feat(memreader): change image llm in file-content as default model

src/memos/mem_reader/read_multi_modal/file_content_parser.py

@@ -326,6 +326,7 @@ def __init__(
         llm: BaseLLM | None = None,
         parser: Any | None = None,
         direct_markdown_hostnames: list[str] | None = None,
+        image_parser: ImageParser | None = None,
     ):
         """
         Initialize FileContentParser.
@@ -341,7 +342,7 @@ def __init__(
         super().__init__(embedder, llm)
         self.parser = parser
         # Initialize ImageParser for processing images in markdown
-        self.image_parser = ImageParser(embedder, llm) if llm else None
+        self.image_parser = image_parser if image_parser is not None else ImageParser(embedder, llm)
 
         # Get inner markdown hostnames from config or environment
         if direct_markdown_hostnames is not None:

src/memos/mem_reader/read_multi_modal/multi_modal_parser.py

@@ -67,11 +67,15 @@ def __init__(
         self.assistant_parser = AssistantParser(embedder, llm)
         self.tool_parser = ToolParser(embedder, llm)
         self.text_content_parser = TextContentParser(embedder, llm)
-        self.file_content_parser = FileContentParser(
-            embedder, llm, parser, direct_markdown_hostnames=direct_markdown_hostnames
-        )
         # Use dedicated image_parser_llm for image parsing (requires vision model)
         self.image_parser = ImageParser(embedder, self.image_parser_llm)
+        self.file_content_parser = FileContentParser(
+            embedder,
+            llm,
+            parser,
+            direct_markdown_hostnames=direct_markdown_hostnames,
+            image_parser=self.image_parser,
+        )
         self.audio_parser = None  # future
 
         self.role_parsers = {