Significant-Gravitas/AutoGPT · warning · ExpertRunPausedError

{expert.name} hit her weekly credit budget ({spent}/{budget}

Error message

{expert.name} hit her weekly credit budget ({spent}/{budget}); schedules are paused until you resume them.

What it means

ExpertRunPausedError raised at scheduling.py:315 when weekly spend >= effective_weekly_budget(expert). The gate first durably pauses the expert's schedules (pause_expert_schedules), posts a chat message to the user via _post_budget_message(breached=True), then raises. This is the budget churn-guardrail — actual billing is handled separately by the credit system.

Source

Thrown at autogpt_platform/backend/backend/api/features/experts/scheduling.py:315

        return
    if expert.isArchived:
        raise ExpertRunPausedError(
            f"{expert.name} is archived; her schedules do not run.", expert_id
        )
    if expert.schedulesPausedAt is not None:
        raise ExpertRunPausedError(f"{expert.name}'s schedules are paused.", expert_id)
    budget = effective_weekly_budget(expert)
    if budget is None:
        return
    spent = await get_weekly_spend(expert_id)
    if spent >= budget:
        await pause_expert_schedules(
            user_id,
            expert_id,
            reason=f"Weekly credit budget reached ({spent}/{budget})",
        )
        await _post_budget_message(user_id, expert, spent, budget, breached=True)
        raise ExpertRunPausedError(
            f"{expert.name} hit her weekly credit budget ({spent}/{budget}); "
            "schedules are paused until you resume them.",
            expert_id,
        )
    if spent >= int(budget * _BUDGET_WARN_FRACTION):
        await _post_budget_message(user_id, expert, spent, budget, breached=False)


async def _post_budget_message(
    user_id: str,
    expert: prisma.models.Expert,
    spent: int,
    budget: int,
    breached: bool,
) -> None:
    """Post the warning/pause message into the expert's thread. Deduped to
    once per expert per week per kind via the deterministic message id.
    Never raises — a failed post must not affect the run decision."""

View on GitHub (pinned to 9c8bb5550f)

Solutions

Call POST /experts/{expert_id}/schedules/resume once you accept more spend, and/or raise the expert's weekly budget.
Reduce schedule cadence or trim preload rosters to keep weekly spend under budget.
Wait for the weekly window to reset if the spend is time-boxed; the pause persists until explicitly resumed.

Defensive patterns

Strategy: try-catch

Validate before calling

spent, budget = await get_weekly_spend(expert_id), effective_weekly_budget(expert)
if budget is not None and spent >= budget:
    surface_budget_notice(spent, budget)  # don't fire the schedule

Try / catch

try:
    await run_schedule(...)
except ExpertRunPausedError as e:
    if 'credit budget' in str(e):
        prompt_user_to_raise_budget_or_resume(expert_id)

Prevention

Set weekly budgets with headroom for the expert's schedule cadence.
Watch for the warning message posted at the spend threshold (it fires before the hard pause).
Resume schedules explicitly after raising a budget — the breach pause persists otherwise.

When it happens

Trigger: A scheduled run fires for an expert with a weekly budget set, and accumulated weekly spend (get_weekly_spend) has reached or exceeded it. The next firing of any of her schedules produces this error.

Common situations: Aggressive schedule cadences (e.g. roster preloads) burning credits early in the week; a lowered budget taking effect mid-week under existing spend; multiple schedules sharing one expert's budget.

Related errors

AI-assisted analysis of Significant-Gravitas/AutoGPT@9c8bb5550f (2026-08-14). Data as JSON: /api/errors/0d65fda4030f9b04. Report an issue: GitHub.