Differences

This shows you the differences between two versions of the page.

--- lctgmeetingsummary20260114 [2026.01.21 09:13] – created Steve Isenberg
+++ lctgmeetingsummary20260114 [2026.01.21 09:47] (current) – Steve Isenberg
@@ Line 1: / Line 1: @@
 ~~NOCACHE~~ <fc #a0a0a0><fs small>//This page last changed ~~LASTMOD~~ visits: {{counter|today| time| times}} today, {{counter|yesterday| time| times}} yesterday, and {{counter|total| time total so far| total times}}//</fs></fc>
-=====Meeting Summary for Lex Computer Group's January 7, 2026  meeting=====
+=====Meeting Summary for Lex Computer Group's January 14, 2026  meeting=====
-<fs xx-large>Potpourri - Email Encryption and Buying Refurbished</fs>
+<fs xx-large>The AI ChatBot Created by LHS Students</fs>
 ====Quick recap====
- blah
+Jerry Xu presented a detailed overview of a chatbot system developed by students from the Youth STEAM Initiative to provide information about Lexington's municipal budget and the new Lexington High School building project. The technical implementation uses a combination of Retrieval Augmented Generation and Reasoning and Acting frameworks, with the system being built on Streamlit frontend and LangChain backend, including features for document retrieval and citation generation. He discussed their testing approach and future plans for improving the system, while addressing concerns about data accuracy and the need for opposing viewpoints in the information provided.
 ====Summary====
+===Lexington Budget Chatbot Development===
+Jerry Xu, a Lexington High School senior, presented a detailed overview of the development process and technical implementation of a chatbot designed to answer questions about Lexington's municipal budget, particularly focusing on the new Lexington High School project. The chatbot uses a framework called GRASP, which combines Retrieval Augmented Generation (RAG) and Reasoning and Acting (React) components. Jerry explained how the RAG system processes user queries by converting textual documents into numerical vector embeddings, allowing the chatbot to retrieve relevant information from external documents. The React framework enables the chatbot to process queries in a more complex manner through a three-step process of thought, action, and observation. The chatbot was developed by nine students from the Youth Steam Initiative and was released in October 2025, providing residents with information before the special tax solution vote on December 8th.
+===Chatbot System Technical Implementation===
+Jerry discussed the technical implementation of their chatbot system, which uses Streamlit for the frontend and LangChain for the backend. They explained how they store user queries and chat history in Streamlit's session state, and how they use LangChain's tools and agent executor to process user inputs. The system is highly customizable through configuration variables, allowing for easy adaptation to different domains. They also described their database setup using Amazon DynamoDB to log user questions and chat box responses anonymously.
+===LangChain Chatbot Technical Implementation===
+He focused on the technical implementation of a chatbot system using LangChain, which includes tools for document retrieval and citation generation. The system processes user queries by embedding them, performing similarity searches on a document collection, and generating responses based on the most relevant information. The chatbot is configured with specific system prompts to maintain neutrality and stay on topic, particularly for the Lexington High School project. The citation feature was highlighted as a unique capability, allowing users to be directed to specific pages within documents rather than just to the document itself.
+===Chatbot Implementation for Municipal Information===
+Jerry discussed the technical implementation of a chatbot system designed to provide accurate and trustworthy information, particularly focusing on municipal budgeting and a new high school project. He explained how the system chunks PDFs by page, generates citation URLs, and provides follow-up questions to users. He detailed their testing approach, which involved multiple groups including town meeting members and residents, and outlined efforts to minimize AI concerns through human review and data transparency. Future plans include improving the UI framework, supporting additional document formats beyond PDFs, and developing an automated testing framework using an external LLM (Large Language Model) for objective evaluation of chatbot responses.
+===Lexington High School Chatbot Project===
+Jerry presented his team's work on developing a chatbot to provide information about the Lexington High School project, which was completed in October. The team focused on using verified, official documents and data from the school building committee, though some concerns were raised about the lack of opposing viewpoints and the complexity of projecting future needs. Jerry explained their approach to using vector embeddings and similarity searches to provide accurate information, and discussed the challenges of keeping the data up-to-date. The team is considering future applications of their framework to other projects, and Jerry may be involved in teaching AI to high school students in the summer.