Update app.py

app.py

@@ -5,7 +5,7 @@ from enum import Enum
 import chainlit as cl
 from huggingface_hub import InferenceClient
 from qdrant_client import QdrantClient
-from sentence_transformers import SentenceTransformer
+from openai import OpenAI
 
 
 # ================================
@@ -15,6 +15,7 @@ from sentence_transformers import SentenceTransformer
 HF_TOKEN = os.getenv("HF_TOKEN")
 QDRANT_URL = os.getenv("QDRANT_URL")
 QDRANT_API_KEY = os.getenv("QDRANT_API_KEY")
+OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
 
 MODEL_ID = "Qwen/Qwen2.5-72B-Instruct"
 
@@ -159,6 +160,8 @@ def detect_stage(history: List[Dict[str, str]], user_text: str) -> Stage:
 def check_env():
     if not HF_TOKEN:
         raise ValueError("HF_TOKEN is missing!")
+    if not OPENAI_API_KEY:
+        raise ValueError("OPENAI_API_KEY is missing!")
 
 
 # ================================
@@ -168,14 +171,18 @@ def check_env():
 def get_context(
     query: str,
     q_client: Optional[QdrantClient],
-    encoder: SentenceTransformer,
+    openai_client: OpenAI,
 ) -> str:
 
     if not q_client:
         return ""
 
     try:
-        vector = encoder.encode(query).tolist()
+        response = openai_client.embeddings.create(
+            model="text-embedding-3-small",
+            input=query
+        )
+        vector = response.data[0].embedding
 
         result = q_client.query_points(
             collection_name=QDRANT_COLLECTION,
@@ -240,11 +247,11 @@ async def start():
         except Exception as e:
             print("❌ Qdrant error:", e)
 
-    encoder = SentenceTransformer("all-MiniLM-L6-v2")
+    openai_client = OpenAI(api_key=OPENAI_API_KEY)
 
     cl.user_session.set("hf_client", hf_client)
     cl.user_session.set("q_client", q_client)
-    cl.user_session.set("encoder", encoder)
+    cl.user_session.set("openai_client", openai_client)
     cl.user_session.set("message_history", [])
 
 
@@ -257,7 +264,7 @@ async def main(message: cl.Message):
 
     hf_client: InferenceClient = cl.user_session.get("hf_client")
     q_client: Optional[QdrantClient] = cl.user_session.get("q_client")
-    encoder: SentenceTransformer = cl.user_session.get("encoder")
+    openai_client: OpenAI = cl.user_session.get("openai_client")
 
     history: List[Dict[str, str]] = cl.user_session.get("message_history") or []
 
@@ -280,7 +287,7 @@ async def main(message: cl.Message):
     # RAG
     # =========================
 
-    context = get_context(user_text, q_client, encoder)
+    context = get_context(user_text, q_client, openai_client)
 
     # =========================
     # BUILD MESSAGES
@@ -351,4 +358,16 @@ async def main(message: cl.Message):
 
     except Exception as e:
         await cl.Message(content=f"Произошла ошибка. Попробуйте еще раз или напишите нам напрямую: @alexdev").send()
-        print(f"LLM Error: {e}")
+        print(f"LLM Error: {e}")
+```
+
+Не забудь:
+
+1. Добавить `openai` в `requirements.txt`:
+```
+chainlit
+huggingface_hub
+qdrant-client
+openai
+uvicorn
+websockets