It serves as the LLM backend for Audioscribe—we took care of all the prompt engineering and the 'brain' of the AI agent on Wordware to ensure high-quality outputs. To do so, we experimented with different models and iterated on the prompts 100 times. Additionally, we used a set of IF statements to build logic for the comprehensive summary based on the note type.
It's much faster to do this on Wordware than on any other LLM framework or typical codebase. We allowed non-technical team members to master the prompts while an engineer focused on the code and simply linked an API to Whisper.
