Support for GPT-5.4-mini/nano

v22.7.16

Update Version.cs

Lots of minor improvements

Mostly to prompting

Trying to avoid stale HL connections

Experimenting with pooled connections timeout

To avoid OpenAI stale requests

Experimenting with timeouts

To avoid hanging HL generator requests

Much better prompting and Hyperlambda Generator

v22.7.12

Update Version.cs

10 seconds ConnectTimeout on HttpClient

To avoid stale connections jamming up things.

More HL workflow examples ++

And MUCH better Hyperlambda Generator!!

Better socket life time handling in core for HTTP connections

Plus some other minor goodies

Significantly better backend codegen

Which combined with the new Hyperlambda LLM creates stunning results!

Using GPT-5.4 in default sw-gen model

The primary vibe coding model is now using GPT-5.4 by default. Notice, you'll need to have access to it or downgrade your model during setup.