Skip to content

Latest commit

 

History

History

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

README.md

🚀 LLM Demonstration Workshop Project

Welcome to the README for our workshop project! This repository serves as a practical demonstration of integrating Large Language Models (LLMs) into applications, showcasing both powerful new capabilities and robust development practices.

What This Project Does

This project is designed to illustrate advanced LLM capabilities through automated content generation and seamless deployment. It dynamically produces presentable content, such as real-time news summaries and LLM cost comparisons, making complex concepts easy to visualize and explore. Our goal is to provide a working example of how LLMs can generate valuable, shareable artifacts.

Key Features & Capabilities

  • LLM-Driven News Summarization: Leveraging Google's Gemini models (gemini-1.5-flash) and newsapi, the project fetches real-time articles and summarizes them on demand, demonstrating live data integration with LLMs.
  • Automated Demonstration Generation: Automatically builds and publishes various "slides" or reports, including LLM cost comparisons, feature recaps, and live factuality demonstrations.
  • Robust CI/CD Pipeline: Utilizes GitHub Actions to automate the generation and deployment of these demonstration artifacts directly to GitHub Pages, ensuring fresh content is always available.
  • Developer Experience Enhancements: Incorporates pre-commit hooks for code quality, formatting, and linting, fostering a consistent and high-quality codebase.
  • Secure Configuration Management: Implements best practices for API key handling via utils modules and params_default.py templates.

Quick Start / Usage

This project primarily functions as a demonstration platform, with its outputs published directly.

  1. Explore Live Demonstrations: The core output of this project is a set of "slides" and generated content deployed to GitHub Pages.

    • Visit the project's deployed content here: https://[YOUR_GITHUB_PAGES_URL_HERE]
    • (Note: The exact URL depends on your repository name and GitHub Pages setup. Please check the "Pages" section in your repository settings if the link above is not active.)
    • Here you can interact with the generated cost comparisons and see the real-time news summarization in action.
  2. Inspect the Codebase: For a deeper dive into how everything works:

    • Clone this repository: git clone https://github.com/[YOUR_ORG/YOUR_REPO].git
    • Explore the utils directory for LLM integrations and helper functions.
    • Review the .github/workflows directory to understand the CI/CD pipeline.

Technical Highlights

  • Advanced LLM Integration: Demonstrates practical use of langchain_google_genai with gemini-1.5-flash, addressing common challenges like model truncation and API key management during development.
  • Intelligent Content Extraction: Utilizes pyparsing to robustly extract structured data from LLM outputs, ensuring reliability for downstream processing.
  • Production-Ready CI/CD: Showcases a sophisticated GitHub Actions setup for continuous artifact generation and deployment, highlighting how to automate content updates from code.
  • Proactive Code Quality: Implementation of pre-commit hooks ensures that code adheres to predefined standards, improving maintainability and collaboration.
  • Strategic Design Choices: Illustrates real-world development decisions, such as pivoting LLM models (gemini-1.5-pro to gemini-1.5-flash) to optimize performance and reliability for specific use cases.

Links to Detailed Documentation

For a comprehensive understanding of the project's development, technical decisions, and deeper insights:

  • Detailed Development Notes: Consult the NOTES.md file for an in-depth log of the project's evolution, design choices, and challenges overcome.
  • Workshop Slides: (If available, link to accompanying workshop presentation slides here.)