That's a really cool approach, love it! Treating agent steps as build artifacts (OCI layers) makes total sense for coding agents or offline evaluations where you need filesystem-level reproducibility.
Verist, on the other hand, is aiming for the application layer: lightweight, low-latency observability for production user-facing apps where spinning up containers per-step isn't feasible. I think there's space for both: Dagger/OCI for heavy 'environment' replay, and Verist for semantic 'decision' replay.
Cool project! Been messing with AI agents for code stuff, and this Kanban setup for parallel tasks looks super handy—gonna try it on my next hack. Quick Q: handling token tracking per user/org? Got tons of experience from building UsagePilot (open-core lib for AI metrics). Happy to chat if you need a second pair of hands on that front!
WebSocket Router for Bun — A type-safe WebSocket router for Bun with Zod-based message validation. Route WebSocket messages to handlers based on message type with full TypeScript support.
Verist, on the other hand, is aiming for the application layer: lightweight, low-latency observability for production user-facing apps where spinning up containers per-step isn't feasible. I think there's space for both: Dagger/OCI for heavy 'environment' replay, and Verist for semantic 'decision' replay.