Create app.py

app.py

@@ -0,0 +1,82 @@
+import gradio as gr
+from llama_index.core.readers import SimpleDirectoryReader
+from llama_index.core import VectorStoreIndex, Document
+from llama_index.core.node_parser import SentenceSplitter
+from llama_index.core import Settings
+from llama_index.embeddings.huggingface import HuggingFaceEmbedding
+from llama_index.llms.huggingface import HuggingFaceLLM
+import csv
+from docx import Document as DocxDocument
+import fitz
+
+# Define the list of LLMs with their names and models
+lm_list = {
+    "google/gemma-2-9b-it": "Google Gemma 2.9B IT",
+    "mistralai/Mistral-7B-Instruct-v0.3": "Mistral 7B Instruct v0.3"
+}
+
+# Initialize the query engine globally
+query_engine = None
+
+def process_file(file):
+    file_extension = file.name.split(".")[-1].lower()
+
+    if file_extension == 'txt':
+        with open(file.name, 'r', encoding='utf-8') as f:
+            text = f.read()
+
+    elif file_extension == 'csv':
+        with open(file.name, 'r', encoding='utf-8') as f:
+            reader = csv.reader(f)
+            text = '\n'.join(','.join(row) for row in reader)
+
+    elif file_extension == 'pdf':
+        pdf_document = fitz.open(file.name, filetype=file_extension)
+        text = ""
+        for page_num in range(pdf_document.page_count):
+            page = pdf_document.load_page(page_num)
+            text += page.get_text("text")
+        pdf_document.close()
+
+    elif file_extension == 'docx':
+        docx_document = DocxDocument(file.name)
+        text = ""
+        for paragraph in docx_document.paragraphs:
+            text += paragraph.text + "\n"
+
+    return [Document(text=text)]
+
+def handle_file_upload(file, llm_name):
+    global query_engine
+
+    Settings.llm = HuggingFaceLLM(model_name=llm_name)
+
+    documents = process_file(file)
+
+    text_splitter = SentenceSplitter(chunk_size=512, chunk_overlap=10)
+    Settings.embed_model = HuggingFaceEmbedding(model_name="nomic-embed-text:latest")
+    Settings.text_splitter = text_splitter
+    index = VectorStoreIndex.from_documents(
+        documents, transformations=[text_splitter], embed_model=Settings.embed_model
+    )
+
+    return index.as_query_engine()
+
+def document_qa(file_upload, llm_choice, question_input):
+    query_engine = handle_file_upload(file_upload, llm_choice)
+    result = query_engine.query(question_input)
+    return str(result)
+
+
+llm_choice = gr.Dropdown(choices=list(lm_list.values()), label="Choose LLM")
+file_upload = gr.File(label="Upload Document")
+question_input = gr.Textbox(label="Enter your question")
+
+gr.Interface(
+    fn=document_qa,
+    inputs=[file_upload, llm_choice, question_input],
+    outputs=gr.Textbox(label="Answer"),
+    title="Document Question Answering",
+    description="Upload a document and choose a language model to get answers.",
+    allow_flagging=False
+).launch()