write_note should preserve frontmatter passed in via content of note

[phernandez]

Feature Description

The write_note tool currently ignores any frontmatter passed in via the content of the note. It should try to preserve it.

For example

{
  `title`: `Custom Frontmatter Test`,
  `folder`: `tests`,
  `content`: `---
Status: In Progress
Priority: High
CustomField: TestValue
Version: 1.0
DueDate: 2025-04-30
---

# Custom Frontmatter Test

This is a test note to verify that custom frontmatter fields are preserved when updating notes.

- [task] Test frontmatter preservation #test #bug
- [note] This note has several custom fields (Status, Priority, CustomField, Version, DueDate)
`
}

will only contain the following frontmatter fields:

title: Custom Frontmatter Test
type: note
permalink: tests/custom-frontmatter-test

[phernandez]

claude: how can we fix this?

[github-actions]

Claude Code Analysis

Based on my analysis of the issue and the codebase, here's what's happening with the frontmatter preservation problem:

Problem Analysis

When using the write_note tool with content that contains frontmatter, the custom frontmatter fields are being stripped out during processing. The issue is in how incoming content with frontmatter is handled:

1. In schema_to_markdown (markdown/utils.py lines 83-84), any frontmatter in the incoming content is detected and removed:

   # if the content contains frontmatter, remove it and merge
   if has_frontmatter(content):
       content = remove_frontmatter(content)

2. Then a new Post is created with only the standard fields (title, type, permalink) plus any metadata from entity_metadata:

   post = Post(
       content,
       title=schema.title,
       type=schema.entity_type,
       permalink=schema.permalink,
       **frontmatter_metadata,
   )

3. For existing notes, the system correctly merges with existing frontmatter, but for new notes or when creating the initial Post object, it discards any frontmatter in the content.

Proposed Solution

The fix would require modifying the schema_to_markdown function in src/basic_memory/markdown/utils.py to parse and preserve frontmatter from the incoming content:

async def schema_to_markdown(schema: Any) -> Post:
    """Convert schema to markdown Post object."""
    # Extract content and metadata
    content = schema.content or ""
    frontmatter_metadata = dict(schema.entity_metadata or {})
    
    # Parse frontmatter from content if present
    content_frontmatter = {}
    if has_frontmatter(content):
        content_frontmatter = parse_frontmatter(content)
        content = remove_frontmatter(content)
        
        # Merge content frontmatter with entity metadata
        # (entity_metadata takes precedence for conflicts)
        content_frontmatter.update(frontmatter_metadata)
        frontmatter_metadata = content_frontmatter

    # Remove special fields for ordered frontmatter
    for field in ["type", "title", "permalink"]:
        frontmatter_metadata.pop(field, None)

    # Create Post with ordered fields
    post = Post(
        content,
        title=schema.title,
        type=schema.entity_type,
        permalink=schema.permalink,
        **frontmatter_metadata,
    )
    return post

This change would:

1. Detect if the incoming content has frontmatter
2. Parse that frontmatter into a dictionary
3. Merge it with any entity_metadata that was passed separately
4. Use the merged metadata when creating the Post object

Files to Modify

Only one file needs to be modified:

• /home/runner/work/basic-memory/basic-memory/src/basic_memory/markdown/utils.py

Testing

A test could be added to verify this behavior:

• Create a note with frontmatter in the content
• Verify the frontmatter fields are preserved in the saved file
• Update the note with modified frontmatter
• Verify all fields (original and updated) are present

The current issue happens because the write_note tool discards any frontmatter in the incoming content, rather than parsing and including it in the note's metadata.

---

🤖 Generated with Claude Code GitHub Action