v0.7.0

What's Changed

• feat(predict-rlm): vendor Codex LM extra by @glesperance in #30
• feat: lifecycle callbacks for RLM iterations by @magix022 in #34
• feat(examples/appworld): add AppWorld RLM benchmark by @glesperance in #32
• feat(predict-rlm): improve RLM logging and verbose traces by @magix022 in #36
• Persistent supervisor execution, pydantic-faithful predict(), mutable workspaces & runtime hooks (term-bench2 & fractal branches) by @magix022 in #37
• ci: split test jobs by optional extra + add no-extras core job by @magix022 in #38
• fix(sbx): generate sandbox name instead of scraping create stdout by @magix022 in #40
• feat(predict-rlm): reattachable persistent sandboxes + on-demand interrupt (#41, #42) by @magix022 in #44
• fix(predict-rlm): bump aiohttp parent dependencies by @magix022 in #45
• v0.7.0 by @magix022 in #46

Full Changelog: v0.6.1...v0.7.0