Two issues were causing telemetry failures: 1. Startup race - memgpt-server sending telemetry before crouton created socket 2. Oversized payloads - large context windows (1M+ tokens) exceeding buffer Changes: - Increase crouton buffer to 128MB max with lazy allocation (64KB initial) - Bump crouton resources (512Mi limit, 128Mi request) - Add retry with exponential backoff in socket backend - Move crouton to initContainers with restartPolicy: Always for deterministic startup 🐙 Generated with [Letta Code](https://letta.com) Co-authored-by: Letta <noreply@letta.com>
15 KiB
15 KiB