第二代仪评指标联识别接口

bbit_ai/app/service/vision.py

@@ -5,10 +5,13 @@ import config.minIO as minIO
 import db.postgres as pg
 from config.minIO import minio_client
 from llm.ticketLLM import *
+from llm.ticketLLMv2 import get_ticket_response_v2
 
 
 def process_ticket_image(
-    img_bytes: bytes,
+    version: int,
+    needBarcode: bool = False,
+    img_bytes=None,
     file_name: str = None,
     project_name: str = None,
     user_id: UUID = None,
@@ -18,6 +21,8 @@ def process_ticket_image(
     """
 
     # 上传到 OSS，使用 UUID 做对象名
+    if img_bytes is None:
+        img_bytes = []
     object_name = str(uuid.uuid4())
     file_bytes = BytesIO(img_bytes)
     bucket_name = "image-ticket"
@@ -28,15 +33,16 @@ def process_ticket_image(
     oss_url = minIO.get_temp_url(bucket_name, object_name)
 
     # 调用分析方法获取 JSON
-    json_data = get_ticket_response(oss_url)
-    # 解析条码
-    barcode = decode_barcode(BytesIO(img_bytes))
-    json_data["barcode"] = barcode
-
+    if version == 1:
+        json_data = get_ticket_response(oss_url)
+    elif version == 2:
+        json_data = get_ticket_response_v2(oss_url, needBarcode)
+    else:
+        json_data = get_ticket_response(oss_url)
     # 获取图片分辨率和大小
     img = Image.open(BytesIO(img_bytes))
     resolution = f"{img.width}x{img.height}"
-    size_kb = len(img_bytes) / 1024
+    size_kb = round(len(img_bytes) / 1024, 2)
 
     # 插入数据库
     pg.insert_ticket_image(
@@ -45,12 +51,13 @@ def process_ticket_image(
         resolution=resolution,
         size=size_kb,
         name=project_name if project_name else object_name[:8],
+        dead_pupa_count=json_data.get("moisture_content") if version == 2 else 0,
         moisture_content=json_data.get("moisture_content"),
         cocoon_weight=json_data.get("cocoon_weight"),
         defective_pupa_count=json_data.get("defective_pupa_count"),
         fresh_shell_weight=json_data.get("fresh_shell_weight"),
         sample_count=json_data.get("sample_count"),
-        barcode=barcode,
+        barcode=json_data.get("barcode"),
         oss=object_name,
         net_weight_total=json_data.get("net_weight_total"),
         evaluator=json_data.get("evaluator"),