Pipeline fails during 'dump' phase if UserProfile path contains non-ASCII characters

[meszkee]

Version

codebase-memory-mcp --0.9.0

Platform

Windows (x64)

Install channel

GitHub release archive / install.sh / install.ps1

Binary variant

standard

What happened, and what did you expect?

Describe the bug

When running index_repository on Windows, the pipeline consistently fails during the dump phase. This happens because my Windows user profile path contains non-ASCII characters (for example: C:\Users\Kovács János).
The SQLite/storage engine seems to fail when attempting to write the graph.db artifact to ~/.codebase-memory due to character encoding issues with accented letters in the path.

Expected behavior

The server should properly handle UTF-8/non-ASCII characters in Windows file paths when resolving the cache or storage directories.

Reproduction

1. Use a Windows machine where the USERPROFILE path contains non-ASCII characters (e.g. C:\Users\Kovács János).

2. Install the MCP server using the official PowerShell installer.

3. Run the index_repository MCP tool on any project.

4. The extraction and resolution phases complete successfully.

5. It crashes exactly at the end with level=error msg=pipeline.err phase=dump.

6. The code being indexed:
   This is not related to the codebase itself. Any standard project (e.g. an empty Laravel or Node.js project) will trigger this if the Windows USERPROFILE path contains non-ASCII characters (e.g. C:\Users\Kovács János).

7. The exact command or MCP tool call:
   Via MCP tool call (index_repository):

{
  "repo_path": "c:\\xampp\\htdocs\\myproject",
  "mode": "fast"
}

### Logs

```text
level=error msg=pipeline.err phase=dump
{"project":"C-xampp-htdocs-myproject","status":"error","hint":"Pipeline failed. Check repo_path exists and contains source files. Try mode='fast' for a quicker diagnostic run."}

Diagnostics trajectory (memory / performance / leak issues)

Project scale (if relevant)

No response

Confirmations

• I searched existing issues and this is not a duplicate.
• My reproduction uses shareable code (a dummy snippet or a public OSS repository), not proprietary code.

[DeusData]

Thanks for reporting this. I have triaged it into 0.9.1-rc as a high-priority Windows/non-ASCII path stability issue. It looks related to the broader non-ASCII path class, but I am keeping this issue separate until the exact failing path is confirmed. If you have the full stderr or pipeline dump failure line, that would help us verify the fix against the right layer.