Zohan Dvir

[Zohan Dvir @ZohanD32951]

@Ashante493 @maria_rcks Can

[Zohan Dvir @ZohanD32951]

@DamianEdwards The ACs in Europe are all from china lol. Midea is currently making a fortune

[Zohan Dvir @ZohanD32951]

@frnkwltroskr @FilioqueX Cope. Das sind deine Leute, Bruder. Hast du dich mal auf dem AfD Sommerfest umgesehen?

[Zohan Dvir @ZohanD32951]

@pierceboggan Need a Jira integration for that 😅

[Zohan Dvir @ZohanD32951]

@KooKiz Crazy to see it approaching C++ realms of performance. Just a few years ago C++ or rust would have been the only sane choice for something low-level like this

[Zohan Dvir @ZohanD32951]

@AJE_Sport Nation of sore losers

[Zohan Dvir @ZohanD32951]

@_Evan_Boyle Does this also apply for BYOK models? I feel like the bedrock antrophic models perform a bit worse. Does it Cache prompts?

[Zohan Dvir @ZohanD32951]

@github The fact that we can STILL not customize which model the tokens are spent at is crazy.

[Zohan Dvir @ZohanD32951]

@KooKiz How well does it handle prompt caching? A lot of harnesses are struggling with that

[Zohan Dvir @ZohanD32951]

@mkristensen Tales too much space. I disable them entirely

[Zohan Dvir @ZohanD32951]

@alain19651 @derspiegel Leg dich wieder hin Oppa

[Zohan Dvir @ZohanD32951]

@davidfowl @aspiredotdev Why not dnx?

[Zohan Dvir @ZohanD32951]

@mjovanovictech Any Provider you want. I tried with our corporate LiteLLM and it worked fine. If you have the hardware for it you can also try running some qwen or gemma models locally

The API just has to be OpenAI-compatible

[Zohan Dvir @ZohanD32951]

@happyadam73 @rogiervdhee @github Agree, @github Copilot is dead technology. They should just double-down on BYOK so the existing tooling can still be used

[Zohan Dvir @ZohanD32951]

@mkristensen Can we please get the full BYOK experience in Visual Studio? It works quite well in VsCode with the customendpoint vendor

[Zohan Dvir @ZohanD32951]

@rogiervdhee For Blazor it's even worse

Example:
https://t.co/lWfLyuIp5l

[Zohan Dvir @ZohanD32951]

@davidfowl @github You can still not configure the model it uses Right?