What causes "Graph execution #{graph_exec_id} not found." in Significant-Gravitas/AutoGPT?

Confirm the execution ID comes from the execute response for the same graph and user.

Significant-Gravitas/AutoGPT · warning · HTTPException

Graph execution #{graph_exec_id} not found.

Error message

Graph execution #{graph_exec_id} not found.

What it means

Raised (404) by GET /graphs/{graph_id}/executions/{graph_exec_id} when execution_db.get_graph_execution returns nothing for the (user, execution_id) pair OR the found execution's graph_id differs from the URI graph_id. Both cases are treated as 'not found' to avoid leaking whether someone else's execution ID exists.

Source

Thrown at autogpt_platform/backend/backend/api/features/v1.py:2196

    path="/graphs/{graph_id}/executions/{graph_exec_id}",
    summary="Get execution details",
    tags=["graphs"],
    dependencies=[Security(requires_user)],
)
async def get_graph_execution(
    graph_id: str,
    graph_exec_id: str,
    user_id: Annotated[str, Security(get_user_id)],
    ctx: Annotated[RequestContext, Security(get_request_context)],
) -> execution_db.GraphExecution | execution_db.GraphExecutionWithNodes:
    result = await execution_db.get_graph_execution(
        user_id=user_id,
        execution_id=graph_exec_id,
        include_node_executions=True,
        organization_id=ctx.org_id,
    )
    if not result or result.graph_id != graph_id:
        raise HTTPException(
            status_code=404, detail=f"Graph execution #{graph_exec_id} not found."
        )

    if not await graph_db.get_graph(
        graph_id=result.graph_id,
        version=result.graph_version,
        user_id=user_id,
        organization_id=ctx.org_id,
    ):
        raise HTTPException(
            status_code=HTTP_404_NOT_FOUND, detail=f"Graph #{graph_id} not found"
        )

    # Apply feature flags to filter out disabled features
    result = await hide_activity_summary_if_disabled(result, user_id)
    onboarding = await get_user_onboarding(user_id)
    if (
        onboarding.onboardingAgentExecutionId == graph_exec_id

View on GitHub (pinned to 9c8bb5550f)

Solutions

Confirm the execution ID comes from the execute response for the same graph and user.
If polling, treat 404 as terminal: stop polling and mark the run as unavailable in the UI.
Check retention settings if executions vanish sooner than expected.

Defensive patterns

Strategy: try-catch

Validate before calling

const runs = await api.listExecutions(graphId);
const known = new Set(runs.map(r => r.execution_id));
if (!known.has(graphExecId)) { dropStalePoller(graphExecId); return; }

Try / catch

catch (e) { if (e.response?.status === 404) { stopPolling(graphExecId); markRunUnavailable(graphExecId); } else throw e; }

Prevention

Treat 404 during polling as a terminal state, not a transient error — never retry it.
Always take execution IDs from the execute response for the same graph_id you put in the URL.
Expire pollers after the execution's expected max age so retention cleanup doesn't strand them.

When it happens

Trigger: Polling an execution that finished and was cleaned up/retention-expired; mistyped or foreign execution ID; polling an execution under the wrong graph_id in the URL (copy-paste mismatch).

Common situations: Frontend run-status polls racing with execution retention cleanup; URLs constructed from stale route state after navigating between agents; API scripts reusing recorded execution IDs after a DB reset.

Related errors

AI-assisted analysis of Significant-Gravitas/AutoGPT@9c8bb5550f (2026-08-14). Data as JSON: /api/errors/16f77e6bab5ba7bb. Report an issue: GitHub.