fix(utils): allow nested output directories (check nearest existing ancestor for writeability)

[Osamaali313]

Problem

build_output_path creates the output directory with mkdir(parents=True, exist_ok=True) — which can create several missing levels at once — but guards it with is_file_writeable, which only checks the immediate parent:

def is_file_writeable(path: Path) -> bool:
    if path.exists():
        return os.access(path, os.W_OK)
    parent_dir = path.parent
    return os.access(parent_dir, os.W_OK)   # immediate parent only

For a multi-level new output directory (e.g. <base>/a/b/c where <base>/a doesn't exist yet), the immediate parent <base>/a/b is itself missing, so os.access(<base>/a/b, W_OK) returns False and build_output_path raises:

MinimaxMcpError: Directory (<base>/a/b/c) is not writeable

…even though mkdir(parents=True) would create the whole chain without any problem. So passing a nested output_directory (e.g. audio/2026/06) to any audio/image/video/music tool fails before the result can be saved.

Fix

Check the writeability of the nearest existing ancestor — the level mkdir(parents=True) actually starts writing into — instead of only the immediate parent:

def is_file_writeable(path: Path) -> bool:
    if path.exists():
        return os.access(path, os.W_OK)
    for ancestor in path.parents:
        if ancestor.exists():
            return os.access(ancestor, os.W_OK)
    return False

Single-level new paths (immediate parent exists) behave exactly as before — no regression.

Tests

Extends test_is_file_writeable with a multi-level new path and adds test_build_output_path_creates_nested_dir (creates a/b/c under a temp dir). The multi-level branch was previously untested (test_make_output_path uses is_test=True, which bypasses the writeability check). pytest tests/test_utils.py passes 7/7.

(Distinct from the open PRs — this is in is_file_writeable/build_output_path, not the validation/timeout/play_audio areas.)

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Adds regression coverage and fixes path writeability detection so nested, non-existent output directories are treated as writable when an existing ancestor is writable.

Changes:

• Update is_file_writeable() to check the nearest existing ancestor instead of only the immediate parent.
• Add tests covering multi-level non-existent paths and verifying build_output_path() creates nested directories.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
tests/test_utils.py	Adds regression tests for multi-level path writeability and nested directory creation via build_output_path().
minimax_mcp/utils.py	Fixes writeability checks for non-existent nested paths by walking up to the nearest existing ancestor.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Osamaali313]

Thanks @copilot —

1. W_OK | X_OK for directories. Good catch, adopted in e5f0ab8. The nearest-existing-ancestor we check is always a directory we'll create entries in, so it now checks os.W_OK | os.X_OK (write + search/execute). Verified it keeps the existing behavior on platforms where directories are always searchable and the tests still pass (7/7); on POSIX it avoids a false positive for a write-but-not-executable directory. I left the path.exists() branch as W_OK since path there may be a regular file (writability ≠ executability), keeping that case's original semantics.

2 & 3. "writeable" vs "writable" in comments. I kept "writeable" deliberately, to match this module's existing public API — the function is named is_file_writeable() and build_output_path raises "Directory (...) is not writeable". Using the same spelling keeps the comments consistent with the symbols they document. Happy to switch the whole module to "writable" if you'd prefer to standardize the spelling, but that felt out of scope for this fix.