from omegaconf import OmegaConf import streamlit as st import os from PIL import Image import re from translate import Translator from pydantic import Field import sys from llama_index.indices.managed.vectara import VectaraIndex from llama_index.core.agent import ReActAgent from llama_index.llms.openai import OpenAI from llama_index.core.tools import QueryEngineTool, ToolMetadata from llama_index.core.utils import print_text from llama_index.core.agent.react.formatter import ReActChatFormatter from llama_index.core.tools import FunctionTool from prompts import prompt_template learning_styles = ['traditional', 'Inquiry-based', 'Socratic'] languages = {'English': 'en', 'Spanish': 'es', 'French': 'fr', 'German': 'de', 'Arabic': 'ar', 'Chinese': 'zh-cn', 'Hebrew': 'he', 'Hindi': 'hi', 'Italian': 'it', 'Japanese': 'ja', 'Korean': 'ko', 'Portuguese': 'pt'} initial_prompt = "How can I help you today?" def launch_bot(): def reset(): cfg = st.session_state.cfg llm = OpenAI(model="gpt-4o", temperature=0) tr_prompt = Translator(to_lang=languages[cfg.language]).translate(initial_prompt) st.session_state.messages = [{"role": "assistant", "content": tr_prompt, "avatar": "🦖"}] st.session_state.thinking_prompt = Translator(to_lang=languages[cfg.language]).translate("Thinking...") vectara = VectaraIndex(vectara_api_key=cfg.api_key, vectara_customer_id=cfg.customer_id, vectara_corpus_id=cfg.corpus_id) # Create tool to adapt output to style, age and language def adjust_response_to_student( text: str = Field(descrition='the original text'), age: int = Field(description='the age of the student. An integer'), style: str = Field(description='teaching style'), language: str = Field(description='the language') ) -> str: """ Rephrase the text to match the student's age, desired teaching style and language """ llm = OpenAI(model="gpt-4o", temperature=0) print(f"DEBUG: Adjusting response to student age {age}, style {style} and language {language}") prompt = f''' The following is response the teacher is planning to provide to a student based on their question. Please adjust the response to match the student's age of {age}, the {style} teaching style. For example, in the inquiry-based teaching style, choose to ask questions that encourage the student to think critically instead of repsonding directly with the answer. Or in the socratic teaching style, choose to ask questions that lead the student to the answer. Always respond in the {language} language. original response: {text} adjusted response: ''' response = llm.complete(prompt) return response # Create the Vectara Tool vectara_tool = QueryEngineTool( query_engine = vectara.as_query_engine(summary_enabled = True, summary_num_results = 10, summary_response_lang = languages[cfg.language], summary_prompt_name = "vectara-summary-ext-24-05-large", vectara_query_mode = "mmr", rerank_k = 50, mmr_diversity_bias = 0.1, n_sentence_before = 5, n_sentence_after = 5), metadata = ToolMetadata(name="vectara", description=""" A tool that is able to answer questions about the justice, morality, politics and related topics. Based on transcripts of recordings from the Justice Harvard class that includes a lot of content on these topics. When using the tool it's best to ask simple short questions. """), ) rephrase_tool = FunctionTool.from_defaults(adjust_response_to_student) # Create the agent prompt = prompt_template.replace("{style}", cfg.style) \ .replace("{language}", cfg.language) \ .replace("{student_age}", str(cfg.student_age)) st.session_state.agent = ReActAgent.from_tools( tools=[vectara_tool, rephrase_tool], llm=llm, verbose=True, react_chat_formatter = ReActChatFormatter(system_header=prompt) ) if 'cfg' not in st.session_state: cfg = OmegaConf.create({ 'customer_id': str(os.environ['VECTARA_CUSTOMER_ID']), 'corpus_id': str(os.environ['VECTARA_CORPUS_ID']), 'api_key': str(os.environ['VECTARA_API_KEY']), 'style': learning_styles[0], 'language': 'English', 'student_age': 18 }) st.session_state.cfg = cfg st.session_state.style = cfg.style st.session_state.language = cfg.language st.session_state.student_age = cfg.student_age reset() cfg = st.session_state.cfg st.set_page_config(page_title="Teaching Assistant", layout="wide") # left side content with st.sidebar: image = Image.open('Vectara-logo.png') st.image(image, width=250) st.markdown("## Welcome to the Justice Harvard e-learning assistant demo.\n\n\n") st.markdown("\n") cfg.style = st.selectbox('Learning Style:', learning_styles) if st.session_state.style != cfg.style: st.session_state.style = cfg.style reset() st.markdown("\n") cfg.language = st.selectbox('Language:', languages.keys()) if st.session_state.language != cfg.language: st.session_state.langage = cfg.language reset() st.markdown("\n") cfg.student_age = st.number_input( 'Student age:', min_value=13, value=cfg.student_age, step=1, format='%i' ) if st.session_state.student_age != cfg.student_age: st.session_state.student_age = cfg.student_age reset() st.markdown("\n\n") if st.button('Start Over'): reset() st.markdown("---") st.markdown( "## How this works?\n" "This app was built with [Vectara](https://vectara.com).\n\n" "It demonstrates the use of Agentic Chat functionality with Vectara" ) st.markdown("---") if "messages" not in st.session_state.keys(): reset() # Display chat messages for message in st.session_state.messages: with st.chat_message(message["role"], avatar=message["avatar"]): st.write(message["content"]) # User-provided prompt if prompt := st.chat_input(): st.session_state.messages.append({"role": "user", "content": prompt, "avatar": '🧑‍💻'}) with st.chat_message("user", avatar='🧑‍💻'): print_text(f"Starting new question: {prompt}\n", color='green') st.write(prompt) # Generate a new response if last message is not from assistant if st.session_state.messages[-1]["role"] != "assistant": with st.chat_message("assistant", avatar='🤖'): with st.spinner(st.session_state.thinking_prompt): res = st.session_state.agent.chat(prompt) cleaned = re.sub(r'\[\d+\]', '', res.response) st.write(cleaned) message = {"role": "assistant", "content": cleaned, "avatar": '🤖'} st.session_state.messages.append(message) sys.stdout.flush() if __name__ == "__main__": launch_bot()