Update functions.py

functions.py

@@ -69,25 +69,22 @@ async def handle_userinput(user_question, custom_graph):
 
 
 
-def create_retriever_from_chroma(vectorstore_path="./docs/chroma/", search_type='mmr', k=7, chunk_size=300, chunk_overlap=30,lambda_mult= 0.7):
-    
+def create_retriever_from_chroma(vectorstore_path="./docs/chroma/", search_type='mmr', k=7, chunk_size=300, chunk_overlap=30, lambda_mult=0.7):
     model_name = "Alibaba-NLP/gte-large-en-v1.5"
-    model_kwargs = {'device': 'cpu',
-                   "trust_remote_code" : 'False'}
+    model_kwargs = {'device': 'cpu', "trust_remote_code": 'False'}
     encode_kwargs = {'normalize_embeddings': True}
+
     embeddings = HuggingFaceEmbeddings(
         model_name=model_name,
         model_kwargs=model_kwargs,
         encode_kwargs=encode_kwargs
     )
 
-    
-
     if os.path.exists(vectorstore_path) and os.listdir(vectorstore_path):
-        vectorstore = Chroma(persist_directory=vectorstore_path,embedding_function=embeddings)
-        
+        vectorstore = Chroma(persist_directory=vectorstore_path, embedding_function=embeddings)
     else:
-        st.write("Vector store doesnt exist and will be created now")
+        st.write("Vector store doesn't exist and will be created now")
+
         urls = [
  
 "https://github.com/zedr/clean-code-python",
@@ -190,38 +187,33 @@ def create_retriever_from_chroma(vectorstore_path="./docs/chroma/", search_type=
     "https://datasciencedojo.com/blog/ensemble-methods-in-machine-learning/",
     "https://datasciencedojo.com/blog/langgraph-tutorial/",
     "https://datasciencedojo.com/blog/data-driven-marketing-in-2024/",
-    "https://datasciencedojo.com/blog/on-device-ai/"
-        ]
+    "https://datasciencedojo.com/blog/on-device-ai/",
+    
     
-    def extract_sentences_from_web(links, chunk_size=500, chunk_overlap=30):
-        data = []
-        for link in links:
-            loader = NewsURLLoader(urls=[link])
-            data += loader.load()
-        return data    
+]
+        
+        def extract_sentences_from_web(links, chunk_size=500, chunk_overlap=30):
+            data = []
+            for link in links:
+                loader = NewsURLLoader(urls=[link])
+                data += loader.load()
+            return data
 
+        docs = extract_sentences_from_web(links=urls)
 
-    docs = extract_sentences_from_web(links=urls)
-        
-        
-    text_splitter = RecursiveCharacterTextSplitter.from_tiktoken_encoder(
-    chunk_size=chunk_size, chunk_overlap=chunk_overlap,
-    separators=["\n\n \n\n","\n\n\n", "\n\n", r"In \[[0-9]+\]", r"\n+", r"\s+"],
-        is_separator_regex = True
-    )
-    split_docs = text_splitter.split_documents(docs)
+        text_splitter = RecursiveCharacterTextSplitter.from_tiktoken_encoder(
+            chunk_size=chunk_size, chunk_overlap=chunk_overlap,
+            separators=["\n\n \n\n", "\n\n\n", "\n\n", r"In \[[0-9]+\]", r"\n+", r"\s+"],
+            is_separator_regex=True
+        )
+        split_docs = text_splitter.split_documents(docs)
 
-        
-    vectorstore = Chroma.from_documents(
+        vectorstore = Chroma.from_documents(
             documents=split_docs, embedding=embeddings, persist_directory=vectorstore_path
         )
-        
-    
-    retriever=vectorstore.as_retriever(search_type = search_type, search_kwargs={"k": k})
 
+    retriever = vectorstore.as_retriever(search_type=search_type, search_kwargs={"k": k})
     
-    
-
     return retriever