cookbook/data_labeling/_07_image_extraction/TEST_LOG.md
Tested 2026-07-18 against gemini-3.5-flash, agno 2.7.4.
Status: PASS
Description: Extracts a typed Scene (subject, setting, time_of_day, dominant_colors, notable_objects) from a photo of Krakow's St. Mary's Basilica via output_schema on a Gemini agent.
Result: Returned Scene(subject="St. Mary's Basilica viewed through the arches of the Cloth Hall in Kraków", setting='outdoor', time_of_day='dawn_or_dusk', dominant_colors=['blue', 'yellow', 'beige']) with five notable_objects including "Sukiennice (Cloth Hall) arches" and "Adam Mickiewicz Monument". Run took 6.9s, 1249 total tokens.
Status: PASS
Description: OCRs a text-heavy image (the Agno intro graphic) into a typed SignReading with primary_text, secondary_text list, and color_scheme.
Result: Returned primary_text='What is Agno', eight secondary_text entries in reading order (from 'Introduction' through 'Level 5: Agentic Workflows with state and determinism.'), and color_scheme='Black, White, Red'. Run took 2.4s, 1246 total tokens.
Status: PASS
Description: Same scene-extraction task with per-field ConfidentStr / ConfidentList wrappers, using a fjord landscape photo from the gstatic gallery.
Result: All five fields populated with confidence='high'; subject value 'A deep fjord valley with a river flowing between steep, green mountains', dominant_colors ['blue', 'green', 'grey', 'brown'], notable_objects ['fjord', 'mountains', 'rocky peak', 'valley', 'river']. Nested wrappers deserialized into the Pydantic models correctly. Run took 3.3s, 1332 total tokens.