video1.0<iframe src="https://www.loom.com/embed/e0f3ae24367841b58b31b2d53a222e5d" frameborder="0" width="1658" height="1243" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>12431658Loomhttps://www.loom.com12431658https://cdn.loom.com/sessions/thumbnails/e0f3ae24367841b58b31b2d53a222e5d-87297abb995b6cc5.gif286.684Hi, this is my submission for the Cosmo Take-Home Engineering assignment, showcasing a PipeCat implementation. I'm utilizing Grok LLAMA 3.18b for low-latency LLM inference and DeepGram for both text-to-speech and speech-to-text, with a SQLite database running locally. During the demo, I’ll be testing order updates and refund policy inquiries while streaming latency metrics in real-time. I also have retry logic in place for handling rate limits, and I'm randomizing inputs to demonstrate the system's capabilities. Please take a look at the K8 image specs in the repo for deployment details and let me know your thoughts!