Upload 4 files

app.py

@@ -0,0 +1,165 @@
+import gradio as gr
+import logging
+
+from config import SHARE_GRADIO_WITH_PUBLIC_URL
+from chains import qa_chain, summarization_chain
+
+logger = logging.getLogger(__name__)
+
+# Translation dictionary
+TRANSLATIONS = {
+    "en": {
+        "title": "# 📚 Study Buddy: AI Learning Assistant",
+        "subtitle": "## 🤖 A smart, user-friendly chatbot for students!",
+        "summary_subtitle": "## 📄 Upload Notes for Summarization",
+        "chat_input_label": "Type your question here:",
+        "chat_placeholder": "e.g., Explain Newton's laws",
+        "chat_button_label": "Get Answer",
+        "summary_button_label": "Summarize Notes",
+        "upload_file_label": "Upload .txt or .pdf file",
+        "summary_output_label": "Summary",
+        "language_label": "Language / Langue",
+        "ai_response_label": "AI Response"
+    },
+    "fr": {
+        "title": "# 📚 Study Buddy: Assistant d'apprentissage IA",
+        "subtitle": "## 🤖 Un chatbot intelligent et convivial pour les étudiants!",
+        "summary_subtitle": "## 📄 Subir notas para resumir",
+        "chat_input_label": "Tapez votre question ici:",
+        "chat_placeholder": "ex : Expliquez les lois de Newton",
+        "chat_button_label": "Obtenir une réponse",
+        "summary_button_label": "Résumer les notes",
+        "upload_file_label": "Téléchargez un fichier .txt ou .pdf",
+        "summary_output_label": "Résumé",
+        "language_label": "Langue / Language",
+        "ai_response_label": "Réponse de l'IA"
+    }
+}
+
+# Function to process user queries
+def chatbot_response(user_input, lang):
+    try:
+        response = qa_chain.invoke({"question": user_input})
+        logger.info("chatbot_response completed")
+        return response
+    except Exception as e:
+        msg = f"Error : {e}"
+        logger.exception(msg)
+        return TRANSLATIONS[lang].get("error_message", "Sorry, an error occurred while processing your request.")
+
+# Function to summarize notes
+def summarize_text(text, lang):
+    try:
+        summary = summarization_chain.invoke({"document_text": text})
+        logger.info("summarize_text completed")
+        return summary
+    except Exception as e:
+        msg = f"Error : {e}"
+        logger.exception(msg)
+        return TRANSLATIONS[lang].get("error_message", "Sorry, an error occurred while summarizing your notes.")
+
+# Function to update UI labels dynamically
+def update_language(lang):
+    return (
+        TRANSLATIONS[lang]["title"],
+        TRANSLATIONS[lang]["subtitle"],
+        TRANSLATIONS[lang]["chat_input_label"],
+        TRANSLATIONS[lang]["chat_placeholder"],
+        TRANSLATIONS[lang]["chat_button_label"],
+        TRANSLATIONS[lang]["upload_file_label"],
+        TRANSLATIONS[lang]["summary_button_label"],
+        TRANSLATIONS[lang]["summary_output_label"],
+        TRANSLATIONS[lang]["ai_response_label"]
+    )
+
+# Gradio UI
+def create_interface():
+    with gr.Blocks(css="body { font-family: sans-serif; background-color: #f9f9f9; }") as study_buddy:
+        
+        # Default to English
+        lang = "en"
+
+        title = gr.Markdown(f"{TRANSLATIONS[lang]['title']}")
+        
+        with gr.Row():
+            with gr.Column():
+                gr.Markdown("", height=4) 
+                
+                language = gr.Radio(
+                    choices=["en", "fr"],
+                    value=lang,
+                    label=TRANSLATIONS[lang]["language_label"]
+                )
+                
+                gr.Markdown("", height=4)
+                               
+                subtitle = gr.Markdown(f"{TRANSLATIONS[lang]['subtitle']}")
+                
+                chat_input = gr.Textbox(
+                    label=TRANSLATIONS[lang]["chat_input_label"],
+                    lines=4,
+                    placeholder=TRANSLATIONS[lang]["chat_placeholder"]
+                )
+
+            with gr.Column():
+                gr.Markdown("", height=4)
+                summary_subtitle = gr.Markdown(f"{TRANSLATIONS[lang]['summary_subtitle']}")
+                file_input = gr.File(file_types=[".pdf", ".txt"])
+                file_label = gr.Markdown(TRANSLATIONS[lang]["upload_file_label"])  # Separate label
+                
+        with gr.Row():
+            with gr.Column():
+                
+                chat_button = gr.Button(TRANSLATIONS[lang]["chat_button_label"], variant="primary")
+                chat_output = gr.Textbox(label=TRANSLATIONS[lang]["ai_response_label"], lines=5, interactive=True)
+
+                # Bind chatbot response function
+                chat_button.click(
+                    chatbot_response, 
+                    inputs=[chat_input, language], 
+                    outputs=chat_output
+                )
+
+            with gr.Column():
+            
+                summary_button = gr.Button(TRANSLATIONS[lang]["summary_button_label"], variant="primary")
+                summary_output = gr.Textbox(label=TRANSLATIONS[lang]["summary_output_label"], lines=5, interactive=True)
+
+                # Bind summarization function
+                summary_button.click(
+                    summarize_text, 
+                    inputs=[file_input, language], 
+                    outputs=summary_output
+                )
+
+        # Update labels dynamically when the language changes
+        def update_labels(lang):
+            return (
+                TRANSLATIONS[lang]["title"],
+                TRANSLATIONS[lang]["subtitle"],
+                TRANSLATIONS[lang]["summary_subtitle"],
+                TRANSLATIONS[lang]["chat_input_label"],
+                TRANSLATIONS[lang]["chat_placeholder"],
+                TRANSLATIONS[lang]["chat_button_label"],
+                TRANSLATIONS[lang]["summary_button_label"],
+                TRANSLATIONS[lang]["summary_output_label"],
+                TRANSLATIONS[lang]["ai_response_label"],
+                TRANSLATIONS[lang]["upload_file_label"]
+            )
+
+        language.change(
+            update_labels, 
+            inputs=[language], 
+            outputs=[
+                title, subtitle, summary_subtitle, 
+                chat_input, chat_input, 
+                chat_button, 
+                summary_button, summary_output, chat_output, file_label  # Update file label separately
+            ]
+        )
+
+    return study_buddy
+
+if __name__ == "__main__":
+    study_buddy = create_interface()
+    study_buddy.launch(share=SHARE_GRADIO_WITH_PUBLIC_URL)

