Feat: add sources chunk content (#639)

* 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

---------

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

Author: fridayL

src/memos/mem_reader/read_multi_modal/file_content_parser.py

@@ -167,14 +167,15 @@ def create_source(
         self,
         message: File,
         info: dict[str, Any],
+        chunk_content: str | None = None,
     ) -> SourceMessage:
         """Create SourceMessage from file content part."""
         if isinstance(message, dict):
             file_info = message.get("file", {})
             return SourceMessage(
                 type="file",
                 doc_path=file_info.get("filename") or file_info.get("file_id", ""),
-                content=file_info.get("file_data", ""),
+                content=chunk_content if chunk_content else file_info.get("file_data", ""),
                 original_part=message,
             )
         return SourceMessage(type="file", doc_path=str(message))
@@ -490,9 +491,6 @@ def parse_fine(
                         f"[FileContentParser] Failed to delete temp file {temp_file_path}: {e}"
                     )
 
-        # Create source
-        source = self.create_source(message, info)
-
         # Extract info fields
         if not info:
             info = {}
@@ -520,8 +518,10 @@ def _make_memory_item(
             mem_type: str = memory_type,
             tags: list[str] | None = None,
             key: str | None = None,
+            chunk_content: str | None = None,
         ) -> TextualMemoryItem:
             """Construct memory item with common fields."""
+            source = self.create_source(message, info, chunk_content)
             return TextualMemoryItem(
                 memory=value,
                 metadata=TreeNodeTextualMemoryMetadata(
@@ -591,6 +591,7 @@ def _process_chunk(chunk_idx: int, chunk_text: str) -> TextualMemoryItem:
                             mem_type=llm_mem_type,
                             tags=tags,
                             key=response_json.get("key"),
+                            chunk_content=chunk_text,
                         )
             except Exception as e:
                 logger.error(f"[FileContentParser] LLM error for chunk {chunk_idx}: {e}")