Crown Cron Evidence
Daily dogfood evidence for OpenClaw/Crown scheduled jobs: which crons run locally, which model handled them, whether they succeeded repeatedly, and what model class appears to fit each cron type.
Local-only routing is clean
No enabled cron job currently has a paid/non-local primary or fallback route.
Daily local-cron reliability history
One snapshot per day, retained for 30 days. This is the proof trail for repeated local-model cron operation.
| Date | Enabled | Local-only | Paid | Risk | Semantic warn | Best observed local model |
|---|---|---|---|---|---|---|
| 2026-07-11 | 0 | 0 | 0 | 0 | 0 | — |
| 2026-07-10 | 0 | 0 | 0 | 0 | 0 | — |
| 2026-07-09 | 65 | 65 | 0 | 0 | 2 | local-qwen-coder/qwen-coder (100%, 1232 runs) |
| 2026-07-08 | 66 | 66 | 0 | 0 | 0 | local-qwen-coder/qwen-coder (100%, 976 runs) |
| 2026-07-07 | 65 | 65 | 0 | 0 | 1 | local-mtplx/mtplx-qwen36-27b-optimized-speed (100%, 826 runs) |
| 2026-07-06 | 66 | 66 | 0 | 7 | 1 | local-mtplx/mtplx-qwen36-27b-optimized-speed (100%, 763 runs) |
| 2026-07-05 | 65 | 65 | 0 | 0 | 3 | local-mtplx/mtplx-qwen36-27b-optimized-speed (100%, 706 runs) |
Model reliability, last 7 days
| Model | Runs | OK | Warn | Avg |
|---|
Cron class fit, last 7 days
| Class | Runs | OK | Warn | Avg |
|---|
Current routing recommendations
Short frequent checks need low overhead; escalate to MTPLX only when DS4 times out or needs broader reasoning.
Longer browsing/metadata synthesis has already been more stable on MTPLX than DS4.
Project/eval jobs need more context and tool discipline; Qwen-Coder is useful for code/test fallback.
DS4 hit a model-call timeout on Daily Lilly email triage; MTPLX passed the shortened prompt locally.
Jobs needing attention
No current risk jobs or latest semantic warnings.