fix: v0.9.1 — clearer messages for exec/redeploy/check_image_updates

Three live-verified MCP weaknesses (update run 2026-06-22):

- exec_in_container: SYNO.Docker.Container/exec does not exist on this DSM
  firmware (code 103, like pause/unpause). Catch it and return a guiding
  "not supported on this firmware" message instead of the raw DSM error;
  the tool stays available for firmwares that do implement exec.
- redeploy_project: the failure hint suggested stop_project + start_project,
  but start_project does not recreate on a new image (it restarts the old
  one). Both failure hints now point back at redeploy_project and warn about
  parallel pull_image lock contention (DSM 2101/2202).
- check_image_updates: docstring and output now state it only reflects DSM
  digest drift on already-pulled tags and does NOT discover newer version
  tags (use list_image_tags + release notes for that).

Tests updated accordingly; CHANGELOG + version bumped to 0.9.1.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

src/mcp_synology_container/modules/containers.py

@@ -269,6 +269,14 @@ def register_containers(mcp: FastMCP, config: AppConfig, client: DsmClient) -> N
                 },
             )
         except Exception as e:
+            if getattr(e, "code", None) == 103 or "does not exist" in str(e).lower():
+                return (
+                    "exec_in_container is not supported on this DSM firmware "
+                    "(Container Manager exposes no 'exec' API method). "
+                    "To read a file inside a container, use the synology-filestation "
+                    "'download' tool on the bind-mount path; to run a command, use the "
+                    "Container Manager UI terminal."
+                )
             return f"Error executing command in '{container_name}': {e}"
 
         output = data.get("output", "")

src/mcp_synology_container/modules/images.py

@@ -369,7 +369,9 @@ def register_images(mcp: FastMCP, config: AppConfig, client: DsmClient) -> None:
 
     @mcp.tool()
     async def check_image_updates(project_name: str | None = None):
-        """Check which local images have updates available (upgradable flag from NAS registry)."""
+        """Report which local images DSM flags as upgradable (digest drift on
+        already-pulled tags). Does NOT discover newer version tags (e.g. 0.8.2 →
+        0.8.3) — use list_image_tags + the image's official releases for that."""
         try:
             data = await client.request(
                 "SYNO.Docker.Image",
@@ -416,4 +418,11 @@ def register_images(mcp: FastMCP, config: AppConfig, client: DsmClient) -> None:
         if not upgradable:
             lines.append("All images are up to date.")
 
+        lines.append("")
+        lines.append(
+            "Note: 'upgradable' reflects digest drift on tags you already have; it "
+            "does NOT detect newer version tags. Use list_image_tags + the image's "
+            "release notes to find newer versions."
+        )
+
         return "\n".join(lines)

src/mcp_synology_container/modules/projects.py

@@ -166,7 +166,9 @@ def register_projects(mcp: FastMCP, config: AppConfig, client: DsmClient) -> Non
         if status not in ("RUNNING", "STOPPED", "BUILD_FAILED", ""):
             return (
                 f"Cannot redeploy '{project_name}': unexpected status '{status}'.\n"
-                f"Workaround: use stop_project + start_project separately."
+                "Wait for the current operation to finish, then retry redeploy_project. "
+                "(start_project alone does NOT recreate on a new image — it only restarts "
+                "the existing container.)"
             )
 
         results: list[str] = []
@@ -269,7 +271,13 @@ def register_projects(mcp: FastMCP, config: AppConfig, client: DsmClient) -> Non
                     f"redeploy_project to recover."
                 )
             else:
-                results.append("Workaround: use stop_project + start_project separately.")
+                results.append(
+                    "The project was not modified. Retry redeploy_project — it stops the "
+                    "project and recreates the containers on the new image. Do NOT use "
+                    "start_project to apply a new tag (it only restarts the old image). A "
+                    "transient DSM 2101/2202 here is usually a lock from a concurrent "
+                    "operation — e.g. a parallel pull_image — so don't pull and deploy at once."
+                )
 
         return "\n".join(results)