fix(grounding): OCR collecte TOUS les matchs + choisit le plus proche de l'ancre
Some checks failed
security-audit / Bandit (scan statique) (push) Successful in 13s
security-audit / pip-audit (CVE dépendances) (push) Successful in 11s
security-audit / Scan secrets (grep) (push) Successful in 8s
tests / Lint (ruff + black) (push) Successful in 14s
tests / Tests unitaires (sans GPU) (push) Failing after 14s
tests / Tests sécurité (critique) (push) Has been skipped
Some checks failed
security-audit / Bandit (scan statique) (push) Successful in 13s
security-audit / pip-audit (CVE dépendances) (push) Successful in 11s
security-audit / Scan secrets (grep) (push) Successful in 8s
tests / Lint (ruff + black) (push) Successful in 14s
tests / Tests unitaires (sans GPU) (push) Failing after 14s
tests / Tests sécurité (critique) (push) Has been skipped
Avant : OCR retournait le premier match → cliquait sur la barre de titre
("CR_patient_demo" dans le path) au lieu du fichier dans la liste.
Après : collecte tous les matchs, choisit le plus proche de la position
originale de l'ancre (anchor_bbox). Si pas de bbox, prend le plus central.
Élimine les clics sur les barres de titre, breadcrumbs, menus.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
@@ -841,7 +841,8 @@ def execute_action(action_type: str, params: dict) -> dict:
|
||||
grounding_result = _shared_find_element(
|
||||
target_text=_fc_target_text,
|
||||
target_description=_fc_target_desc,
|
||||
anchor_image_base64=screenshot_base64
|
||||
anchor_image_base64=screenshot_base64,
|
||||
anchor_bbox=anchor_bbox
|
||||
)
|
||||
if grounding_result:
|
||||
x, y = grounding_result['x'], grounding_result['y']
|
||||
|
||||
Reference in New Issue
Block a user