new code

src/pdfchatbot.py

@@ -11,7 +11,7 @@ from langchain.document_loaders import PyPDFLoader
 from langchain.prompts import PromptTemplate
 from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
 import spaces
-from langchain_text_splitters import CharacterTextSplitter
+from langchain_text_splitters import CharacterTextSplitter,RecursiveCharacterTextSplitter
 
 
 class PDFChatBot:
@@ -96,7 +96,12 @@ class PDFChatBot:
         """
         Load the vector database from the documents and embeddings.
         """
-        text_splitter = CharacterTextSplitter(chunk_size=self.chunk_size, chunk_overlap=100)
+        text_splitter = RecursiveCharacterTextSplitter(
+            chunk_size=256,
+            chunk_overlap=100,
+            length_function=len,
+            add_start_index=True,
+        )
         docs = text_splitter.split_documents(self.documents)
         self.vectordb = Chroma.from_documents(docs, self.embeddings)
 
@@ -132,7 +137,8 @@ class PDFChatBot:
     def create_organic_pipeline(self):
         self.pipeline = pipeline(
             "text-generation",
-            model=self.config.get("autoModelForCausalLM"),
+            model=self.model,
+            tokenizer=self.tokenizer,
             model_kwargs={"torch_dtype": torch.bfloat16},
             device="cuda",
         )