Day 5 · Week 1
Week 1 · Integrate and verify
Scheduled session · Friday, 18 September 2026
Our planned session
We implement and demonstrate a minimal local agent loop: accept a goal, let Ollama select one allowlisted tool, validate and execute it in application code, return the observation to the model, and present the final answer for human review.
Weekly outcome
A working AI agent powered entirely by a local Ollama model.
Readiness check
We need an editor and one available assistant; Day 1 includes a browser fallback for JavaScript.
Human approval gate
We can explain the difference between a prompted chatbot and an agent, then demonstrate that only an allowlisted tool with validated arguments can run.
This chapter has not been published yet.
We plan to publish each chapter one day before its scheduled session. Session results record work we actually performed and reviewed. Neither the calendar date nor publication alone marks this activity complete.
Expected file: content/365-days-to-fullstack-ai-engineer/day-005.md