Agent environments gain precision controls.
Fixes streaming output timeouts so long-running local tools do not prematurely abort model responses.
Isolates step-level execution budgets from individual tool call durations.
Enhances cache cleanup, audit reporting, and mirror configuration for alternate Python runtimes.
Adds explicit command handling for orphan temporary build environments and GraalPy/Pyodide mirrors.
Automates background model upgrades for llama.cpp compatibility while keeping local backups.
Eliminates duplicate list entries for upgraded models and preserves safe downgrade paths.
Recent releases refine execution time, local caching, and model compatibility. Teams running agentic workflows or local runtime infrastructure benefit from tighter timeout isolation and automated migration paths that prevent brittle state errors during complex interactions.
The same card posted on social. Open it for the full size.
