptoms data to start the ranking process."
-        confidence = 0
+        message = "Please add the symptoms data to start the ranking process."
     
-    return outputs, majority_output, confidence
+    return message
 
 def update_model_components(selected_model):
     model_map = {
@@ -188,7 +211,7 @@ with gr.Blocks() as ui:
             llm_btn = gr.Button(value="Diagnose Disease", variant="primary", elem_id="diagnose", interactive=False)            
             symptoms.input(toggle_button, inputs=[symptoms, gpt_key, llama_key, gemini_key], outputs=llm_btn)
             output = gr.Textbox(label="LLM output with majority vote and confidence", interactive=False, placeholder="Output will appear here...")
-            llm_btn.click(fn=diagnose, inputs=[gpt_key, llama_key, gemini_key, top_p, temperature, symptoms], outputs=output, api_name="LLM_Comparator")
-
+            llm_btn.click(fn=diagnose, inputs=[gpt_key, llama_key, gemini_key, top_p, temperature, symptoms], 
+                          outputs=output, api_name="LLM_Comparator")
 
 ui.launch(share=True)