The drop-down list only provides the ability to add a model for testing. How can I remove a model from testing?

Thanks,

Bon

Hi @bontruong , while hovering a model you can decide to remove it (see screenshot below)
just one thing: once the prompt has been run from a model you can't delete it anymore, but can still remove it for the future responses

This keeps occuring even on re-login {"error":"Failed to perform inference: Invalid username or password."}

Yes its on purpose @inventivecornflower, it to kept the main flow. You can fully reset it or add another flow with a right click if needed

逐项解析一下商业模型，看看你对中国市场了解有多深
（1）1单回本2单提现3单分红模式（高波利裂变，用户躺平收益）
（2）279模式七人拼团模式（“2人回本、7人成团、9（久）享分红”）
（3）139久分红模式（1 次购买返本,3 单分佣,9 池分红）
（4）倍莱鲜羊奶模式
（5）商家让利红包补贴模式（5倍分红收益、商家红包、排队免单）
（6）元气先锋水机排队返模式（水机+门店、进二出一）（7）链动3+1（滑落机制）321模式三三复制，链动2+1（裂变快 快速上线 升级新功能）
（8）推三返一七星团队机制（低波比，快速裂变）
（9）一亩田合伙人共同富裕系统（商业模式导师-王冲团队策划模式）
（10）循环购模式（消费返利、贡献值与分红）
（11）绿色积分增值模式（单边上扬，只涨不跌）
（12）代理分销系统（团队级差 区域代理 团队分润 自动分账）
（13）新零售分销商城（拼团，积分商城，秒杀，砍价，积分兑换，门店自提等）
最后判别哪一种模型最有效裂变最快
哪一种模型既能快速裂变又具有可持续性
有没有可能通过整合优化以上模型创新一种更加适合当前市场的模式？

Keep getting inference failures:
{"error":"Failed to perform inference: JWT has expired: "exp" claim timestamp check failed"}

Makes it pretty un-usable as a tool. Above was in the middle of a chat with Kimi-2.6 though Novita, tried other models & providers, all with the same error in the same chat.

Also Playground needs a way to stop an inference in progress if you see that the thinking is wrong so you can correct your query other than just waste tokens on something that you know will not meet your requirements.

Need a way to increase response limits and timeouts as have in many cases been cut-off in output generation which wastes time and money making the endeavour to use playground also useless.

Hi @SenecaInExile sorry for the delay, didnt see your message.
1 - stop button: I'll add one to make it stoppable
2 - regarding your first answer it was a known issue, do you still have the issue ?

Hi @SenecaInExile sorry for the delay, didnt see your message.
1 - stop button: I'll add one to make it stoppable
Thanks, would help.

2 - regarding your first answer it was a known issue, do you still have the issue ?
Yes, seems to be more prevalent if you are repositioning in the 'desktop' while one of the models is working. (i.e. moving around, trying to re-position, etc). don't know if it's a direct correlation but if there is any movement while one or more models is processing the preponderance of errors (truncated responses, timestamp errors, etc) are much more frequent.

Some other items I'm noticing:
3) should have a means to copy the entire entry or response prompt block for that turn. You can grab the entire flow in json format, but would really like the means similar to say lm studio and other tools where you can copy the entire block /turn (including embedded code, etc in that step/block) to memory or to save it in markdown format. Currently have to basically highlight & scroll to copy the block manually and then paste that in another document. Not too user friendly.

4. should be able to collapse/expand blocks. If you paste in say a handover document or other description document that is long you are scrolling down in playground for page and pages to get to the top/bottom. The quick navigation item on the bottom helps but it's way too small when you have say 500 or even 1000 lines of a document to paste in for review. (turns it into a vertical line). To collapse old steps would be great.

5. should be able to save or default to what models to use when starting a new chat. (perhaps as a profile setting or just restore last used models) opposed to defaulting to the whatever is 'popular' at the moment. Just an annoyance that I have to remove and then add in the models that I want each time I start a new chat.

Thanks for the idea @SenecaInExile
I'll look into it this week

I keep getting this in playground: ("error:" Failed to perform inference. JWT has expired:"exp" claim timestamp check failed")

You've got to extend the timeout in playground to allow the user to answer clarifying questions in a chat session. Currently if output is provided the session terminates way too fast (less than a minute). And there is no way to continue the session without starting over from scratch which just wastes thousands of tokens for no reason. Either lengthen the time to allow the user to respond meaningfully, OR allow a continued session by sending current turn history context with a new connection (sub optimal as it still wastes tokens). But the current state of affairs is the worst option.