chains.py

@@ -0,0 +1,47 @@
+import logging
+
+from llms.local import llm
+# from llms.openai import llm
+from prompts import QA_PROMPT
+from prompts import SUMMARIZATION_PROMPT
+
+logger = logging.getLogger(__name__)
+
+def get_qa_chain():
+    logger.info("Creating QA chain")
+    print("> Creating QA chain")
+    
+    try:
+        # qa_chain = QA_PROMPT
+        qa_chain = QA_PROMPT | llm
+        
+        logger.info("QA chain created")
+        print("> QA chain created")
+        return qa_chain
+    
+    except Exception as e:
+        msg = f"Error: {e}"
+        logger.exception(msg)
+        print(msg)
+        raise e
+
+def get_summarization_chain():
+    logger.info("Creating summarization chain")
+    print("> Creating summarization chain")
+    
+    try:
+        # summarization_chain = SUMMARIZATION_PROMPT
+        summarization_chain = SUMMARIZATION_PROMPT | llm
+        
+        logger.info("Summarization chain created")
+        print("> Summarization chain created")
+        return summarization_chain
+    
+    except Exception as e:
+        msg = f"Error: {e}"
+        logger.exception(msg)
+        print(msg)
+        raise e
+
+qa_chain = get_qa_chain()
+summarization_chain = get_summarization_chain()

local.py

@@ -0,0 +1,70 @@
+from dotenv import load_dotenv
+import logging
+
+import torch
+from transformers import (
+    AutoModelForCausalLM,
+    AutoTokenizer,
+    BitsAndBytesConfig,
+    AutoTokenizer,
+    BitsAndBytesConfig,
+    pipeline,
+)
+
+from langchain_huggingface import HuggingFacePipeline
+from langchain.globals import set_debug
+from langchain.globals import set_verbose
+
+from config import HF_MODEL_ID
+from config import LLM_VERBOSE
+
+set_verbose(LLM_VERBOSE)
+set_debug(LLM_VERBOSE)
+
+logger = logging.getLogger(__name__)
+load_dotenv()
+
+cuda_check = torch.cuda.is_available()
+logger.info(f"torch.cuda.is_available : {cuda_check}")
+print(f"> torch.cuda.is_available : {cuda_check}")
+
+# Load Llama3 model and tokenizer
+model_id = HF_MODEL_ID
+
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+
+# BitsAndBytesConfig int-4 config
+# device_map = {"": 0}
+device_map = "auto"
+compute_dtype = getattr(torch, "float16")
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_use_double_quant=False,
+    bnb_4bit_quant_type="nf4",
+    # bnb_4bit_compute_dtype=torch.bfloat16
+    bnb_4bit_compute_dtype=compute_dtype,
+    # bnb_4bit_use_double_quant=False,
+)
+
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    device_map=device_map,
+    # attn_implementation="flash_attention_2",
+    quantization_config=bnb_config,
+)
+
+model.generation_config.pad_token_id = tokenizer.eos_token_id
+
+pipe = pipeline(
+    "text-generation", 
+    model=model, 
+    tokenizer=tokenizer, 
+    max_new_tokens=50,
+    return_full_text=False,
+    num_return_sequences=1,
+    eos_token_id=tokenizer.eos_token_id,
+    temperature=0.0001,
+    do_sample=True,
+)
+
+llm = HuggingFacePipeline(pipeline=pipe)

prompts.py

@@ -0,0 +1,53 @@
+from langchain.prompts import PromptTemplate
+
+qa_template = (
+"""
+<|start_header_id|>system<|end_header_id|>
+You are an AI assistant specialized in answering academic queries with accuracy and clarity.
+
+### Instructions
+- Read the user's question carefully.
+- Provide a concise and informative answer based on available knowledge.
+- If the question is unclear, ask for clarification.
+- If you do not know the answer, respond with "I'm not sure."
+
+### User Question
+{question}
+
+<|eot_id|><|start_header_id|>user<|end_header_id|>
+
+Question: Please provide a detailed answer.
+<|eot_id|><|start_header_id|>assistant<|end_header_id|>
+Answer:
+""")
+
+summarization_template = (
+"""
+<|start_header_id|>system<|end_header_id|>
+You are a highly skilled AI that specializes in summarizing academic content.
+
+### Instructions
+- Read the provided text carefully.
+- Extract the most important key points while maintaining clarity.
+- Summarize in a way that retains the core meaning but is concise.
+- Output only the summary without additional explanations.
+
+### Text to Summarize
+{document_text}
+
+<|eot_id|><|start_header_id|>user<|end_header_id|>
+
+Task: Provide a concise summary.
+<|eot_id|><|start_header_id|>assistant<|end_header_id|>
+Summary:
+""")
+
+QA_PROMPT = PromptTemplate(
+    input_variables=["question"],
+    template=qa_template
+)
+
+SUMMARIZATION_PROMPT = PromptTemplate(
+    input_variables=["document_text"],
+    template=summarization_template
+)