AI Conference Bot

Bot is working, link:  https://t.me/Token2049_Events_bot

- GaiaNet node running any model. For smaller servers it's advised to use the smallest LLM with no RAG data added.

LLM is used to write an SQL query and select data from our event database. After retrieval, the data used to compile templated-answers with features like pagination, add to feature and more.

- Smart aggregation.

We use several crawlers to aggregate data from luma, main conference website and google spreadsheet.

- After data collection, LLM is used to categorize events, event summary completion, and data validation.

- After that, users are notified about new events and featured events with updated info. (If user liked and event and location has changed - the user will receive an update).

- RAG & Larger models

If you have access to a bigger server, the bigger model can be used with data from mySQL ported to a RAG database. By doing so, you can allow users to "chat" with the event DB and recieve non-templated responses.

- Any event

The bot can work with any event and any number of side events. We tested it during KBW and Token2049, aggregating over 400 main event talks & panels with additional 1000 side events.

To make the bot work with a new event, you simply need to parse the google doc with side events and the main conference agenda, which takes around 2 hours on average.