From 1a3d60fcf2592edc0be0d4bc01134dd75ec448e7 Mon Sep 17 00:00:00 2001
From: zhangheng <zhangheng@cowarobot.com>
Date: Mon, 18 May 2026 16:34:58 +0800
Subject: [PATCH] fix(image): record reference source metadata in request.json

---
 gpt_image2_generate.py | 34 ++++++++++++++++++++++++++++++++++
 1 file changed, 34 insertions(+)

diff --git a/gpt_image2_generate.py b/gpt_image2_generate.py
index 786aadf..0dedbc9 100644
--- a/gpt_image2_generate.py
+++ b/gpt_image2_generate.py
@@ -214,6 +214,32 @@ def resolve_image_source(prefix: str) -> str | None:
     return provided[0][1]
 
 
+def describe_image_source(source: str | None, fallback_name: str) -> dict[str, str] | None:
+    if source is None:
+        return None
+
+    value = source.strip()
+    if not value:
+        return None
+
+    lowered = value.lower()
+    if lowered.startswith("data:image/"):
+        source_kind = "data_url"
+        filename = fallback_name
+    elif lowered.startswith(("http://", "https://")):
+        source_kind = "url"
+        filename = Path(urlparse(value).path).name or fallback_name
+    else:
+        source_kind = "file"
+        filename = Path(value).expanduser().name or fallback_name
+
+    return {
+        "source": value,
+        "kind": source_kind,
+        "filename": filename,
+    }
+
+
 def safe_filename(filename: str, fallback: str, mime_type: str) -> str:
     candidate = Path(filename).name.strip()
     if not candidate:
@@ -476,6 +502,8 @@ def main() -> None:
     poll_interval = env_int("IMAGE_POLL_INTERVAL", DEFAULT_POLL_INTERVAL)
     endpoint_path = "/v1/images/edits" if request_mode == "edit" else "/v1/images/generations"
     min_output_edge = resolve_min_output_edge(size, resolution, request_style)
+    reference_image_info = describe_image_source(reference_source, "reference-image")
+    mask_image_info = describe_image_source(mask_source, "mask-image")
 
     print(
         "开始生成图片:",
@@ -508,7 +536,13 @@ def main() -> None:
         "request_style": request_style,
         "min_output_edge": min_output_edge,
         "reference_image_present": bool(reference_source),
+        "reference_image_source": reference_image_info["source"] if reference_image_info else None,
+        "reference_image_source_kind": reference_image_info["kind"] if reference_image_info else None,
+        "reference_image_filename": reference_image_info["filename"] if reference_image_info else None,
         "mask_image_present": bool(mask_source),
+        "mask_image_source": mask_image_info["source"] if mask_image_info else None,
+        "mask_image_source_kind": mask_image_info["kind"] if mask_image_info else None,
+        "mask_image_filename": mask_image_info["filename"] if mask_image_info else None,
         "payload": payload,
     })