Update app.py

app.py

@@ -17,27 +17,28 @@ import bs4
 from langchain_core.rate_limiters import InMemoryRateLimiter
 from urllib.parse import urljoin
 
-rate_limiter = InMemoryRateLimiter(
-    requests_per_second=0.1,  # <-- MistralAI free. We can only make a request once every second
-    check_every_n_seconds=0.01,  # Wake up every 100 ms to check whether allowed to make a request,
-    max_bucket_size=10,  # Controls the maximum burst size.
-)
-
-retriever = ArxivRetriever(
-    load_max_docs=2,
-    get_ful_documents=True,
-)
-
-# LLM model
-llm = ChatMistralAI(model="mistral-large-latest", rate_limiter=rate_limiter)
-
-# Embeddings
-embed_model = "sentence-transformers/multi-qa-distilbert-cos-v1"
-# embed_model = "nvidia/NV-Embed-v2"
-embeddings = HuggingFaceInstructEmbeddings(model_name=embed_model)
-# embeddings = MistralAIEmbeddings()
 
 def initialize(arxivcode):  
+    rate_limiter = InMemoryRateLimiter(
+        requests_per_second=0.1,  # <-- MistralAI free. We can only make a request once every second
+        check_every_n_seconds=0.01,  # Wake up every 100 ms to check whether allowed to make a request,
+        max_bucket_size=10,  # Controls the maximum burst size.
+    )
+    
+    retriever = ArxivRetriever(
+        load_max_docs=2,
+        get_ful_documents=True,
+    )
+    
+    # LLM model
+    llm = ChatMistralAI(model="mistral-large-latest", rate_limiter=rate_limiter)
+    
+    # Embeddings
+    embed_model = "sentence-transformers/multi-qa-distilbert-cos-v1"
+    # embed_model = "nvidia/NV-Embed-v2"
+    embeddings = HuggingFaceInstructEmbeddings(model_name=embed_model)
+    # embeddings = MistralAIEmbeddings()
+    
     docs = retriever.invoke(str(arxivcode))
     for i in range(len(docs)): 
         docs[i].metadata['Published'] = str(docs[i].metadata['Published'])