Create app.py

app.py

@@ -0,0 +1,95 @@
+import gradio as gr
+import pandas as pd
+import os
+import tempfile
+from typing import List, Dict, Any
+import json
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_community.vectorstores import Chroma
+from langchain_community.embeddings import HuggingFaceEmbeddings
+from langchain.schema import Document
+from langchain.chains import RetrievalQA
+import logging
+import uuid
+import docx
+import PyPDF2
+import openpyxl
+import pptx
+import shutil
+import re
+from transformers import pipeline
+
+# Set up logging
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+CHROMA_DB_DIR = "./chroma_db"
+
+class HFZeroGPULLM:
+    def __init__(self, model_id="mistralai/Mistral-7B-Instruct-v0.1"):
+        try:
+            self.generator = pipeline("text-generation", model=model_id, device=-1)
+            logger.info("Loaded HuggingFace text-generation pipeline on CPU.")
+        except Exception as e:
+            logger.error(f"Failed to load HuggingFace pipeline: {e}")
+            self.generator = None
+
+    def invoke(self, prompt):
+        if not self.generator:
+            raise RuntimeError("HFZeroGPULLM not initialized properly.")
+        result = self.generator(prompt, max_new_tokens=512, do_sample=True)[0]
+        return result['generated_text'] if 'generated_text' in result else result['text']
+
+class CSVRAGSystem:
+    def __init__(self):
+        self.vectorstore = None
+        self.qa_chain = None
+        self.uploaded_files = []
+        self.text_splitter = RecursiveCharacterTextSplitter(
+            chunk_size=1000,
+            chunk_overlap=200,
+            length_function=len,
+        )
+
+        # Initialize HuggingFace LLM (CPU-based)
+        try:
+            self.llm = HFZeroGPULLM()
+            logger.info("HuggingFace LLM initialized successfully.")
+        except Exception as e:
+            logger.error(f"Failed to initialize HuggingFace LLM: {e}")
+            self.llm = None
+
+        # Always try to load persistent ChromaDB
+        self.load_vectorstore()
+
+    def load_vectorstore(self):
+        try:
+            if os.path.exists(CHROMA_DB_DIR) and os.listdir(CHROMA_DB_DIR):
+                embeddings = HuggingFaceEmbeddings(
+                    model_name="sentence-transformers/all-MiniLM-L6-v2",
+                    model_kwargs={'device': 'cpu'}
+                )
+                self.vectorstore = Chroma(
+                    embedding_function=embeddings,
+                    persist_directory=CHROMA_DB_DIR
+                )
+                if self.llm:
+                    self.qa_chain = RetrievalQA.from_chain_type(
+                        llm=self.llm,
+                        chain_type="stuff",
+                        retriever=self.vectorstore.as_retriever(search_kwargs={"k": 3}),
+                        return_source_documents=True
+                    )
+                logger.info("Loaded persistent ChromaDB vectorstore.")
+            else:
+                logger.info("No existing ChromaDB found. Will create on first upload.")
+        except Exception as e:
+            logger.error(f"Error loading persistent ChromaDB: {e}")
+
+    # [REMAINDER OF CODE UNCHANGED ... your previous class logic continues here]
+
+# NOTE:
+# - The Ollama import was removed.
+# - Replaced Ollama usage with `HFZeroGPULLM` that uses Hugging Face Transformers.
+# - You can adjust the `model_id` (e.g., to llama2 models or phi models) depending on availability.
+# - Ensure `transformers` is added to requirements.txt