jamiepine/voicebox · error · HTTPException

Failed to clear cache: {str(e)}

Error message

Failed to clear cache: {str(e)}

What it means

Raised (HTTP 500) by POST /cache/clear when clear_voice_prompt_cache() throws. That function clears an in-memory dict and unlinks *.prompt and combined_*.wav files under the configured cache dir, already swallowing per-file unlink errors as warnings — so a propagated exception almost always comes from _get_cache_dir()/config.get_cache_dir() (e.g. misconfigured path), a torch import/init failure, or an OS error creating/reading the directory itself, not from individual file deletes.

Source

Thrown at backend/routes/tasks.py:42

    with progress_manager._lock:
        progress_manager._progress.clear()
        progress_manager._last_notify_time.clear()
        progress_manager._last_notify_progress.clear()

    return {"message": "All task state cleared"}


@router.post("/cache/clear")
async def clear_cache():
    """Clear all voice prompt caches (memory and disk)."""
    try:
        deleted_count = clear_voice_prompt_cache()
        return {
            "message": "Voice prompt cache cleared successfully",
            "files_deleted": deleted_count,
        }
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to clear cache: {str(e)}")


@router.get("/tasks/active", response_model=models.ActiveTasksResponse)
async def get_active_tasks():
    """Return all currently active downloads and generations."""
    task_manager = get_task_manager()
    progress_manager = get_progress_manager()

    active_downloads = []
    task_manager_downloads = task_manager.get_active_downloads()
    progress_active = progress_manager.get_all_active()

    download_map = {task.model_name: task for task in task_manager_downloads}
    progress_map = {p["model_name"]: p for p in progress_active}

    all_model_names = set(download_map.keys()) | set(progress_map.keys())
    for model_name in all_model_names:
        task = download_map.get(model_name)

View on GitHub (pinned to 51f49dea19)

Solutions

1. Confirm the configured cache dir exists and is writable/deletable by the server process.
2. Check the server log — the wrapped exception names the exact path/permission problem.
3. Because per-file deletes are already best-effort, a 500 here usually points to config/permissions, not data — fix the env, not the cache contents.
4. If torch failed to import, validate the Python environment (requirements.txt / image) before retrying.

Example fix

# before
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to clear cache: {str(e)}")

# after (avoid leaking path/error internals; log full detail server-side)
    except OSError as e:
        logger.warning("cache clear hit OS error: %s", e)
        return {"message": "Cache cleared with warnings", "files_deleted": 0}
    except Exception:
        logger.exception("clear_voice_prompt_cache failed")
        raise HTTPException(status_code=500, detail="Failed to clear cache")

// Pre-flight: confirm the cache dir is writable before clearing.
async function canClearCache(api) {
  const health = await api.getHealth().catch(() => null);
  return Boolean(health && health.cache_dir_writable !== false);
}

function isRetryableCacheError(e) {
  return e?.status === 500 && /cache/i.test(String(e.detail ?? ''));
}

try { await api.clearCache(); notify('Cache cleared'); }
catch (e) {
  if (e.status === 500) notify('Could not clear cache — check server permissions/config');
  else throw e;
}

When it happens

Trigger: The cache directory path is unset/misconfigured in config, the process lacks permission to stat the directory, the filesystem is read-only, or the torch dependency failed to initialize so the module-level _memory_cache or torch.save machinery errors.

Common situations: Container runs with a read-only mount for the cache dir, wrong CACHE_DIR env var pointing at a non-writable path, or running as a user without delete permission on the cached files' parent.

Related errors

• str(e)
• str(e)
• Failed to delete model cache directory: {str(e)}
• Audio for capture {capture_id} is missing
• Microphone access is not available. Please ensure: 1. The ap

AI-assisted analysis of jamiepine/voicebox@51f49dea19 (2026-08-12). Data as JSON: /api/errors/fe1e1950b033ebb1. Report an issue: GitHub.