Update app.py

app.py

@@ -1,70 +1,83 @@
 import gradio as gr
 import os
 import torch
-from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
+from transformers import pipeline
 from langchain.embeddings import SentenceTransformerEmbeddings
 from langchain.vectorstores import Chroma
 from langchain.llms.huggingface_pipeline import HuggingFacePipeline
 from langchain.chains import RetrievalQA
 from langchain.document_loaders import PDFMinerLoader
 from langchain.text_splitter import RecursiveCharacterTextSplitter
-from chromadb.config import Settings
+from chromadb.utils.embedding_functions import OpenAIEmbeddingFunction
+import chromadb
 
-# Initialize Chroma settings once
-CHROMA_SETTINGS = Settings(
-    chroma_db_impl='duckdb+parquet',
-    persist_directory="db",
-    anonymized_telemetry=False
-)
-
-# Initialize the Chroma database on app start (assuming the database will be initialized only once)
-def init_db_if_not_exists(pdf_path):
-    try:
-        # Check if the database exists and load it
-        db = Chroma(persist_directory=CHROMA_SETTINGS.persist_directory, client_settings=CHROMA_SETTINGS)
-        db.get_collection()  # This line will raise an error if the collection doesn't exist
-    except Exception:
-        # If not, initialize the database
-        loader = PDFMinerLoader(pdf_path)
-        documents = loader.load()
-        text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
-        texts = text_splitter.split_documents(documents)
-        embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
-        db = Chroma.from_documents(texts, embeddings, persist_directory=CHROMA_SETTINGS.persist_directory)
-        db.persist()
+# Define Chroma Settings
+CHROMA_SETTINGS = {
+    "chroma_db_impl": "duckdb+parquet",
+    "persist_directory": "db",
+    "anonymized_telemetry": False
+}
 
-# Load model and create pipeline once
+# Load model and tokenizer
 checkpoint = "MBZUAI/LaMini-Flan-T5-783M"
 tokenizer = AutoTokenizer.from_pretrained(checkpoint)
-base_model = AutoModelForSeq2SeqLM.from_pretrained(checkpoint, device_map="auto", torch_dtype=torch.float32)
-llm_pipeline = HuggingFacePipeline(pipeline=pipeline("text2text-generation", model=base_model, tokenizer=tokenizer))
+base_model = AutoModelForSeq2SeqLM.from_pretrained(checkpoint, device_map=torch.device("cpu"), torch_dtype=torch.float32)
+
+# Define functions
+def data_ingestion(file_path):
+    loader = PDFMinerLoader(file_path)
+    documents = loader.load()
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=500)
+    texts = text_splitter.split_documents(documents)
+    embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
+    db = Chroma.from_documents(texts, embeddings, persist_directory=CHROMA_SETTINGS["persist_directory"])
+    db.persist()
+    print(texts)
+    return db
 
-def process_answer(instruction):
+def llm_pipeline():
+    pipe = pipeline(
+        "text2text-generation",
+        model=base_model,
+        tokenizer=tokenizer,
+        max_length=256,
+        do_sample=True,
+        temperature=0.3,
+        top_p=0.95
+    )
+    local_llm = HuggingFacePipeline(pipeline=pipe)
+    return local_llm
+
+def qa_llm():
+    llm = llm_pipeline()
     embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
-    vectordb = Chroma(persist_directory=CHROMA_SETTINGS.persist_directory, embedding_function=embeddings)
+    vectordb = Chroma(persist_directory=CHROMA_SETTINGS["persist_directory"], embedding_function=embeddings)
     retriever = vectordb.as_retriever()
-    qa = RetrievalQA.from_chain_type(llm=llm_pipeline, chain_type="stuff", retriever=retriever)
-    generated_text = qa(instruction)
-    return generated_text["result"]
+    qa = RetrievalQA.from_chain_type(
+        llm=llm,
+        chain_type="stuff",
+        retriever=retriever,
+        return_source_documents=True
+    )
+    
+    return qa
 
-def chatbot(pdf_file, user_question):
-    if pdf_file:  # Only initialize if a new PDF is uploaded
-        init_db_if_not_exists(pdf_file.name)
-    try:
-        answer = process_answer(user_question)
-        return answer
-    except Exception as e:
-        return f"An error occurred: {str(e)}"
+def process_answer(file, instruction):
+    # Ingest the data from the uploaded PDF
+    data_ingestion(file.name)
+    # Process the question
+    qa = qa_llm()
+    generated_text = qa(instruction)
+    answer = generated_text["result"]
+    return answer
 
-# Create Gradio Interface
+# Define Gradio interface
 iface = gr.Interface(
-    fn=chatbot,
-    inputs=[gr.File(type="binary", label="Upload your PDF"), gr.Textbox(lines=1, label="Ask a Question")],
-    outputs="text",
-    title="PDF Chatbot",
-    description="Upload a PDF and ask questions about its content.",
+    fn=process_answer, 
+    inputs=["file", "text"], 
+    outputs="text"
 )
 
-
-# Run the Gradio interface
-iface.launch()
+# Launch the interface
+iface.launch()