Release v0.5.1

CHANGELOG.md

@@ -2,6 +2,8 @@
 
 ## Unreleased
 
+## 0.5.1 (2026-05-01)
+
 ### Fixed
 
 - **`get_file_info` and `delete_files` correctly handle multi-path inputs on real DSM 7.x** (#77) — closes #68. DSM 7.x's `SYNO.FileStation.List getinfo` and `SYNO.FileStation.Delete start` do **not** honor the documented comma-joined multi-path format, even on v2: a request with `path=/a,/b` is treated as a single literal path. For `getinfo` this surfaces as one synthetic record whose `path` field IS the literal comma-joined string (the handler's `len(files) == 1` branch then renders it as a single info card). For `delete_files` this surfaces as a successful task that actually no-ops on every input, returning `[+] Deleted N item(s)` with all paths listed but none removed. The round-1 hypothesis (pin `getinfo` to v2 to dodge a v3 quirk) was disproven by vdsm CI on DSM 7.2.2 — the comma-joined-as-single-path symptom reproduces on v2 too. Fix matches the user's documented workaround on #68: **one DSM call per input path**. Both tools now iterate `paths` and issue per-path requests, aggregating results into the same response shape callers already expect (single info card for one path, table for multiple; per-share recycle-bin messaging unchanged for delete). Trade-off is N round-trips for N paths, which is fine for typical small-N usage and trivially correct. Refactor extracted `_delete_one_path` from `delete_files` so the per-path async-task pattern (start → poll → stop in `try`/`finally`) lives in one place; `get_file_info` simply loops `client.request` since it's synchronous. Bumped `tests/conftest.py` `SYNO.FileStation.List max_version` from 2 to 3 to match DSM 7.x reality so future regression tests don't get fooled by a max-resolves-to-2 default. New `TestGetFileInfo::test_multipath_uses_per_path_serial_calls` asserts (a) N requests for N paths, (b) each request carries a single path with no commas, (c) all pinned to v2, (d) results aggregate correctly. New `TestMultiPathDelete` integration test (re-exported in `tests/vdsm/test_vdsm_integration.py`) creates two folders, deletes both in one multi-path call, verifies via `list_files` that both are actually gone — would have caught the original #68 regression before it shipped if it had existed in v0.5.0. 550 unit tests pass at 96.13% coverage.

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "mcp-synology"
-version = "0.5.0"
+version = "0.5.1"
 description = "MCP server for Synology NAS — manage files on your NAS via Claude"
 readme = "README.md"
 license = "Apache-2.0"

server.json

@@ -3,7 +3,7 @@
   "name": "io.github.cmeans/mcp-synology",
   "title": "Synology NAS",
   "description": "MCP server for Synology NAS — browse files, monitor health, and automate operations",
-  "version": "0.5.0",
+  "version": "0.5.1",
   "repository": {
     "url": "https://github.com/cmeans/mcp-synology",
     "source": "github"
@@ -12,7 +12,7 @@
     {
       "registryType": "pypi",
       "identifier": "mcp-synology",
-      "version": "0.5.0",
+      "version": "0.5.1",
       "transport": {
         "type": "stdio"
       }