Update app.py

app.py

@@ -1,6 +1,6 @@
 from fastapi import FastAPI, HTTPException, Query
 from fastapi.responses import JSONResponse
-from webscout import WEBS, transcriber, LLM, FastAI
+from webscout import WEBS, transcriber, LLM
 from typing import Optional, List, Dict, Union
 from fastapi.encoders import jsonable_encoder
 from bs4 import BeautifulSoup
@@ -151,20 +151,31 @@ async def llm_chat(
     except Exception as e:
         raise HTTPException(status_code=500, detail=f"Error during LLM chat: {e}")
 
-fast_ai = FastAI()
+def snova_ai(user, model="llama3-70b", system="Answer as concisely as possible."):
+    env_type = "tp16405b" if "405b" in model else "tp16"
+    data = {'body': {'messages': [{'role': 'system', 'content': system}, {'role': 'user', 'content': user}], 'stream': True, 'model': model}, 'env_type': env_type}
+    with requests.post('https://fast.snova.ai/api/completion', headers={'content-type': 'application/json'}, json=data, stream=True) as response:
+        output = ''
+        for line in response.iter_lines(decode_unicode=True):
+            if line.startswith('data:'):
+                try:
+                    data = json.loads(line[len('data: '):])
+                    output += data.get("choices", [{}])[0].get("delta", {}).get("content", '')
+                except json.JSONDecodeError:
+                    if line[len('data: '):] == '[DONE]':
+                        break
+        return output
 
 @app.get("/api/fastAI")
-async def fast_ai_endpoint(user: str, model: str = "llama3-70b", system: str = "Answer as concisely as possible."):
+async def fast_ai(user: str, model: str = "llama3-70b", system: str = "Answer as concisely as possible."):
     """Get a response from the Snova AI service."""
-    fast_ai.model = model
-    fast_ai.system = system
     try:
-        # Use asyncio.to_thread to run generate_response in a separate thread
-        response = await asyncio.to_thread(fast_ai.generate_response, user)
+        response = await asyncio.to_thread(snova_ai, user, model, system)
         return JSONResponse(content={"response": response})
     except Exception as e:
         raise HTTPException(status_code=500, detail=f"Error during Snova AI request: {e}")
 
+
     
 @app.get("/api/answers")
 async def answers(q: str, proxy: Optional[str] = None):