Add plain Markdown import and clean export functionality

[TheZupZup]

Summary

This PR adds support for importing plain Markdown files (without NexaNote frontmatter) and exporting notes as clean Markdown files suitable for use in Obsidian and other plain-text tools. Plain .md files dropped into the notes/ directory are now automatically discovered and surfaced as notes without being rewritten, while exports produce frontmatter-free output with sanitized filenames and collision handling.

Key Changes

Plain Markdown Import

• Synthetic ID system: Plain .md files are assigned stable synthetic IDs using the md. prefix followed by URL-safe base64-encoded filenames, allowing the original filename to be recovered without an external index
• Non-invasive reading: Plain Markdown files are read and synthesized into Note objects on-the-fly without rewriting the files, preserving user edits made outside NexaNote
• Seamless coexistence: Plain and NexaNote-managed notes (with frontmatter) can coexist in the same directory and are listed together
• In-place conversion: When a plain Markdown note is edited and saved through NexaNote, it is converted to a managed note with frontmatter in place

Clean Markdown Export

• New export.py module: Provides export_note() and export_all() functions that write notes as body-only Markdown files without YAML frontmatter or internal metadata
• Filename sanitization: sanitize_filename() removes filesystem-unsafe characters (control chars, <>:"/\|?*), collapses whitespace, handles Windows reserved names, and provides a fallback for blank names
• Collision handling: Duplicate titles are disambiguated with (2), (3) suffixes using case-insensitive matching to prevent overwrites on case-insensitive filesystems
• Existing file protection: Export never overwrites pre-existing files in the target directory
• Multi-page support: Multiple pages are joined with blank lines; internal NexaNote markers are stripped

API Endpoint

• POST /export/markdown: New endpoint accepts optional target_dir and include_archived parameters, defaults to <data_dir>/export, and returns a report with the count and paths of exported files

Storage Layer Updates

• file_store.py: Added helper functions plain_md_id_from_stem(), stem_from_plain_md_id(), and synthesize_plain_md_note() to handle plain Markdown discovery and synthesis
• _read_note() and list_notes(): Updated to synthesize plain Markdown notes when deserialization fails (no frontmatter found)
• get_stats(): Fixed to correctly count plain Markdown notes and their pages

Comprehensive Test Coverage

• 378-line test suite covering:
  • Plain Markdown ID round-tripping with Unicode and special characters
  • Import of plain .md files with stable IDs and non-invasive reads
  • Coexistence of plain and managed notes
  • In-place conversion on save
  • Filename sanitization edge cases (invalid chars, reserved names, truncation, whitespace)
  • Clean export with collision detection and case-insensitive deduplication
  • API endpoint behavior and internal storage integrity

Notable Implementation Details

• Plain Markdown files are identified by the absence of NexaNote frontmatter during deserialization, not by a separate index
• Timestamps for plain Markdown notes are derived from filesystem metadata (st_ctime/st_mtime) so external edits are reflected on the next listing
• The export pipeline never modifies internal NexaNote storage; all output is written to a separate target directory
• Soft-deleted notes are automatically excluded from exports via the existing list_notes() filtering

https://claude.ai/code/session_017Mf5G58RPUBKqKQL7zET2m

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d11d84f00d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Sanitize decoded plain-md stems before filesystem access

The new plain-markdown ID path bypasses _safe_id() and directly uses the base64-decoded stem as a path component, so a crafted note_id like md.<base64('../../outside')> resolves outside notes_dir. This affects all operations that call _note_path() (get_note, save_note, delete_note_permanent, etc.), allowing reads/writes/deletes of unintended files whenever an untrusted or malformed ID reaches storage.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Reject Windows reserved basenames even with extensions

The reserved-name guard only checks exact matches (e.g., CON) and misses reserved device names followed by extensions (e.g., CON.txt, NUL.tar.gz), which Windows still treats as invalid. As a result, exporting notes with those titles can fail with OSError on Windows despite the function claiming cross-platform-safe filenames.

Useful? React with 👍 / 👎.