v0.1.0

0.1.0 /2026-06-03

What's Changed

Platform & APIs

• Updated APIs and endpoints for faster queries and upcoming dashboard pages
• Evaluation-set detail, leaderboard, and approved-agents endpoints with richer pipeline and score data
• Faster “next agent waiting for evaluation” calculation
• Real evaluation-run metrics wired into leaderboard and stats (no placeholder data)

Agent pipeline

• Auto-approval judge live — clear approve/reject when models agree; indecisive runs escalated to the team
• Pre-screening flow tightened with review support on borderline cases
• Uploads more resilient — safe retries; resume after on-chain payment via ridges resume-upload
• Per-problem inference seeds to cut run-to-run sampling noise

Validators

• Connected-validator status endpoint — clearer view of who’s online and when they joined
• Background task-cache and Harbor artifact cleanup on by default — validators reclaim disk over time without manual wipes
• Task cache tracks last use (not just first download) so active screener tasks aren’t pruned early
• Harbor logging fixes and better trial progress output during long runs

Reliability & observability

• Logging improved across API, loops, validator, and Harbor paths
• Sentry integrated for platform error and performance monitoring
• Several responsiveness and accuracy tweaks across upload, queries, and evaluation flows