ci: parallelize tests and use uv for faster CI (#629)

feat(agent): task execution checkpointing with crash recovery (#626)

fix(dashboard): recognize credit proxy as LLM credential for onboarding (#628)

fix(agent): reduce per-call token overhead by ~6.7K tokens (~18%) (#627)

Merge pull request #625 from openlegion-ai/fix/402-permanent-error

Fix fallback cost/context lookups for openlegion/ models

Strip openlegion/ prefix in cost and context window lookups

Add openlegion provider for credit proxy, fix model rewrite for custom providers

Add 402 (insufficient credits) to permanent errors to prevent useless failover

fix(dashboard): purge dropdown not opening due to click.outside race (#624)

feat(browser): add inter-action delay setting for stealth pacing (#623)

fix(host): emit llm_call events for streaming LLM calls (#622)

fix(dashboard): purge dropdown not opening due to x-transition + @click.outside race (#619)

fix(dashboard): use idle timeout for SSE stream and close TOCTOU race (#621)

fix(agent): align workspace caps and add bootstrap limit feedback (#618)