| # Visual Schema | |
| Every visual result includes: | |
| - `asset_id` | |
| - `view_type` | |
| - `lens_name` | |
| - `lens_version` | |
| - `parameters_hash` | |
| - `confidence` | |
| - `payload_json` | |
| - `provenance_json` | |
| Derived views are cached by: | |
| sha256(asset bytes) + lens name + lens version + canonicalized parameters | |
| This lets a later text-only turn reuse previous visual evidence without re-uploading or recomputing the image. | |