MarkdownDataStore

MarkdownDataStore stores named records as Markdown files whose data is divided into numbered level-three sections. It is useful when people must be able to inspect and edit persisted session or profile data without a database client.

Example record

The following calls create data/sessions/session-1.md and append a second history entry:

const store = new MarkdownDataStore({ dataDir: './data' });

await store.updateFile('sessions', 'session-1', {
    Profile: '- Developer',
    History: '- **User**: Document the request flow'
});

await store.appendToFile('sessions', 'session-1', {
    sections: {
        History: '- **Agent**: Added the MainAgent diagram'
    }
});

The resulting file remains normal Markdown:

### 1. Profile
- Developer

### 2. History
- **User**: Document the request flow
- **Agent**: Added the MainAgent diagram

Main operations

Operation	Behavior
`listFiles(type)`	Lists Markdown record names at the root or within a validated type directory.
`getFile(type, name, selectors)`	Returns all sections and raw Markdown, or selected sections by name or one-based index.
`getSectionMap()`	Returns section content keyed by section name.
`replaceFile()`	Replaces the complete ordered section set.
`appendToFile()`	Appends content to matching sections and can deduplicate normalized lines.
`updateFile()`	Appends to existing sections and creates missing sections.
`deleteFile()`	Deletes selected sections and renumbers the remainder, or deletes the complete file.
`parseList()`, `parseDialogue()`, `parseKeyValue()`	Convert supported Markdown shapes to arrays or records; matching render methods produce canonical Markdown.

Safety and scope

The constructor requires dataDir. Type and file-name normalization prevent traversal outside that directory. Empty section content is normalized to *None*. The store does not interpret the business meaning of a section and does not provide cross-file transactions.