v0.9 — Observability & Execution History

Every execution now leaves a trace. jhansi.io gains a full run model — every code execution is stored, searchable, and never silently lost.

What's in it:

• Typed SSE events — output and done event types, so AI agents can act on structured results not raw text

• Run model — every execution stored with full output, exit code, duration, and error summary

• Soft delete — sandboxes are retained after deletion, full history preserved

• Agent tracking — tag sandboxes with agent and created_by at creation time

• Metrics endpoint — live counts, success rate, average duration, broken down by agent

• Rolling log retention — configurable storage cap, oldest logs auto-evicted when limit hit