feat: add file ids

src/memos/mem_reader/read_multi_modal/file_content_parser.py

@@ -324,7 +324,7 @@ def parse_fast(
         # For file content parts, default to LongTermMemory
         # (since we don't have role information at this level)
         memory_type = "LongTermMemory"
-
+        file_ids = [file_id] if file_id else []
         # Create memory items for each chunk
         memory_items = []
         for chunk_idx, chunk_text in enumerate(content_chunks):
@@ -351,6 +351,7 @@ def parse_fast(
                     confidence=0.99,
                     type="fact",
                     info=info_,
+                    file_ids=file_ids,
                 ),
             )
             memory_items.append(memory_item)
@@ -373,6 +374,7 @@ def parse_fast(
                     confidence=0.99,
                     type="fact",
                     info=info_,
+                    file_ids=file_ids,
                 ),
             )
             memory_items.append(memory_item)
@@ -499,7 +501,7 @@ def parse_fine(
         session_id = info_.pop("session_id", "")
         if file_id:
             info_["file_id"] = file_id
-
+        file_ids = [file_id] if file_id else []
         # For file content parts, default to LongTermMemory
         memory_type = "LongTermMemory"
 
@@ -536,6 +538,7 @@ def _make_memory_item(
                     confidence=0.99,
                     type="fact",
                     info=info_,
+                    file_ids=file_ids,
                 ),
             )
 

src/memos/memories/textual/item.py

@@ -133,6 +133,11 @@ class TreeNodeTextualMemoryMetadata(TextualMemoryMetadata):
         description="background of this node",
     )
 
+    file_ids: list[str] | None = Field(
+        default_factory=list,
+        description="The ids of the files associated with the memory.",
+    )
+
     @field_validator("sources", mode="before")
     @classmethod
     def coerce_sources(cls, v):