server_status

[debug]

Returns workspace, warm-up and watcher status. watcherError reports a watcher failure separately from workspaceReloadError.

Try it locally in the TGlider playground

This tool requires no parameters.

Examples

Replace sample paths and symbol keys with values from your workspace.

Check whether a workspace is loaded

{}

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "server_status",
    "arguments": {}
  }
}

Response Notes

Returns loaded-state, workspace root, project count, revision, watcher state, and background compilation warm-up status (workspaceWarmup: enabled, in-progress, current phase — the symbol index first, then the call graph — projects warmed in that phase, and ETA) when available. A load in flight appears under workspaceLoading. That block reports what triggered the load, elapsed time, an ETA, and advice from the previous load of this workspace. lastLoad reports the cost of the last load in this process. A load is synchronous, so a status call that arrives during one returns after it ends. Read lastLoad for the load itself and workspaceWarmup for the phase you can poll. The workspaceLoading ETA and the warm-up ETA read timings that persist across restarts. workspaceReloadError reports a load, reload, or sync that did not leave a complete workspace. It carries the message, timestamp, and paths involved. Failed watcher candidate validation keeps the last useful workspace, so results can omit the failed disk change until recovery. Publication reads the disk again, so a later change can still make reload fail after replacement starts. A watcher-driven reload lists the changed paths. An explicit load or reload leaves them empty. A load that skipped a configuration lists the configurations it could not read. A load rejected before it replaced anything never sets the field. A source sync does not clear a structural candidate or partial-load error. The next complete load or reload clears it. Separately, watcherError reports a watcher failure with message and at fields, or null when clear. HTTP /health reports the same watcherError and watching state. A watcher failure sets watching=false and stops automatic file updates. The loaded workspace and open text remain available, but later disk changes can be absent. See load for watcher recovery and sync for explicit updates.