- Lab
-
Libraries: If you want this lab, consider one of these libraries.
- Cloud
Triage Multimodal Submissions with Gemini
A customer care team wants Gemini to triage damaged-delivery reports, each one a customer photo plus a short note. Gemini answers every report with confidence, including the reports where the photo cannot support the answer. In this lab, you will send a small set of reports through a Gemini triage service on Google Cloud, classify each one against the team's rubric, find the reports whose evidence is missing, insufficient, or contradictory, and configure the routing rule that sends those reports to a person instead of settling them automatically.
Lab Info
Table of Contents
-
Challenge
Submit image-and-text reports and verify what evidence the model can reliably extract
- Submit the four pending damage reports to the Gemini triage service.
- Compare Gemini's description of each photo and note with the originals.
At the end of this objective, you'll know which parts of Gemini's answer you can check directly against the evidence.
-
Challenge
Apply the supplied rubric to classify each report
- Rate the three evidence requirements for each pending report from its photo and note.
- Choose the one outcome the rubric allows and compare it with Gemini's outcome and confidence.
- Open the auto-routed folder to confirm that the starting rule settled all four reports.
At the end of this objective, you'll have seen the service settle every report automatically, including the reports whose evidence does not support Gemini's outcome.
-
Challenge
Identify reports with missing, insufficient, or contradictory evidence
- Match each pending report to the kind of problem it has, whether missing, insufficient, or contradictory evidence.
- Compare the evidence fields of the three problem reports with those of the complete report.
- Compare the four confidence scores with the starting threshold.
At the end of this objective, you'll know that the evidence fields separate the problem reports from the complete one and the confidence threshold does not.
-
Challenge
Configure the routing rule to send insufficient-evidence reports for human review
- Review the routing rule in the environment variables of the report-triage service.
- Switch on the three evidence requirements and lower the instance limit.
- Deploy the change and confirm that the new revision serves all traffic.
At the end of this objective, you'll have a new revision live that settles a report automatically only when every evidence requirement is shown.
-
Challenge
Test complete and incomplete reports and verify that each reaches the expected outcome
- Resubmit the four pending reports under the new rule.
- Submit the two test reports the rule has not seen.
- Check which folder each report reached and read the condition that decided it.
At the end of this objective, you'll have verified that complete reports settle automatically and incomplete reports go to a person, and you'll know which rule condition decided each route.
About the author
Real skill practice before real-world application
Hands-on Labs are real environments created by industry experts to help you learn. These environments help you gain knowledge and experience, practice without compromising your system, test without risk, destroy without fear, and let you learn from your mistakes. Hands-on Labs: practice your skills before delivering in the real world.
Learn by doing
Engage hands-on with the tools and technologies you’re learning. You pick the skill, we provide the credentials and environment.
Follow your guide
All labs have detailed instructions and objectives, guiding you through the learning process and ensuring you understand every step.
Turn time into mastery
On average, you retain 75% more of your learning if you take time to practice. Hands-on labs set you up for success to make those skills stick.