Update app.py

app.py

@@ -1,73 +1,77 @@
 import gradio as gr
 import os
-from pathlib import Path
-import argparse
 from huggingface_hub import snapshot_download
-
+from llama_cpp import Llama
+import sys
+from io import StringIO
+from transformers import pipeline
+from datasets import load_dataset
+import torch
+import json
+from pathlib import Path
 
 repo_name = "TheBloke/Mistral-7B-v0.1-GGUF"
 model_file = "mistral-7b-v0.1.Q4_K_M.gguf"
 
-
-#repo_name = 'ECUiVADE/Mistral'
-#model_file = "openhermes-2.5-mistral-7b.Q4_K_M.gguf"
+llm = Llama(model_path=model_file, n_gpu_layers=100, n_ctx=2048)
 
 
-print('Fetching model:', repo_name, model_file)
-snapshot_download(repo_id=repo_name, local_dir=".", allow_patterns=model_file)
-print('Done fetching model:')
+# Global flag for first run
+isFirstRun = True
+context = ""
+history = []  # Initialize history
+initContext = """<|im_start|>You are playing the role of an aggressive patient called Barry Lawson. Barry has been waiting for hours at the hospital with severe pain in his chest and hasn't been seen yet. You are incredibly frustrated and angry that things are taking so long, you will only calm down if your concerns are addressed. Waiting will only make you more angry and aggressive towards the staff. At any moment you could break into a RAGE! Make sure you don't censor swear words. Make sure to explain your emotional state and thinking before responding, for example, Barry: (tired of waiting and very angry) What the fuck do I have to do around here to get some treatment!"""
+feedback_file = Path("/content/datalog.json")
 
-DEFAULT_MODEL_PATH = model_file
+def AIPatient(message):
 
-from llama_cpp import Llama
-llm = Llama(model_path=model_file, model_type="mistral")
+    global isFirstRun, history,context
 
+    if isFirstRun:
+      context = initContext
+      isFirstRun = False
+    #else:
+      #for turn in history:
+      #  context += f"\n<|im_start|> Nurse: {turn[0]}\n<|im_start|> Barry: {turn[1]}"
+    context += """
+                  <|im_start|>nurse
+                  Nurse: """+message+"""
+                  <|im_start|>barry
+                  Barry:
+                  """
 
-def predict(input, chatbot, max_length, top_p, temperature, history):
-    chatbot.append((input, ""))
     response = ""
-    history.append(input)
-
-    for output in llm(input, stream=True, temperature=temperature, top_p=top_p, max_tokens=max_length, ):
-        piece = output['choices'][0]['text']
-        response += piece
-        chatbot[-1] = (chatbot[-1][0], response)
-
-        yield chatbot, history
-
-    history.append(response)
-    yield chatbot, history
-
-
-def reset_user_input():
-    return gr.update(value="")
+    # Here, you should add the code to generate the response using your model
+    # For example:
+    while(len(response) < 1):
+      output = llm(context, max_tokens=400, stop=["Nurse:"], echo=False)
+      response = output["choices"][0]["text"]
+      response = response.strip()
 
+    with feedback_file.open("a") as f:
+            f.write(json.dumps({"Nurse": message, "Barry": response},indent=4))
+            f.write("\n")
 
-def reset_state():
-    return [], []
+    context += response
+    print (context)
 
+    history.append((message,response))
+    return history
 
 with gr.Blocks() as demo:
-    gr.HTML("""<h1 align="center">So Rude</h1>""")
+    gr.Markdown("# AI Patient Chatbot")
+    with gr.Group():
+        with gr.Tab("Patient Chatbot"):
+            chatbot = gr.Chatbot()
+            message = gr.Textbox(label="Enter your message to Barry", placeholder="Type here...", lines=2)
+            send_message = gr.Button("Submit")
+            send_message.click(AIPatient, inputs=[message], outputs=[chatbot])
+            save_chatlog = gr.Button("Save Chatlog")
+            #send_message.click(SaveChatlog, inputs=[message], outputs=[chatbot])
 
-    chatbot = gr.Chatbot()
-    with gr.Row():
-        with gr.Column(scale=4):
-            user_input = gr.Textbox(show_label=False, placeholder="Input...", lines=8, elem_id="user_input")
-            submitBtn = gr.Button("Submit", variant="primary", elem_id="submit_btn")
-        with gr.Column(scale=1):
-            max_length = gr.Slider(0, 256, value=64, step=1.0, label="Maximum Length", interactive=True)
-            top_p = gr.Slider(0, 1, value=0.7, step=0.01, label="Top P", interactive=True)
-            temperature = gr.Slider(0, 2.0, value=0.95, step=0.01, label="Temperature", interactive=True)
-            emptyBtn = gr.Button("Clear History")
 
-    history = gr.State([])
+            #message.submit(AIPatient, inputs=[message], outputs=[chatbot])
 
-    submitBtn.click(
-        predict, [user_input, chatbot, max_length, top_p, temperature, history], [chatbot, history], show_progress=True
-    )
-    submitBtn.click(reset_user_input, [], [user_input])
+demo.launch(debug=True)
 
-    emptyBtn.click(reset_state, outputs=[chatbot, history], show_progress=True)
 
-demo.queue().launch(share=False, inbrowser=True)