Thursday, August 14, 2025

🚀 Two More Certificates, a Setup Gauntlet, and the Agents Road Ahead

I wrapped up the second part of Module 1 in Trustworthy Generative AI and picked up two certificates on the same day: the course certificate and the Prompt Engineering umbrella certificate that completes the three-course sequence. That night I lined up the next step toward Level 07: “AI Agents and Agentic AI with Python & Generative AI.”

(Amusing glitch: I did Module 02 first by mistake, then started Module 01 the next day.)

The setup gauntlet (short + real)

Getting Google Colab + LiteLLM + OpenAI API working took some untangling:

  • API keys changed since the course was written.
    The notebooks were created when OpenAI issued general keys (sk-…). Today keys are project-scoped (sk-proj…). Out of the box, that makes the old code unrunnable until you add a Project ID/header (or equivalent config). Once I did, requests flowed.

  • First time actually enabling API billing.
    I’d generated a key back in Oct 2024, but I’d never set up charges. I funded the project, sent my first two calls, and watched usage register (164 tokens). It’s useful to see real numbers!

Outcome: environment is stable; exercises run cleanly; I’m ready to build agentic workflows.

Takeaway: platform realities evolve faster than course notebooks. A few targeted fixes (key storage, project header, billing) and you’re through.

No comments:

Post a Comment

🧠 Transfer of Consciousness: Moving Long-Running AI Projects Between Chats

When you work with conversational AI over time, you quickly discover an odd limitation: Your project stays in the old chat. The capabiliti...