Update app.py

Add langfuse log.

app.py

@@ -3,6 +3,8 @@ from threading import Thread
 from typing import Iterator
 
 import gradio as gr
+from langfuse import Langfuse
+from langfuse.decorators import observe
 import spaces
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
@@ -44,6 +46,16 @@ pre, code {
 
 system_prompt = str(os.getenv("SYSTEM_PROMPT"))
 
+secret_key = str(os.getenv("LANGFUSE_SECRET_KEY"))
+public_key = str(os.getenv("LANGFUSE_PUBLIC_KEY"))
+host = str(os.getenv("LANGFUSE_HOST"))
+
+langfuse = Langfuse(
+  secret_key=secret_key,
+  public_key=public_key,
+  host=host
+)
+
 
 def execution_time_calculator(start_time, log=True):
     delta = time.time() - start_time
@@ -70,6 +82,10 @@ def get_generation_speed():
 
     return generation_speed
 
+@observe()
+def log_to_langfuse(message, chat_history, max_new_tokens, temperature, top_p, top_k, repetition_penalty, do_sample, model_outputs, generation_speed):
+    return  "".join(model_outputs) 
+
 
 @spaces.GPU
 def generate(
@@ -128,9 +144,13 @@ def generate(
 
     generation_speed = token_per_second_calculator(sum_tokens, time_delta)
 
+    log_function = log_to_langfuse(message, chat_history, max_new_tokens, temperature, top_p, top_k, repetition_penalty, do_sample, outputs, generation_speed)
+
     print(f"generation_speed: {generation_speed}")
 
 
+
+
 chatbot = gr.Chatbot(placeholder=PLACEHOLDER, scale=1, show_copy_button=True, height="68%", rtl=True) #,  elem_classes=["chatbot"])
 chat_input = gr.Textbox(show_label=False, lines=2, rtl=True, placeholder="ورودی", show_copy_button=True, scale=4)
 submit_btn = gr.Button(variant="primary", value="ارسال", size="sm", scale=1, elem_classes=["_button"])