On an evaluation dataset run, executes the AI agent and scores response correctness with OpenAI using the Ragas methodology.
On a chat message, runs an OpenAI copywriting agent that pulls reference material from Google Docs to draft copy.
On a Telegram message, generates a reply with OpenAI and sends it back via Telegram.
On a Telegram message, runs an OpenAI/Gemini agent with calculator and Wikipedia tools, logs to Airtable, and replies via Telegram.
