Add Gemini OCR and fix VRC log world handling

This commit is contained in:
every_holiday
2026-06-09 02:37:06 +09:00
parent fad1a20089
commit 0521af2933
4 changed files with 148 additions and 35 deletions

View File

@@ -56,16 +56,24 @@ def getSecretValue(section_name, key, default=""):
def loadOcrConfig():
config = loadTomlFile(CONFIG_FILE)
ocr = config.get("ocr", {})
secrets = loadSecretConfig()
ai_secrets = secrets.get("ai", {})
crop = ocr.get("crop", {})
return {
"provider": str(ai_secrets.get("provider", ocr.get("provider", "paddleocr"))).strip().lower(),
"model_name": str(ai_secrets.get("model_name", ocr.get("model_name", ""))).strip(),
"api_key": str(ai_secrets.get("GEMINI_API_KEY", "")).strip(),
"lang": str(ocr.get("lang", "japan")).strip(),
"crop_left": float(crop.get("left", 0.05)),
"crop_top": float(crop.get("top", 0.35)),
"crop_right": float(crop.get("right", 0.95)),
"crop_bottom": float(crop.get("bottom", 0.95)),
"preprocess": bool(ocr.get("preprocess", True)),
"scale": float(ocr.get("scale", 1.5)),
"use_angle_cls": bool(ocr.get("use_angle_cls", False)),
"scale": float(ocr.get("scale", 2.0)),
"use_angle_cls": bool(ocr.get("use_angle_cls", True)),
"delay_seconds": float(ocr.get("delay_seconds", 1.0)),
"show_log": bool(ocr.get("show_log", False)),
}
@@ -74,7 +82,7 @@ def loadVisionConfig():
vision = config.get("vision", {})
return {
"model_name": str(vision.get("model_name", "google/gemma-3-4b-it")).strip(),
"model_name": str(vision.get("model_name", "google/gemma-4-31B-it")).strip(),
"prompt": str(
vision.get(
"prompt",