Skip to content

ForkCell 0.1.0a2

Pre-release
Pre-release

Choose a tag to compare

@senseb senseb released this 20 Jun 09:30
· 11 commits to main since this release

ForkCell 0.1.0a2 preview focuses on benchmark evidence, receipt safety, and the first restore-path latency cut.

Highlights:

  • Adds representative agent-workspace benchmarks via scripts/benchmark_typical_scenarios.py.
  • Updates README and docs/evidence-summary.md with native/local/volume performance data.
  • Makes native-overlay log collection best-effort non-blocking by default; use --sync-logs for policy/event tests that need to wait for OCSF/log events.
  • Adds scripts/validate_security_bindings.py and CI coverage for common secret-file exclusion plus policy/checkpoint receipt binding.
  • Excludes common local secret files from source import: .env, .env.*, .ssh, .aws, *.pem, *.key.

Representative macOS + Docker Desktop numbers:

backend scenario files restore path
native-overlay small repo 500 550ms
native-overlay medium webapp 2408 445ms
native-overlay dependency/cache 6024 490ms
local-overlay dependency/cache 6024 243ms restore
volume-delta dependency/cache 6024 418ms restore

Boundary note: native-overlay still reports restore_sync_ms=0ms for the workspace substrate. Full governed restore path includes OpenShell sandbox lifecycle and delete; this is not a claim of <60ms fully serviceable MicroVM sandbox startup.

PyPI: https://pypi.org/project/forkcell/0.1.0a2/