support modifying the prefix of base64.

Author: JDOxygen

oxygent/oxy/llms/base_llm.py

@@ -79,6 +79,8 @@ class BaseLLM(Oxy):
         default=2 * 1024 * 1024,
         description="Maximum non-media file size (bytes) for base64 embedding.",
     )
+    base64_image_prefix: str = Field(default="data:image", description="")
+    base64_video_prefix: str = Field(default="data:video", description="")
     is_disable_system_prompt: bool = Field(default=False)
 
     async def _get_messages(self, oxy_request: OxyRequest):
@@ -170,11 +172,15 @@ async def _get_messages(self, oxy_request: OxyRequest):
 
                 if item_type == "image_url":
                     item[item_type]["url"] = await image_to_base64(
-                        item[item_type]["url"], self.max_image_pixels
+                        item[item_type]["url"],
+                        self.max_image_pixels,
+                        self.base64_image_prefix,
                     )
                 elif item_type == "video_url":
                     item[item_type]["url"] = await video_to_base64(
-                        item[item_type]["url"], self.max_video_size
+                        item[item_type]["url"],
+                        self.max_video_size,
+                        self.base64_video_prefix,
                     )
                 else:
                     logger.warning(

oxygent/utils/common_utils.py

@@ -96,7 +96,9 @@ async def source_to_bytes(source: str):
             return await f.read()
 
 
-async def image_to_base64(source: str, max_image_pixels: int = 10000000) -> str:
+async def image_to_base64(
+    source: str, max_image_pixels: int = 10000000, base64_prefix="data:image"
+) -> str:
     image_bytes = await source_to_bytes(source)
 
     def process_image(image_bytes):
@@ -118,16 +120,18 @@ def process_image(image_bytes):
             return output.getvalue()
 
     image_bytes = await asyncio.to_thread(process_image, image_bytes)
-    return f"data:image;base64,{base64.b64encode(image_bytes).decode('utf-8')}"
+    return f"{base64_prefix};base64,{base64.b64encode(image_bytes).decode('utf-8')}"
 
 
 # 512 * 1024 * 1024 bytes == 512MB
-async def video_to_base64(source: str, max_video_size: int = 512 * 1024 * 1024) -> str:
+async def video_to_base64(
+    source: str, max_video_size: int = 512 * 1024 * 1024, base64_prefix="data:video"
+) -> str:
     video_bytes = await source_to_bytes(source)
     if len(video_bytes) > max_video_size:
         return source
     else:
-        return f"data:video;base64,{base64.b64encode(video_bytes).decode('utf-8')}"
+        return f"{base64_prefix};base64,{base64.b64encode(video_bytes).decode('utf-8')}"
 
 
 async def table_to_base64(source: str, max_table_size: int = 50 * 1024 * 1024) -> str: