Hello,
We’ve encountered a serious issue that see...
# ask-questionsb
busy-megabyte-43386
10/28/2025, 7:47 AMHello,
We’ve encountered a serious issue that seems to be connected to your recent implementation changes. After investigating, I noticed that the response format for /api/eval/apiKey has been completely changed.
As a result, our app can no longer parse the experiment data, which has broken all our experiments. This issue is critical and quite costly for us.
Could you please revert to the previous response model and release a fix as soon as possible?
Thank you in advance for your quick assistance.
b
busy-window-96391
10/28/2025, 8:02 AM@calm-dog-24239 Hi!
Could you please ask the team to review this?
We are using remote evaluation, and it seems that the Flutter SDK is currently not handling experiment tracking callbacks correctly due to a changed contract.
🙌 1
c
calm-dog-24239
10/28/2025, 8:04 AMHi @busy-window-96391. Ok, we will review that.
❤️ 1
b
busy-megabyte-43386
10/28/2025, 10:49 AM@calm-dog-24239 Could you please let me know approximately when to expect your response?
c
calm-dog-24239
10/28/2025, 12:53 PMThe only change related to this functionality was that we fixed the deserialization of tracks, since in the Flutter SDK it was using the key
featureResultresultb
busy-megabyte-43386
10/28/2025, 1:04 PMI don’t think the problem is in the Flutter SDK.
The SDK is receiving the response from the data layer (server) — the issue seems to be there.
Could you please check if the tracking callback is working for you now?
Also, could you share the response with me via remote evaluation?
c
calm-dog-24239
10/28/2025, 1:33 PMOh, got it. Ok, please let check that.
❤️ 1
calm-dog-24239
10/28/2025, 3:01 PM• On this PR - tracking callback is working
• Here response for /api/eval/apiKey
{
"features": {
"sloboda-feature": {
"defaultValue": "var_1",
"rules": [
{
"force": "var_1",
"tracks": [
{
"experiment": {
"variations": [
"var_1",
{}
],
"key": "sloboda-feature",
"coverage": 1,
"weights": [
0.5,
0.5
],
"hashAttribute": "id",
"meta": [
{
"key": "0",
"name": "Control"
},
{
"key": "1",
"name": "Variation 1"
}
],
"name": "Sloboda Experiment",
"phase": "0",
"seed": "a23a184c-6dae-4a58-9c17-422cecd055f9",
"hashVersion": 2
},
"result": {
"value": "var_1",
"on": true,
"off": false,
"source": "experiment",
"ruleId": "fr_19g61mmfuukowl",
"experiment": {
"variations": [
"var_1",
"var_2"
],
"key": "sloboda-feature",
"coverage": 1,
"weights": [
0.5,
0.5
],
"hashAttribute": "id",
"meta": [
{
"key": "0",
"name": "Control"
},
{
"key": "1",
"name": "Variation 1"
}
],
"name": "Sloboda Experiment",
"phase": "0",
"seed": "a23a184c-6dae-4a58-9c17-422cecd055f9",
"hashVersion": 2
},
"experimentResult": {
"key": "0",
"featureId": "sloboda-feature",
"inExperiment": true,
"hashUsed": true,
"variationId": 0,
"value": "var_1",
"hashAttribute": "id",
"hashValue": "user_2",
"stickyBucketUsed": false,
"name": "Control",
"bucket": 0.4718
}
}
}
]
}
]
}
},
"experiments": [],
"dateUpdated": "2025-10-28T134936.392Z"
}
b
busy-megabyte-43386
10/28/2025, 3:19 PMYes, it is correct response, that’s odd…
Ok, I will recheck it on my side and come back with more info. Thanks 🤝
c
calm-dog-24239
10/28/2025, 3:20 PMOk, feel free to ask. Looking forward to on your feedback.
🙌 1
h
happy-autumn-40938
10/28/2025, 5:57 PMHi there, unfortunately the old format (right screenshot) is actually the incorrect format and led to several problems (too much information revealed, problems with tracking callbacks when using prerequisites). We recently patched this.
Are you able to update your codebase to reflect the change? Unfortunately we will not be able to revert these changes.
b
busy-megabyte-43386
10/28/2025, 6:16 PMI spent quite a lot of time figuring this out, and yes - I’ve confirmed that in version 1.27 everything works as expected, but in 1.28 the response is completely different.
However, because of these changes, our experiments are now broken. If we make adjustments on our side, it will require releasing a new version of the app, after which users will need to update. Realistically, it takes around 4-6 months to reach a decent update rate among users.
What should we do during this time?
Did you notify anyone that you were going to make changes that are critical for running experiments?
As I mentioned earlier, this has completely broken our experiments.
busy-megabyte-43386
10/28/2025, 6:42 PMP.S. We’ve created our own SDK, but this issue also occurs in the growthbook_sdk_flutter package that you maintain.
h
happy-autumn-40938
10/28/2025, 7:37 PMI believe we didn’t realize this was an unsafe change, but clearly it has affected your application. Are you self-hosting or using cloud?
b
busy-megabyte-43386
10/28/2025, 7:48 PMIn general, we’re self-hosted.
h
happy-autumn-40938
10/28/2025, 8:33 PMAre you able to revert to 1.27 until you're able to roll out app updates?
b
busy-megabyte-43386
10/28/2025, 8:38 PMWe are considering this approach now.
🙌 1
c
calm-dog-24239
10/29/2025, 9:54 AMHi @busy-megabyte-43386. We have updated this PR with changes to deserialize tracking data so that the tracking callback can be invoked. Looking forward for your feedback.
b
busy-megabyte-43386
10/29/2025, 9:59 AMHi, as I mentioned earlier, we are using our own implementation of the GrowthBook SDK. But In any case, I’ll check your PR.
🙌 1
c
calm-dog-24239
10/29/2025, 10:15 AMOh, got it. Ok, we are looking forward for your feedback.
b
busy-megabyte-43386
10/30/2025, 1:07 PMWe checked version 1.27, and it’s actually even worse… :( If the key value is false, it’s not returned by the endpoint at all - it seems that was the main reason for creating version 1.28.
As a result, the core GB feature for running experiments is currently unavailable to us (until we roll out the release to all users, which will take 2–3 months :( ). Due to the data inaccuracies caused by changes in 1.28, we’re canceling all ongoing experiments.
A very disappointing and unfortunate experience…
@calm-dog-24239 @happy-autumn-40938 FYI
h
happy-autumn-40938
10/30/2025, 5:35 PM@busy-megabyte-43386 would it potentially address the issue for you if we add a flag to serve the old payload format? And not scrub the false values? If so I can rig something up
c
calm-dog-24239
10/30/2025, 5:43 PMThere were some changes: https://github.com/growthbook/growthbook-proxy/pull/96
This PR includes a fix that was previously request.
It also contains other changes — in particular, a modification of the trackData response format, which was made to resolve another issue.
As a result, you received the needed fix,
but at the same time, app broke due to other changes in this PR — specifically because of the change in the trackData format, which caused deserialization to fail in their production version.
If we make a new proxy release with the old trackData format, it will fix the issue, but it won’t resolve the other trackData-related issue.
So, there is option:
Make a new proxy release with the old format — this will help, but leave the other issue unresolved.
b
busy-megabyte-43386
10/31/2025, 8:30 AMif we add a flag to serve the old payload formatHello, it will not help because it is required changes in our side => release
h
happy-autumn-40938
10/31/2025, 3:46 PMI’m talking about a flag on the proxy / remote eval lib
b
busy-megabyte-43386
11/04/2025, 3:52 PMHello guys,
We decided to solve this problem on our side and move to a new contract. Kindly ask you not to make any fixes in this way and to inform us in advance about such changes.
Thanks for your attention to this issue 🤝
🙌 1
c
calm-dog-24239
11/04/2025, 4:16 PMHi @busy-megabyte-43386. Thank you for the update.
❤️ 1
26 Views