├── .gitignore ├── Architecture.png ├── Home.py ├── Libs.py ├── README.md ├── demo ├── Chat.mov ├── CreateQuestions.mov ├── QnA.mov ├── Suggest_Writing.mov └── Summary.mov ├── pages ├── 0_Summary_Document.py ├── 1_Question_Anwers_Document.py ├── 2_Create_Multi_Choice.py ├── 3_Improve_Writing.py ├── 4_Rewrite_Essay.py └── 5_Search.py ├── requirements.txt └── samples ├── 2022DemandForecastingBasedMachineLearning.pdf ├── Data-struture.pdf ├── DeepAR.pdf └── Financial_Time_Series_Forecasting_Algorithm_Based_.pdf /.gitignore: -------------------------------------------------------------------------------- 1 | *.pyc 2 | stock_ticker_database.db 3 | *venv -------------------------------------------------------------------------------- /Architecture.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/nguyendinhthi0705/Study-Assistant/fe52ec6778a64bf84be74f042a1c21c677ffed2d/Architecture.png -------------------------------------------------------------------------------- /Home.py: -------------------------------------------------------------------------------- 1 | import streamlit as st 2 | import Libs as glib 3 | import json 4 | 5 | st.set_page_config(page_title="Home") 6 | 7 | 8 | st.markdown("Ask me anything as below samples:") 9 | st.markdown("Top 10 interview questions for OOP in Java") 10 | st.markdown("Write a recursive function.") 11 | st.markdown("Phân biệt giữa classifcation and object detection trong computer vision.") 12 | st.markdown("Thuật toán nào được dùng để xây dựng hệ thống recommendation.") 13 | 14 | input_text = st.text_input("Input your question") 15 | if input_text: 16 | with st.chat_message("user"): 17 | st.markdown(input_text) 18 | response = glib.call_claude_sonet_stream(input_text) 19 | st.write_stream(response) 20 | 21 | 22 | 23 | 24 | 25 | 26 | -------------------------------------------------------------------------------- /Libs.py: -------------------------------------------------------------------------------- 1 | import os 2 | import boto3, json 3 | from dotenv import load_dotenv 4 | from langchain_community.retrievers import AmazonKnowledgeBasesRetriever 5 | from langchain.chains import RetrievalQA 6 | from langchain_community.chat_models import BedrockChat 7 | from langchain_core.prompts import ChatPromptTemplate 8 | from langchain_core.runnables import RunnablePassthrough 9 | from langchain_core.output_parsers import StrOutputParser 10 | 11 | load_dotenv() 12 | 13 | def call_claude_sonet_stream(prompt): 14 | 15 | prompt_config = { 16 | "anthropic_version": "bedrock-2023-05-31", 17 | "max_tokens": 2000, 18 | "temperature": 0, 19 | "top_k": 0, 20 | "messages": [ 21 | { 22 | "role": "user", 23 | "content": [ 24 | {"type": "text", "text": prompt}, 25 | ], 26 | } 27 | ], 28 | } 29 | 30 | body = json.dumps(prompt_config) 31 | 32 | modelId = "anthropic.claude-3-sonnet-20240229-v1:0" 33 | accept = "application/json" 34 | contentType = "application/json" 35 | 36 | bedrock = boto3.client(service_name="bedrock-runtime") 37 | response = bedrock.invoke_model_with_response_stream( 38 | body=body, modelId=modelId, accept=accept, contentType=contentType 39 | ) 40 | 41 | stream = response['body'] 42 | if stream: 43 | for event in stream: 44 | chunk = event.get('chunk') 45 | if chunk: 46 | delta = json.loads(chunk.get('bytes').decode()).get("delta") 47 | if delta: 48 | yield delta.get("text") 49 | 50 | def rewrite_document(input_text): 51 | prompt = """Your name is good writer. You need to rewrite content: 52 | \n\nHuman: here is the content 53 | """ + str(input_text) + """ 54 | \n\nAssistant: """ 55 | return call_claude_sonet_stream(prompt) 56 | 57 | 58 | def summary_stream(input_text): 59 | prompt = f"""Based on the provided context, create summary the lecture 60 | \n\nHuman: here is the content 61 | """ + str(input_text) + """ 62 | \n\nAssistant: """ 63 | return call_claude_sonet_stream(prompt) 64 | 65 | def query_document(question, docs): 66 | prompt = """Human: here is the content: 67 | """ + str(docs) + """ 68 | Question: """ + question + """ 69 | \n\nAssistant: """ 70 | 71 | return call_claude_sonet_stream(prompt) 72 | 73 | def create_questions(input_text): 74 | system_prompt = """You are an expert in creating high-quality multiple-choice quesitons and answer pairs 75 | based on a given context. Based on the given context (e.g a passage, a paragraph, or a set of information), you should: 76 | 1. Come up with thought-provoking multiple-choice questions that assess the reader's understanding of the context. 77 | 2. The questions should be clear and concise. 78 | 3. The answer options should be logical and relevant to the context. 79 | 80 | The multiple-choice questions and answer pairs should be in a bulleted list: 81 | 1) Question: 82 | 83 | A) Option 1 84 | 85 | B) Option 2 86 | 87 | C) Option 3 88 | 89 | Answer: A) Option 1 90 | 91 | 92 | Continue with additional questions and answer pairs as needed. 93 | 94 | MAKE SURE TO INCLUDE THE FULL CORRECT ANSWER AT THE END, NO EXPLANATION NEEDED:""" 95 | 96 | prompt = f"""{system_prompt}. Based on the provided context, create 10 multiple-choice questions and answer pairs 97 | \n\nHuman: here is the content 98 | """ + str(input_text) + """ 99 | \n\nAssistant: """ 100 | return call_claude_sonet_stream(prompt) 101 | 102 | def suggest_writing_document(input_text): 103 | prompt = """Your name is good writer. You need to suggest and correct mistake in the essay: 104 | \n\nHuman: here is the content 105 | """ + str(input_text) + """ 106 | \n\nAssistant: """ 107 | return call_claude_sonet_stream(prompt) 108 | 109 | def search(question, callback): 110 | retriever = AmazonKnowledgeBasesRetriever( 111 | knowledge_base_id="JAHBTIXPHK", 112 | retrieval_config={"vectorSearchConfiguration": {"numberOfResults": 1}}, 113 | 114 | ) 115 | 116 | model_kwargs_claude = {"max_tokens": 1000} 117 | llm = BedrockChat(model_id="anthropic.claude-3-sonnet-20240229-v1:0" 118 | , model_kwargs=model_kwargs_claude 119 | , streaming=True 120 | , callbacks=[callback]) 121 | 122 | chain = RetrievalQA.from_chain_type( 123 | llm=llm, retriever=retriever, return_source_documents=True 124 | ) 125 | return chain.invoke(question) -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # Overview Study-Assistant 2 | This is a simple demo of Amazon Bedrock and Anthropic Claude 3 Sonnet model with langchain and streamlit. For more detail please reference the following link:
3 | - https://aws.amazon.com/bedrock/ 4 | - Claude 3 5 | # To view demo and sample data: 6 | Access folder demo for demo video 7 | Access folder samples for sample videos 8 | 9 | # To Setup 10 | Setup Python
11 | Setup Python Env
12 | Setup AWS CLI
13 | > git clone https://github.com/nguyendinhthi0705/Study-Assistant.git
14 | > cd Study-Assistant
15 | > pip3 install -r requirements.txt
16 | > streamlit run Home.py --server.port 8080
17 | 18 | # Architecture 19 | ![Architecture](./Architecture.png) 20 | 21 | # Learn more about prompt and Claude 3 22 | Introduction to prompt design 23 | Model Card 24 | 25 | # Demo 26 | 27 | ## A Simple Chat 28 | [![Chat](http://img.youtube.com/vi/PdX7i0A4a-M/0.jpg)](https://www.youtube.com/watch?v=PdX7i0A4a-M)] 29 | 30 | ## Questions and Anwsers 31 | [![Questions and Anwsers](http://img.youtube.com/vi/ciJfAhyRjTI/0.jpg)](https://www.youtube.com/watch?v=ciJfAhyRjTI)] 32 | 33 | ## Summary a Lecture 34 | [![Summary](http://img.youtube.com/vi/5JpeWmbHMi0/0.jpg)](https://www.youtube.com/watch?v=5JpeWmbHMi0)] 35 | 36 | ## Create Multi Choice Questions 37 | [![Create Multi Choice Questions](http://img.youtube.com/vi/AE9gj19a9t0/0.jpg)](https://www.youtube.com/watch?v=AE9gj19a9t0)] 38 | 39 | ## Suggest a Better Writing 40 | [![Suggest Better Writing](http://img.youtube.com/vi/7xBR5rtcp30/0.jpg)](https://www.youtube.com/watch?v=7xBR5rtcp30)] 41 | 42 | 43 | -------------------------------------------------------------------------------- /demo/Chat.mov: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/nguyendinhthi0705/Study-Assistant/fe52ec6778a64bf84be74f042a1c21c677ffed2d/demo/Chat.mov -------------------------------------------------------------------------------- /demo/CreateQuestions.mov: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/nguyendinhthi0705/Study-Assistant/fe52ec6778a64bf84be74f042a1c21c677ffed2d/demo/CreateQuestions.mov -------------------------------------------------------------------------------- /demo/QnA.mov: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/nguyendinhthi0705/Study-Assistant/fe52ec6778a64bf84be74f042a1c21c677ffed2d/demo/QnA.mov -------------------------------------------------------------------------------- /demo/Suggest_Writing.mov: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/nguyendinhthi0705/Study-Assistant/fe52ec6778a64bf84be74f042a1c21c677ffed2d/demo/Suggest_Writing.mov -------------------------------------------------------------------------------- /demo/Summary.mov: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/nguyendinhthi0705/Study-Assistant/fe52ec6778a64bf84be74f042a1c21c677ffed2d/demo/Summary.mov -------------------------------------------------------------------------------- /pages/0_Summary_Document.py: -------------------------------------------------------------------------------- 1 | import streamlit as st 2 | import Libs as glib 3 | from PyPDF2 import PdfReader 4 | import Libs as glib 5 | 6 | st.set_page_config(page_title="Summary a Lecture/Paper") 7 | 8 | uploaded_file = st.file_uploader("Upload Lecture/Paper in PDF") 9 | docs = [] 10 | 11 | if uploaded_file is not None: 12 | reader = PdfReader(uploaded_file) 13 | for page in reader.pages: 14 | docs.append(page.extract_text()) 15 | 16 | response = glib.summary_stream(docs) 17 | st.write_stream(response) 18 | -------------------------------------------------------------------------------- /pages/1_Question_Anwers_Document.py: -------------------------------------------------------------------------------- 1 | import streamlit as st 2 | import Libs as glib 3 | from PyPDF2 import PdfReader 4 | import Libs as glib 5 | 6 | 7 | st.set_page_config(page_title="Questions on Lecture") 8 | 9 | uploaded_file = st.file_uploader("Upload Your Lecture in PDF") 10 | docs = [] 11 | 12 | st.markdown("Ask me anything as below samples:") 13 | st.markdown("Summary the lecture") 14 | 15 | input_text = st.text_input("Your question!") 16 | if uploaded_file is not None and input_text: 17 | reader = PdfReader(uploaded_file) 18 | for page in reader.pages: 19 | docs.append(page.extract_text()) 20 | 21 | response = glib.query_document(input_text, docs) 22 | st.write_stream(response) 23 | 24 | 25 | -------------------------------------------------------------------------------- /pages/2_Create_Multi_Choice.py: -------------------------------------------------------------------------------- 1 | import streamlit as st 2 | import Libs as glib 3 | from PyPDF2 import PdfReader 4 | import Libs as glib 5 | 6 | st.set_page_config(page_title="Create question for a Lecture/Paper") 7 | 8 | st.markdown("Upload your lecture to create multiple choice questions!") 9 | uploaded_file = st.file_uploader("Upload Lecture/Paper in PDF") 10 | docs = [] 11 | 12 | if uploaded_file is not None: 13 | reader = PdfReader(uploaded_file) 14 | for page in reader.pages: 15 | docs.append(page.extract_text()) 16 | 17 | response = glib.create_questions(docs) 18 | st.write_stream(response) 19 | -------------------------------------------------------------------------------- /pages/3_Improve_Writing.py: -------------------------------------------------------------------------------- 1 | import streamlit as st 2 | import Libs as glib 3 | 4 | st.set_page_config(page_title="Improve writing an essay") 5 | 6 | 7 | input_text = st.text_area("Input your whole or apart of your essay") 8 | if input_text: 9 | response = glib.suggest_writing_document(input_text) 10 | st.write_stream(response) 11 | 12 | -------------------------------------------------------------------------------- /pages/4_Rewrite_Essay.py: -------------------------------------------------------------------------------- 1 | import streamlit as st 2 | import Libs as glib 3 | 4 | st.set_page_config(page_title="Rewrite an Essay") 5 | 6 | 7 | input_text = st.text_area("Input your whole or apart of your essay") 8 | if input_text: 9 | response = glib.rewrite_document(input_text) 10 | st.write_stream(response) 11 | 12 | -------------------------------------------------------------------------------- /pages/5_Search.py: -------------------------------------------------------------------------------- 1 | import streamlit as st 2 | import Libs as glib 3 | from langchain.callbacks import StreamlitCallbackHandler 4 | 5 | st.set_page_config(page_title="Search Knowledge base") 6 | st.markdown("Một số bệnh phổ biến của trẻ em là gì?") 7 | st.markdown("Tóm tắt tài chính apple?") 8 | 9 | input_text = st.text_input("Search Knowledge base") 10 | if input_text: 11 | st_callback = StreamlitCallbackHandler(st.container()) 12 | response = glib.search(input_text, st_callback) 13 | st.write(response["result"]) 14 | st.write(response) 15 | 16 | -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | transformers 2 | Pillow 3 | pypdf 4 | jq 5 | langchain 6 | streamlit 7 | boto3 8 | awscli 9 | pandas-datareader 10 | langchain-experimental 11 | bs4 12 | setuptools 13 | PyPDF2 14 | pypdf 15 | langchain_community 16 | python-dotenv 17 | botocore -------------------------------------------------------------------------------- /samples/2022DemandForecastingBasedMachineLearning.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/nguyendinhthi0705/Study-Assistant/fe52ec6778a64bf84be74f042a1c21c677ffed2d/samples/2022DemandForecastingBasedMachineLearning.pdf -------------------------------------------------------------------------------- /samples/Data-struture.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/nguyendinhthi0705/Study-Assistant/fe52ec6778a64bf84be74f042a1c21c677ffed2d/samples/Data-struture.pdf -------------------------------------------------------------------------------- /samples/DeepAR.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/nguyendinhthi0705/Study-Assistant/fe52ec6778a64bf84be74f042a1c21c677ffed2d/samples/DeepAR.pdf -------------------------------------------------------------------------------- /samples/Financial_Time_Series_Forecasting_Algorithm_Based_.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/nguyendinhthi0705/Study-Assistant/fe52ec6778a64bf84be74f042a1c21c677ffed2d/samples/Financial_Time_Series_Forecasting_Algorithm_Based_.pdf --------------------------------------------------------------------------------