A Q&A bot for Procedures at the Italian Consulate in Mendoza

• 

  Q&A Bot answering to questions with the new data given

Inspiration

Many people seek for guidance when trying to understand how to go about some procedures at Italian Consulate in Mendoza. Even if there is the offcial information in their website, people get lost and prefer to ask a specific question. This friendly bot can guide them in that process.

What it does

It ingests the data from some of the pages of the and embeds them in an OpenAi structure. So we have a chatbot up to date with the consulate website and other related documentation.

How we built it

I used Langchain, OpenAi, Chromadb, Tiktoken, and a Google Colab notebook.

Challenges we ran into

Understanding how LLM models work. Working with different languages (usually code snipets are in English, some pages of the website in Italian and most questions are in Spanish, given that is the official language in Argentina) Getting the libraries to be installed.

Accomplishments that we're proud of

Getting a working model! Now I have a template, and can keep on adding new features to the existing model.

What we learned

How LLM models customization, OpenAI and Langchain work.

What's next for A Q&A bot for Procedures at the Italian Consulate in Mendoza

Setting up the integration with a telegram that can provide valuable data to improve the database. Setting up a Telegram Bot, for people to be able to ask questions. Setting it online with Streamlit.

Built With

• chromadb
• llm
• openai
• python