Welcome to the Low Latency on LLMs and Agent Workflows session of the DevFest AI Workshop! This session will guide you through agentic workflows using Groq.
Instructor: Jose Menendez
Duration: 20min.
Objective:
- Get an agentic workflow running with low latency LLMs in minutes.
- Orchestrate another workflow to formulate pytho code and execute it.
By the end of this session, you will have a deeper understanding of
- Groq, LPU, and Low Latency LLMs.
- Agentic workflows.
- How to leverage crewAI to build them without the hassle.
- Basic knowledge of LLM prompting
- Basic knowledge of Python
- Basic understanding of YAML
-
Introduction
- Overview of Groq, LPU, and Low Latency LLMs, open source models.
-
Hands-on Activity
- Create an out-of-the-box workflow with crewAI.
- Make small adjustments to implement a coder workflow.
-
Q&A and Discussion
- Open floor for questions / ideas / discussions.
It's a true quickstart, so you can just follow the instructions.
The examples in this folder are:
researchers: Base workflow out of the box.xmas-tree: A modified version where a couple of agents generate a X-Max tree ASCII art.
Following the docs of CrewAI, use configurations to build a workflows that:
- Exercise 1: A workflow that takes a URL to read the content and generate a markdown summary of the content. For this you have to implement the
Scrape Websitetool. - Exercise 2: Take it up a notch. Modify the basic researcher workflow to use web search and scraping in order to do an online investigation.
- Exercise 3: Create a custom tool, that can take an audio file, send it to Groq to transcribe it and then summarize the content.
- Documentation:
If you have questions during the workshop, please reach out to Jose Menendez or open an issue in the repository with label 6_groq.
Happy coding!