Update app.py

app.py

@@ -1,4 +1,5 @@
 import streamlit as st
+import pandas as pd
 from llama_index.core import StorageContext, load_index_from_storage, VectorStoreIndex, SimpleDirectoryReader, ChatPromptTemplate
 from llama_index.llms.huggingface import HuggingFaceInferenceAPI
 from llama_index.embeddings.huggingface import HuggingFaceEmbedding
@@ -8,7 +9,6 @@ import shutil
 import os
 import time
 
-
 icons = {"assistant": "robot.png", "user": "man-kddi.png"}
 
 # Configure the Llama index settings
@@ -17,7 +17,6 @@ Settings.llm = HuggingFaceInferenceAPI(
     tokenizer_name="meta-llama/Meta-Llama-3-8B-Instruct",
     context_window=3900,
     token=os.getenv("HF_TOKEN"),
-    # max_new_tokens=1000,
     generate_kwargs={"temperature": 0.1},
 )
 Settings.embed_model = HuggingFaceEmbedding(
@@ -33,33 +32,39 @@ os.makedirs(DATA_DIR, exist_ok=True)
 os.makedirs(PERSIST_DIR, exist_ok=True)
 
 def data_ingestion():
-    documents = SimpleDirectoryReader(DATA_DIR).load_data()
+    documents = []
+
+    # Load documents from the data directory
+    documents += SimpleDirectoryReader(DATA_DIR).load_data()
+
+    # Process and load CSV files
+    for file in os.listdir(DATA_DIR):
+        if file.endswith(".csv"):
+            csv_path = os.path.join(DATA_DIR, file)
+            df = pd.read_csv(csv_path)
+            # Convert DataFrame to a list of text strings (or any other format suitable for your embeddings)
+            csv_texts = df.apply(lambda row: " ".join(row.astype(str)), axis=1).tolist()
+            documents += csv_texts
+
     storage_context = StorageContext.from_defaults()
     index = VectorStoreIndex.from_documents(documents)
     index.storage_context.persist(persist_dir=PERSIST_DIR)
 
 def remove_old_files():
-    # Specify the directory path you want to clear
     directory_path = "data"
-
-    # Remove all files and subdirectories in the specified directory
     shutil.rmtree(directory_path)
-
-    # Recreate an empty directory if needed
     os.makedirs(directory_path)
 
 def extract_transcript_details(youtube_video_url):
     try:
-        video_id=youtube_video_url.split("=")[1]
-        
-        transcript_text=YouTubeTranscriptApi.get_transcript(video_id)
+        video_id = youtube_video_url.split("=")[1]
+        transcript_text = YouTubeTranscriptApi.get_transcript(video_id)
 
         transcript = ""
         for i in transcript_text:
             transcript += " " + i["text"]
        
         return transcript
-
     except Exception as e:
         st.error(e)
 
@@ -67,21 +72,20 @@ def handle_query(query):
     storage_context = StorageContext.from_defaults(persist_dir=PERSIST_DIR)
     index = load_index_from_storage(storage_context)
     chat_text_qa_msgs = [
-    (
-        "user",
-        """You are Q&A assistant named CHATTO, created by Pachaiappan [linkdin](https://www.linkedin.com/in/pachaiappan) an AI Specialist. Your main goal is to provide answers as accurately as possible, based on the instructions and context you have been given. If a question does not match the provided context or is outside the scope of the document, you only say the user to 'Please ask a questions within the context of the document'.
-        Context:
-        {context_str}
-        Question:
-        {query_str}
-        """
-    )
+        (
+            "user",
+            """You are Q&A assistant named CHATTO, created by Pachaiappan [linkdin](https://www.linkedin.com/in/pachaiappan) an AI Specialist. Your main goal is to provide answers as accurately as possible, based on the instructions and context you have been given. If a question does not match the provided context or is outside the scope of the document, you only say the user to 'Please ask a questions within the context of the document'.
+            Context:
+            {context_str}
+            Question:
+            {query_str}
+            """
+        )
     ]
     text_qa_template = ChatPromptTemplate.from_messages(chat_text_qa_msgs)
     query_engine = index.as_query_engine(text_qa_template=text_qa_template)
     answer = query_engine.query(query)
     
-    
     if hasattr(answer, 'response'):
         return answer.response
     elif isinstance(answer, dict) and 'response' in answer:
@@ -94,7 +98,6 @@ def streamer(text):
         yield i
         time.sleep(0.001)
 
-
 # Streamlit app initialization
 st.title("Chat with your PDF📄")
 st.markdown("**Built by [Pachaiappan❤️](https://mr-vicky-01.github.io/Portfolio/)**")
@@ -108,15 +111,15 @@ for message in st.session_state.messages:
 
 with st.sidebar:
     st.title("Menu:")
-    uploaded_file = st.file_uploader("Upload your PDF Files and Click on the Submit & Process Button")
+    uploaded_file = st.file_uploader("Upload your PDF/CSV Files and Click on the Submit & Process Button")
     video_url = st.text_input("Enter Youtube Video Link: ")
     if st.button("Submit & Process"):
         with st.spinner("Processing..."):
-            if len(os.listdir("data")) !=0:
+            if len(os.listdir("data")) != 0:
                 remove_old_files()
                 
             if uploaded_file:
-                filepath = "data/saved_pdf.pdf"
+                filepath = os.path.join(DATA_DIR, uploaded_file.name)
                 with open(filepath, "wb") as f:
                     f.write(uploaded_file.getbuffer())
         
@@ -125,10 +128,10 @@ with st.sidebar:
                 with open("data/saved_text.txt", "w") as file:
                     file.write(extracted_text)
                 
-            data_ingestion()  # Process PDF every time new file is uploaded
+            data_ingestion()  # Process every time new file is uploaded
             st.success("Done")
 
-user_prompt = st.chat_input("Ask me anything about the content of the PDF:")
+user_prompt = st.chat_input("Ask me anything about the content of the PDF/CSV:")
 
 if user_prompt and (uploaded_file or video_url):
     st.session_state.messages.append({'role': 'user', "content": user_prompt})
@@ -140,4 +143,3 @@ if user_prompt and (uploaded_file or video_url):
         response = handle_query(user_prompt)
     with st.chat_message("user", avatar="robot.png"):
         st.write_stream(streamer(response))
-    st.session_state.messages.append({'role': 'assistant', "content": response})