Mia(@MiaAI_lab):Update for GLM 5.3 Flash for 2x RTX PRO 6000s ✨ Same treatment as the DGX Sparks. - Repo now has the tool-call guard: no-tools turns get a note, and stray tool markup is hidden from the reply. - Each rejected tool call gets its own log line with the reason and name. A debug flag adds the raw text. - The guard never empties a reply now: broken calls stay as text, so agents stop retrying blank replies. - The build has zero compiler warnings. Get it here:

Mia

@MiaAI_lab

Building with AI & LLMs // Running frontier open models locally

Joined July 2022

414 Following 48.7K Followers

Mia@MiaAI_lab

2026.11.12 11:36

Update for GLM 5.3 Flash for 2x RTX PRO 6000s ✨ Same treatment as the DGX Sparks. - Repo now has the tool-call guard: no-tools turns get a note, and stray tool markup is hidden from the reply. - Each rejected tool call gets its own log line with the reason and name. A debug flag adds the raw text. - The guard never empties a reply now: broken calls stay as text, so agents stop retrying blank replies. - The build has zero compiler warnings. Get it here: