Llama-3 70b OMNI-complete - AUTO Improving AUTOcomplete Prompt for EVERYTHING (Groq)

---

---

AI Nuggets

CLI Commands, URLs, and Tips from Video Transcript

CLI Commands

• yarn test - Runs prompt validation tests using prompFu.

URLs

• Uno CSS (Atomic CSS engine) - No specific URL provided, but you can search for it online.
• Gro (LLM provider) - No specific URL provided, but it’s mentioned as being used with a llama 70b model.
• Vue.js - No specific URL provided, but it’s a JavaScript framework that can be found at https://vuejs.org/.
• Tailwind CSS - No specific URL provided, but it’s a utility-first CSS framework that can be found at https://tailwindcss.com/.
• WindiCSS - No specific URL provided, but it’s a utility-first CSS framework similar to Tailwind CSS.
• BAP video - The URL is not provided in the transcript, but it’s mentioned to be linked in the video description.
• Prompt Fu videos - The URL is not provided in the transcript, but it’s mentioned to be linked in the video description.
• Code for the tool demonstrated - The URL is not provided in the transcript, but it’s mentioned to be linked in the video description.

Tips

• LLM autocompletes can self-improve with every single completion.
• The code base shared in the video has a simplistic yet effective way to reinforce the most popular autocomplete.
• LLM autocompletes are easily reusable across tools and applications.
• The prompt shared in the video can be used across different domains with minor tweaks.
• LLM autocompletes reveal actionable information about your prospects, users, and customers.
• The architecture shared in the video is a client-server architecture with a prompt-centered design.
• The prompt is the most important piece of the application.
• The prompt uses markdown syntax with H1 and H2 headers to structure the information.
• The prompt contains generation rules and variables for the topic, previous completions, domain knowledge, and the desired completion.
• The previous completions JSON file acts as a database that is self-improving.
• The architecture integrates with Prompt Fu for prompt testing and validation.
• The server looks directly at the test file for prompt validation.
• The frontend framework used (Vue.js, React, Svelte, etc.) is considered irrelevant as long as it provides value to users efficiently.
• The Python Flask server serves up two routes: use autocomplete and get autocomplete.
• The domain knowledge is a plain text file that grows over time and may eventually become a BAP (Big Ass Prompt).
• The video emphasizes the potential of LLMs to enhance existing workflows and development processes with self-improving, prompt-based software.

(Note: The above information is extracted from the transcript provided and does not contain additional details that may be present in the video description or the video itself.)