muni_bot

Inspiration

• recent interaction with Zoning Board of Appeals in Cato Township while hearing a request for variance.

What it does

• Provides an interface to allow residents and board members to access upto date information about the current zoning requirements for that locale. Additionally it can rationalize about wether a variance meets requirements to assist in removing bias when granting/rejecting variances or regular zoning permits. ## How we built it embedding:
• utilized langchain for extracting chunks from the pdf's for embedding
• openai for embeddings
• pinecone for storing embeddings in an index.

chat botting:

• utilized a gradio interface for the chat experience
• utilized a Langchain ConversationalRetrievalChain for indexing the user query against the indexed zoning ordinance pdf.
• utilized openai's gpt-4 model for this experiment ## Challenges we ran into
• pdf format chunking for embedding is not always easy

Accomplishments that we're proud of

What we learned

• GPT-4 based bots have the potential to remove a lot of bias from common interactions.

What's next for muni_bot

• deploy logic to an AWS Lambda, deploy UI and connect to lambda for message processing, add history to the chat, add option for utilizing pdf url links to chat across different set of zoning ordinances/implement a multi vector chain to auto select which zoning index to query against.

Built With

• gradio
• langchain
• openai
• pinecone