Significant-Gravitas/AutoGPT · error · HTTPException

Graph execution #{graph_exec_id} not found.

Error message

Graph execution #{graph_exec_id} not found.

What it means

Raised (HTTP 404) by the external get-graph-execution endpoint when `execution_db.get_graph_execution(user_id, execution_id, include_node_executions=True)` returns nothing — no execution with that id is visible to the API key's user. Ownership is enforced inside the DB query, so other users' executions look identical to nonexistent ones. (A second 404, `Graph #{graph_id} not found.`, fires when the execution exists but the caller-supplied graph_id doesn't match it.)

Source

Thrown at autogpt_platform/backend/backend/api/external/v1/routes.py:279

@v1_router.get(
    path="/graphs/{graph_id}/executions/{graph_exec_id}/results",
    tags=["graphs"],
)
async def get_graph_execution_results(
    graph_id: str,
    graph_exec_id: str,
    auth: APIAuthorizationInfo = Security(
        require_permission(APIKeyPermission.READ_GRAPH)
    ),
) -> GraphExecutionResult:
    graph_exec = await execution_db.get_graph_execution(
        user_id=auth.user_id,
        execution_id=graph_exec_id,
        include_node_executions=True,
    )
    if not graph_exec:
        raise HTTPException(
            status_code=404, detail=f"Graph execution #{graph_exec_id} not found."
        )

    if not await graph_db.get_graph(
        graph_id=graph_exec.graph_id,
        version=graph_exec.graph_version,
        user_id=auth.user_id,
    ):
        raise HTTPException(status_code=404, detail=f"Graph #{graph_id} not found.")

    return GraphExecutionResult(
        execution_id=graph_exec_id,
        status=graph_exec.status.value,
        nodes=[
            ExecutionNode(
                node_id=node_exec.node_id,
                input=node_exec.input_data.get("value", node_exec.input_data),
                output={k: v for k, v in node_exec.output_data.items()},

View on GitHub (pinned to 9c8bb5550f)

Solutions

1. Use the execution id returned by the execute endpoint (`{"id": ...}`) exactly, and the graph id that created it.
2. List executions for the graph to discover valid, retained execution ids.
3. Confirm the API key belongs to the user who started the execution.
4. If executions were pruned by retention, restart the graph run to produce a fresh execution.

Example fix

# before: swapped ids
GET /graphs/{execution_id}/executions/{graph_id}  # 404

# after
run = POST /graphs/{graph_id}/executions
GET /graphs/{graph_id}/executions/{run.id}

run = client.post(f"/graphs/{graph_id}/executions", json=payload).json()
exec_id = run["id"]  # always use the id returned by the execute endpoint
# poll only this id, and keep graph_id from the same call
client.get(f"/graphs/{graph_id}/executions/{exec_id}")

try:
    client.get(f"/graphs/{graph_id}/executions/{exec_id}")
except HTTPError as e:
    if e.response.status_code == 404:
        execs = client.get(f"/graphs/{graph_id}/executions").json()  # re-discover valid ids
        if exec_id not in {x["id"] for x in execs}:
            log("execution pruned or owned by another user")
    raise

When it happens

Trigger: GET `/api/external-api/v1/graphs/{graph_id}/executions/{graph_exec_id}` with a typo'd/deleted execution id, an execution owned by another user, or a graph_id that doesn't match the execution's actual graph/version.

Common situations: Polling an execution id after it was pruned/retention-expired; mixing ids between environments (staging vs prod); passing the graph id where the execution id belongs or vice versa; version skew after re-uploading a graph (old executions reference old versions invisible to the current graph lookup).

Related errors

• Provider '{provider}' not found
• Block #{block_id} not found.
• Graph #{graph_id} not found.
• Agent #{graph_id} v{graph_version} not found for this user (
• Agent graph #{graph_id} not found

AI-assisted analysis of Significant-Gravitas/AutoGPT@9c8bb5550f (2026-08-14). Data as JSON: /api/errors/8fe4a3e7cea27ef8. Report an issue: GitHub.