Update app.py

app.py

@@ -1,43 +1,82 @@
+# from fastapi import FastAPI
+# from fastapi.responses import JSONResponse
+# from fastapi import Request
+# from huggingface_hub import InferenceClient
+
+# app = FastAPI()
+
+# @app.post("/")
+# async def greet_json(request: Request):
+#     input_data = await request.json()
+#     # number = input_data.get("number")
+ 
+#     # tripled_number = number * 2
+#     # return {"message": f"Your input number is: {number}, your doubled number is: {tripled_number}"}
+#     user_input = input_data.get("user_input")
+    
+#     client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+#     # Get the response from the model
+#     response = client(user_input)
+    
+#     # assistant_response = client.text_generation(user_input)
+#     assistant_response = "I am assistant."
+#     return {"assistant_message": f"Your input message is: {user_input}, assistant_response is: {response}"}
 from fastapi import FastAPI, HTTPException
 from pydantic import BaseModel
 from huggingface_hub import InferenceClient
 import os
 from datasets import load_dataset
 
+
 app = FastAPI()
 
 # Get the token from the environment variable
 hf_token = os.environ.get("HF_TOKEN")
+
 dataset = load_dataset("Lhumpal/youtube-hunting-beast-transcripts")
 
+
+
 if hf_token:
     client = InferenceClient("HuggingFaceH4/zephyr-7b-beta", token=hf_token)
 else:
     raise ValueError("HF_TOKEN environment variable not set. Please add it as a secret in your Hugging Face Space.")
 
+# Rest of your code...
+
 class ChatRequest(BaseModel):
     message: str
+    history: list[tuple[str, str]] = []
     system_message: str = "You are a friendly Chatbot."
+    max_tokens: int = 512
+    temperature: float = 0.7
+    top_p: float = 0.95
 
 class ChatResponse(BaseModel):
     response: str
-
 @app.post("/chat", response_model=ChatResponse)
 async def chat(request: ChatRequest):
     try:
-        # Use only the latest message, no history
+
         messages = [{"role": "system", "content": request.system_message}]
+        for val in request.history:
+            if val[0]:
+                messages.append({"role": "user", "content": val[0]})
+            if val[1]:
+                messages.append({"role": "assistant", "content": val[1]})
         messages.append({"role": "user", "content": request.message})
 
         response = ""
         for message in client.chat_completion(
             messages,
-            max_tokens=512,
-            temperature=0.7,
+            max_tokens=request.max_tokens,
+            stream=True,
+            temperature=request.temperature,
+            top_p=request.top_p,
         ):
             token = message.choices[0].delta.content
             response += token
 
         return {"response": response}
     except Exception as e:
-        raise HTTPException(status_code=500, detail=str(e))
+        raise HTTPException(status_code=500, detail=str(e))