Serverless LLMs and Agentic AI with Modal – Lesson 3

💖 Support BrainOmega ☕ Buy Me a Coffee: https://buymeacoffee.com/brainomega 💳 Stripe: https://buy.stripe.com/aFa00i6XF7jSbfS9T218c00 💰 PayPal: https://paypal.me/farhadrh 🎥 In this video, we continue our Serverless LLMs and Agentic AI course with Lesson 3: Custom Images in Modal. After learning how to run functions remotely and control scaling in the previous lessons, we now focus on a critical question: what environment does your code actually run in? In this lesson, you’ll learn how Modal images define the container environment for your serverless AI workloads, without writing Dockerfiles or managing infrastructure. This lesson is fully hands-on and environment-focused. You’ll learn how to build custom Modal images by installing system packages, Python dependencies, and running shell commands during image build time. You’ll also see how to ship local assets—such as templates, configuration files, and prompts—directly into the container. To make everything concrete, we build a small Report Generator Service that downloads a public-domain book during image build, processes it at runtime, and returns both Markdown and HTML reports using a Jinja2 template. By the end of this lesson, you’ll understand how image layers work, why build-time steps matter, and how to design fast, reproducible, and production-ready environments for LLM inference, embeddings, and agentic systems. This lesson gives you the mental model needed to move beyond toy examples and start building real serverless AI applications. 💻 Code on GitHub: https://github.com/frezazadeh/serverless-llm-agentic-ai/blob/main/Lesson3.ipynb ⸻ 📚 What You’ll Learn • What Modal images are and how they define container environments • How to install system packages with apt_install • How to install Python dependencies with pip_install • How to run shell commands during image build time • How to ship local assets using add_local_dir and add_local_file • Why image layer ordering affects rebuild speed • How