Update app.py

app.py

@@ -1,12 +1,15 @@
+import os
 from threading import Thread
 from typing import Iterator
 import gradio as gr
+import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
+from modelGuards.suicideModel import predictSuicide
 from openai import OpenAI
-import requests, json
 
 MAX_MAX_NEW_TOKENS = 2048
 DEFAULT_MAX_NEW_TOKENS = 1024
+MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 DEFAULT_SYSTEM_PROMPT = """\
 You are a helpful and joyous mental therapy assistant. Always answer as helpfully and cheerfully as possible, while being safe.  Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content.Please ensure that your responses are socially unbiased and positive in nature.\n\nIf a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information.
 """
@@ -14,56 +17,32 @@ You are a helpful and joyous mental therapy assistant. Always answer as helpfull
 DESCRIPTION = """
 # LLama-2-Mental-Therapy-Chatbot
 """
+LICENSE = "open-source"
 
-client = OpenAI(
-    base_url="http://192.168.3.74:8080/v1",
-    api_key="-"
-)
-    
-def response_guard(text):
-    url = 'http://192.168.3.74:6006/safety'
-    data = {'message': text}
-    response = requests.post(url, data=json.dumps(data), headers={'Content-Type': 'application/json'})
-    if response.status_code == 200:
-        result = response.json()
-        return(result)
-
-    
+from llamaModel.model import get_input_token_length, get_LLAMA_response_stream
 def generate(
     message: str,
     chat_history: list[tuple[str, str]],
     system_prompt: str,
     max_new_tokens: int = 1024,
-    temperature: float = 1,
+    temperature: float = 0.6,
     top_p: float = 0.9,
+    top_k: int = 50
 ) -> Iterator[str]:
-    llmGuardCheck = response_guard(message)
-    if(llmGuardCheck != "safe"):
-        raise gr.Error(llmGuardCheck)
-        yield(llmGuardCheck)
+    if os.getenv("PREDICT_SUICIDE")=="True" and predict_suicide(message)=='suicide':
+        yield("I am sorry that you are feeling this way. You need a specialist help. Please consult a nearby doctor.")
     else:
-        messages = []
+        conversation = []
         if system_prompt:
-            messages.append({"role": "system", "content": system_prompt})
+            conversation.append({"role": "system", "content": system_prompt})
         for user, assistant in chat_history:
-            messages.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
-        messages.append({"role": "user", "content": message})
-        chat_completion = client.chat.completions.create(
-            model="tgi", messages=messages, stream=True,max_tokens=max_new_tokens,temperature=temperature,top_p=top_p
-        )
-        response = ""
-        first_chunk = True
-        for chunk in chat_completion:
-            token = chunk.choices[0].delta.content
-            if first_chunk:
-                token= token.strip() ## the first token Has a leading space, due to some bug in TGI
-                response += token
-                yield response
-                first_chunk = False
-            else:
-                if token!="</s>":
-                    response += token
-                    yield response
+            conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
+        conversation.append({"role": "user", "content": message})
+        if(get_input_token_length(conversation) > MAX_INPUT_TOKEN_LENGTH):
+            raise gr.InterfaceError(f"The accumulated input is too long ({get_input_token_length(conversation)} > {MAX_INPUT_TOKEN_LENGTH}). Clear your chat history and try again.")
+        generator = get_LLAMA_response_stream(conversation, max_new_tokens, temperature, top_p, top_k)
+        for response in generator:
+            yield response
 
 
 chat_interface = gr.ChatInterface(
@@ -93,6 +72,13 @@ chat_interface = gr.ChatInterface(
             step=0.05,
             value=0.95,
         ),
+        gr.Slider(
+            label="Top-k",
+            minimum=1,
+            maximum=1000,
+            step=1,
+            value=50,
+        ),
     ],
     stop_btn="Stop",
 )
@@ -100,6 +86,5 @@ chat_interface = gr.ChatInterface(
 with gr.Blocks(css="style.css") as demo:
     gr.Markdown(DESCRIPTION)
     chat_interface.render()
-    
 if __name__ == "__main__":
     demo.queue(max_size=20).launch()