Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

Vertex AI Agent Engine sandbox `:execute` returning 404 NOT_FOUND in us-central1

Open
#6,706 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
25/100
Issue type
Bug
Clarity
Needs clarification
Activity status
Quiet
Tech stack
google-cloud, python
Domain
ai, api

Research direction

Start by reproducing the failure through client.agent_engines.sandboxes.execute_code(...) and compare it with the sandbox GET response, using the request shapes and SDK versions listed in the issue. Check whether the missing connectionInfo and routing_token can be addressed in the client; done means identifying a client-side change or confirming that a server-side rollout or workaround is required.

Written by the indexing model from the issue text.

Description

api: vertex-ai

Summary

Since 2026-04-28 ~18:51 UTC, calls to the Vertex AI Agent Engine sandbox :execute endpoint return 404 NOT_FOUND ("Requested entity was not found.") for newly created sandboxes in us-central1. The sandbox is verified to be in STATE_RUNNING immediately before the failing call. Affects both production and staging projects, consistent with a server-side rollout.

Versions

  • google-cloud-aiplatform==1.133.0
  • google-genai==1.73.1
  • google-adk==1.29.0
  • Python 3.11

Repro

client.agent_engines.sandboxes.execute_code(...) posts:

POST https://us-central1-aiplatform.googleapis.com/v1/<sandbox>/:execute
Content-Type: application/json

{ "inputs": [{ "mimeType": "application/json",
               "data": "<base64 of {\"code\":\"...\"}>" }] }

Manual probing confirms :execute is reachable and validates input:

Body HTTP
{"inputs":[{"mimeType":"application/json","data":<...>}]} (current SDK shape) 404
{"inputs":[{"mimeType":"text/x-python","data":<...>}]} 400 INVALID_ARGUMENT
{"inputs":[{"mimeType":"text/plain","data":<...>}]} 400 INVALID_ARGUMENT
{"inputData": ...} (legacy) 400, "Unknown name 'inputData'"
{"code": "..."} 400, "Unknown name 'code'"
Legacy :executeCode 404 (empty body)

Additional findings

  1. Race / propagation ruled out. Re-running with a 10s wait after STATE_RUNNING reproduces the same 404.
  2. GET <sandbox> returns only name, displayName, createTime, updateTime, state, spec, expireTime. No connectionInfo field. Per commit 677b55e, SandboxEnvironmentConnectionInfo is where routing_token lives in the new schema. The server isn't returning that block, so no client (old or new) has anything to forward.
  3. HEAD-of-main SDK reproduces the same 404. Built python-aiplatform from main at HEAD (commit 0bb2ecd, past 677b55e), called execute_code() against a fresh staging sandbox in STATE_RUNNING. Result was identical: ClientError 404 NOT_FOUND. {'error': {'code': 404, 'message': 'Requested entity was not found.', 'status': 'NOT_FOUND'}}. The HEAD SDK still emits the mimeType=application/json shape because the server never populates connectionInfo.
  4. The reproducer ran as the same service account our live staging Cloud Run service runs as, ruling out an IAM artefact specific to a probing identity.

Hypothesis

A server-side rollout to :execute in us-central1 on or around 2026-04-28T18:00 UTC changed the contract to require routing information that the API does not currently expose on the sandbox GET payload. The fix may need to ship server-side.

Asks

  1. Can you confirm whether a server-side change to Agent Engine sandbox :execute was rolled out in us-central1 on/around 2026-04-28T18:00 UTC.
  2. Any interim server-side workaround / route still accepting the previous shape?

Tracked internally on Google IssueTracker 507490536.

Dominant language
Python
Stars
907
Forks
467
Avg merge
1d 8h
Merged PRs (30d)
40

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from googleapis/python-aiplatform

All issues in googleapis/python-aiplatform

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.