Responses API, fixed and added diagnostics, Custom OpenAI model sort (up down buttons)

[sapphirepro]

1. Used as base partly broken Agent side Customer OpenAI (Responses), but fixed it. Now there models work, tools call also operational.
2. Fixed Diagnostics (in chat => "+" => Diagnostics, which previously added random 1 file). Now provides all attached context diagnostics with selection of "errors" only or "all".
3. Added diagnostics tool to agent mode
4. Added sort of Custom OpenAI models, so they can be now manually organized with up/down keys.

Ordinary OpenAI (predefined) remained untouched. Fixed both Custom OpenAI and responses version from broken state in agent mode to operational state.

Please review and if possible accept fully/partly/or modify on need.

Review recommended. For me all works. With kind regards.

[vercel]

@sapphirepro is attempting to deploy a commit to the carlrobertohgmailcom's projects Team on Vercel.

A member of the Team first needs to authorize it.

[sapphirepro]

PS. "Test connection button" for Custom OpenAI(Responses) was not touched and broken, as it seems to expect chat completions protocol. Had no time to fix it from author I took base from for this commit. Probably not a big deal and will be easier for you knowing your code to fix it easier.

And that explains why Responses API custom was needed:

Cache hit rate is much higher then it ever was for me on completions api