feat: add file_info for file parser (#651)

* feat: update memos headers

* feat: headers add

* feat: update search agent

* feat: upadte mem story

* feat: update mem scehduler

* feat: update deepsearch mem code

* feat:  update deepsearch agent

* feat: update test code

* fix: remove dup config

* feat: dock search pipeline

* fix: code test

* feat: add test scripts

* feat: add test

* feat: update need_raw process

* fix: add initter

* fix: change agent search func name

* feat: update logs and defined

* feat:  update full text mem search

* feat: cp  plugin  to dev

* feat: add one recall for fulltext retrieval

* fix: set default for fulltext search

* feat: add langchain chunk

* feat: fix playground for query

* feat:  update file content memory extract

* feat: update code

* feat: update import

* code: reformat suffix

* feat: update file_id

* remove langchain-text-splitters==1.0.0

* feat: add reqiuement

* feat: make test

* feat: fix  markdown

* feat: fix simple chunker

* feat: add file sources

* feat:  add concat doc source

* add: file_info

---------

Co-authored-by: CaralHsi <[email protected]>

Author: fridayL

src/memos/mem_reader/read_multi_modal/file_content_parser.py

@@ -170,6 +170,7 @@ def create_source(
         chunk_index: int | None = None,
         chunk_total: int | None = None,
         chunk_content: str | None = None,
+        file_url_flag: bool = False,
     ) -> SourceMessage:
         """Create SourceMessage from file content part."""
         if isinstance(message, dict):
@@ -178,6 +179,7 @@ def create_source(
                 "type": "file",
                 "doc_path": file_info.get("filename") or file_info.get("file_id", ""),
                 "content": chunk_content if chunk_content else file_info.get("file_data", ""),
+                "file_info": file_info if file_url_flag else {},
             }
             # Add chunk ordering information if provided
             if chunk_index is not None:
@@ -202,10 +204,7 @@ def rebuild_from_source(
         # Rebuild from source fields
         return {
             "type": "file",
-            "file": {
-                "filename": source.doc_path or "",
-                "file_data": source.content or "",
-            },
+            "file": source.file_info,
         }
 
     def _parse_file(self, file_info: dict[str, Any]) -> str:
@@ -278,7 +277,7 @@ def parse_fast(
         file_data = file_info.get("file_data", "")
         file_id = file_info.get("file_id", "")
         filename = file_info.get("filename", "")
-
+        file_url_flag = False
         # Build content string based on available information
         content_parts = []
 
@@ -297,6 +296,7 @@ def parse_fast(
                     content_parts.append(f"[File Data (base64/encoded): {len(file_data)} chars]")
                 # Check if it looks like a URL
                 elif file_data.startswith(("http://", "https://", "file://")):
+                    file_url_flag = True
                     content_parts.append(f"[File URL: {file_data}]")
                 else:
                     # TODO: split into multiple memory items
@@ -348,6 +348,7 @@ def parse_fast(
                 chunk_index=chunk_idx,
                 chunk_total=total_chunks,
                 chunk_content=chunk_text,
+                file_url_flag=file_url_flag,
             )
 
             memory_item = TextualMemoryItem(
@@ -384,6 +385,7 @@ def parse_fast(
                 chunk_index=None,
                 chunk_total=0,
                 chunk_content=content,
+                file_url_flag=file_url_flag,
             )
             memory_item = TextualMemoryItem(
                 memory=content,

src/memos/mem_reader/read_multi_modal/user_parser.py

@@ -80,6 +80,7 @@ def create_source(
                                 message_id=message_id,
                                 doc_path=file_info.get("filename") or file_info.get("file_id", ""),
                                 content=file_info.get("file_data", ""),
+                                file_info=file_info,
                             )
                         )
                     elif part_type == "image_url":

src/memos/memories/textual/item.py

@@ -28,6 +28,7 @@ class SourceMessage(BaseModel):
             source is a chat turn.
         - content: Minimal reproducible snippet from the source. If omitted,
             upstream may fall back to `doc_path` / `url` / `message_id`.
+        - file_info: File information for file source.
         - chat_time / message_id / doc_path: Locators for precisely pointing back
             to the original record (timestamp, message id, document path).
         - Extra fields: Allowed (`model_config.extra="allow"`) to carry arbitrary
@@ -40,7 +41,7 @@ class SourceMessage(BaseModel):
     message_id: str | None = None
     content: str | None = None
     doc_path: str | None = None
-
+    file_info: dict | None = None
     model_config = ConfigDict(extra="allow")