improved code

app.py

@@ -12,16 +12,21 @@ import streamlit as st
 FILE_PATH = "anjibot_chunks.json"
 BATCH_SIZE = 384
 INDEX_NAME = "groq-llama-3-rag"
-PINECONE_API_KEY = os.getenv["PINECONE_API_KEY"]
-GROQ_API_KEY = os.getenv["GROQ_API_KEY"]
+PINECONE_API_KEY = os.getenv("PINECONE_API_KEY") 
+GROQ_API_KEY = os.getenv("GROQ_API_KEY")
 DIMENSIONS = 768
 
+# Load data once at the start
+data = load_data(FILE_PATH)
+
+# Initialize Pinecone and SentenceTransformer once
+index = initialize_pinecone(PINECONE_API_KEY, INDEX_NAME, DIMENSIONS)
+encoder = SentenceTransformer('dwzhu/e5-base-4k')
 
 def load_data(file_path: str) -> dict:
     with open(file_path, 'r') as file:
         return json.load(file)
 
-
 def initialize_pinecone(api_key: str, index_name: str, dims: int) -> any:
     pc = Pinecone(api_key=api_key)
     spec = ServerlessSpec(cloud="aws", region='us-east-1')
@@ -38,10 +43,7 @@ def initialize_pinecone(api_key: str, index_name: str, dims: int) -> any:
 
     return pc.Index(index_name)
 
-
 def upsert_data_to_pinecone(index: any, data: dict):
-    encoder = SentenceTransformer('dwzhu/e5-base-4k')
-    
     for i in tqdm(range(0, len(data['id']), BATCH_SIZE)):
         # Find end of batch
         i_end = min(len(data['id']), i + BATCH_SIZE)
@@ -60,17 +62,15 @@ def upsert_data_to_pinecone(index: any, data: dict):
         to_upsert = list(zip(batch["id"], embeds, batch["metadata"]))
         index.upsert(vectors=to_upsert)
 
-
 def get_docs(query: str, index: any, encoder: any, top_k: int) -> list[str]:
     xq = encoder.encode(query)
     res = index.query(vector=xq.tolist(), top_k=top_k, include_metadata=True)
     return [x["metadata"]['content'] for x in res["matches"]]
 
-
 def get_response(query: str, docs: list[str], groq_client: any) -> str:
     system_message = (
-        "You are Anjibot, the AI course rep of 400 Level Computer Science department. You are always helpful, jovial, can be sarcastica but still sweet.\n"
-        "Provide the answer to class related queries using\n"
+        "You are Anjibot, the AI course rep of 400 Level Computer Science department. You are always helpful, jovial, can be sarcastic but still sweet.\n"
+        "Provide the answer to class-related queries using\n"
         "context provided below.\n"
         "If you don't the answer to the user's question based on your pretrained knowledge and the context provided, just direct the user to Anji the human course rep.\n"
         "Anji's phone number: 08145170886.\n\n"
@@ -88,19 +88,11 @@ def get_response(query: str, docs: list[str], groq_client: any) -> str:
     )
     return chat_response.choices[0].message.content
 
-
 def handle_query(user_query: str):
-    # Load data
-    data = load_data(FILE_PATH)
-
-    # Initialize Pinecone
-    index = initialize_pinecone(PINECONE_API_KEY, INDEX_NAME, DIMENSIONS)
-
-    # Upsert data into Pinecone
+    # Upsert data into Pinecone (if necessary)
     upsert_data_to_pinecone(index, data)
 
-    # Initialize encoder and Groq client
-    encoder = SentenceTransformer('dwzhu/e5-base-4k')
+    # Initialize Groq client
     groq_client = Groq(api_key=GROQ_API_KEY)
 
     # Get relevant documents
@@ -132,5 +124,3 @@ def main():
 
 if __name__ == "__main__":
     main()
-
-