Update app/vectorstore.py

app/vectorstore.py

@@ -29,7 +29,7 @@ class vectorstore:
             self.index_path = path
             self.vectorstore = self.load_vectorstore(self.index_path)
             
-    def chunk_text(self,text, chunk_size=800):
+    def chunk_text(self,text, chunk_size=700):
         """
         Split text into chunks of approximately 'chunk_size' words.
         
@@ -47,7 +47,7 @@ class vectorstore:
             chunks.append(chunk)
         return chunks
 
-    def create_vectorstore_from_pdf(self,pdf_path, chunk_size=800, embedder_model="all-MiniLM-L6-v2"):
+    def create_vectorstore_from_pdf(self,pdf_path, chunk_size=700, embedder_model="all-MiniLM-L6-v2"):
         """
         Extract text from a PDF using pypdf, chunk the text, compute embeddings,
         and create a FAISS index.
@@ -78,7 +78,7 @@ class vectorstore:
         index_path = self.save_vectorstore_with_timestamp_and_without(vectorstore, embedder_model)
         return index_path
 
-    def create_vectorstore_from_website(self, url, chunk_size=800, embedder_model="all-MiniLM-L6-v2"):
+    def create_vectorstore_from_website(self, url, chunk_size=700, embedder_model="all-MiniLM-L6-v2"):
         """
         Fetch text from a website, chunk the text, compute embeddings,
         and create a FAISS index.
@@ -109,7 +109,7 @@ class vectorstore:
         index_path = self.save_vectorstore_with_timestamp_and_without(vectorstore, embedder_model)
         return index_path
 
-    def add_to_vectorstore_web(self, url, index_file_path, chunk_size=800):
+    def add_to_vectorstore_web(self, url, index_file_path, chunk_size=700):
         """
         Fetch text from a website, chunk the text, compute embeddings using the existing embedder,
         and add them to the existing FAISS vectorstore.
@@ -141,7 +141,7 @@ class vectorstore:
         return updated_index_path
 
 
-    def add_to_vectorstore_from_pdf(self, pdf_path, chunk_size=800):
+    def add_to_vectorstore_from_pdf(self, pdf_path, chunk_size=700):
         """
         Extract text from a PDF using PyPDF2, chunk the text, compute embeddings using the existing embedder,
         and add them to the existing FAISS vectorstore.
@@ -206,7 +206,7 @@ class vectorstore:
         return vectorstore
     
     
-    def search_vectorstore(self, query, top_k=6):
+    def search_vectorstore(self, query, top_k=5):
         # Search using LangChain's similarity_search
         docs = self.vectorstore.similarity_search(query, k=top_k)
         results = [doc.page_content for doc in docs]