Session 13 — Project presentations
Goals
- Show a system that runs, not only describe it.
- Present one failure and explain where it happened.
- Compare with a simpler baseline, with numbers.
- Explain one decision in the code, each student on their own.
Three things to show
In the presentation you need to be able to show three things:
- The system runs on a real example from your own collection.
- One case fails, and you know whether the failure is in search, decision, tool or coordination.
- There is a comparison with a simpler version: the same cases, with the number of successes and the number of calls.
Every student needs to explain a decision in the code, even if the work was submitted as a pair. The submission questions and the official files are in Moodle.
A comparison table
One table is enough. It answers whether the complexity you added bought anything:
| Measure | Simple baseline | Your system |
|---|---|---|
| Successes | ||
| Correct refusals | ||
| Calls | ||
| Time |
If your system is not better than the baseline, that is a finding. Present it, and explain what you would test next.
Presenting a failure
Pick one case that failed, and show its log. Answer four questions:
- What the right answer was, and how you know.
- At which step in the log the path went wrong.
- Whether the failure is in search, decision, tool or coordination.
- What you would change, and how you would check that the change helped without breaking other cases.
A failure you can locate shows that you understand the system. A presentation with no failure at all raises the question of whether you tested enough cases.
Before the presentation
- A full run from a clean folder, following the README, on the computer you will present from.
- A fallback in case the network or the model service is unavailable: a saved log of a real run, marked as such.
- Both partners can explain every part of the code, not only the part they wrote.
Hold a full rehearsal as a pair, with a clock. One presents, and the other asks two questions: "Why not a fixed workflow?" and "What happens if this tool fails?". Swap roles.
Check your understanding
Summary
Found a mistake on this page? Email the lecturer