Update app.py

app.py

@@ -13,17 +13,20 @@ import gradio as gr
 
 class DocumentRetrievalAndGeneration:
     def __init__(self, embedding_model_name, lm_model_id, data_folder, faiss_index_path):
-        self.documents = self.load_documents(data_folder)
+        self.all_splits = self.load_documents(data_folder)
         self.embeddings = SentenceTransformer(embedding_model_name)
         self.gpu_index = self.load_faiss_index(faiss_index_path)
         self.llm = self.initialize_llm(lm_model_id)
-        self.all_splits = self.split_documents()
+        # self.all_splits = self.split_documents()
 
     def load_documents(self, folder_path):
         loader = DirectoryLoader(folder_path, loader_cls=TextLoader)
+        text_splitter = RecursiveCharacterTextSplitter(chunk_size=5000, chunk_overlap=250)
         documents = loader.load()
+        all_splits = text_splitter.split_documents(documents)
         print('Length of documents:', len(documents))
-        return documents
+        print("LEN of all_splits", len(all_splits))
+        return all_splits
 
     def load_faiss_index(self, faiss_index_path):
         cpu_index = faiss.read_index(faiss_index_path)
@@ -51,12 +54,6 @@ class DocumentRetrievalAndGeneration:
         )
         return generate_text
 
-    def split_documents(self):
-        text_splitter = RecursiveCharacterTextSplitter(chunk_size=5000, chunk_overlap=250)
-        all_splits = text_splitter.split_documents(self.documents)
-        print("LEN of all_splits", len(all_splits))
-        return all_splits
-
     def query_and_generate_response(self, query):
         query_embedding = self.embeddings.encode(query, convert_to_tensor=True).cpu().numpy()
         distances, indices = self.gpu_index.search(np.array([query_embedding]), k=5)
@@ -65,6 +62,7 @@ class DocumentRetrievalAndGeneration:
         for idx in indices[0]:
             content += "-" * 50 + "\n"
             content += self.all_splits[idx].page_content + "\n"
+            print("CHUNK",idx)
             print(self.all_splits[idx].page_content)
             print("############################")
         prompt=f"""
@@ -136,7 +134,7 @@ if __name__ == "__main__":
                 font-weight: bold; /* Make text bold */
             }
             """
-        EXAMPLES = ["Does the VIP modules & CSI2 module could work simultaneously? ", 
+        EXAMPLES = ["Can the VIP and CSI2 modules operate simultaneously? ", 
                     "I'm using Code Composer Studio 5.4.0.00091 and enabled FPv4SPD16 floating point support for CortexM4 in TDA2. However, after building the project, the .asm file shows --float_support=vfplib instead of FPv4SPD16. Why is this happening?", 
                     "Could you clarify the maximum number of cameras that can be connected simultaneously to the video input ports on the TDA2x SoC, considering it supports up to 10 multiplexed input ports and includes 3 dedicated video input modules?"]