Test Log - _13_video_classification

Tested 2026-07-18 against gemini-3.5-flash, agno 2.7.4.

basic.py

Status: PASS

Description: Downloads sample_seaview.mp4 from agno-public S3 and runs single-label closed-set classification (scene_type Literal with 7 options) via output_schema on Gemini with native video input.

Result: Returned Classification(scene_type='indoor'). Run took 9.8s (677 input / 10 output tokens, 313 reasoning). Note: despite the filename, the clip shows a scientist at a microscope in a lab, so indoor is correct.

---

with_confidence.py

Status: PASS

Description: Same clip and label set, with an added confidence Literal field (high/medium/low) and instructions defining each confidence tier.

Result: Returned Classification(scene_type='people', confidence='high'). Run took 9.0s (724 input / 15 output tokens, 304 reasoning). Model picks a different valid label than basic.py — both indoor and people fit a person working in a lab.

---