Skip to content

Releases: tiramitree/benchhandoff

Release list

BenchHandoff v0.5.0

BenchHandoff v0.5.0 Latest
Latest

Choose a tag to compare

View all tags
@tiramitree tiramitree released this 29 Jul 17:20
v0.5.0
c9837dc

BenchHandoff v0.5.0 is a GitHub-only early release of the fail-closed,
resumable experiment-recovery CLI and its optional source-only Kubernetes
AgentRun controller.

Highlights:

  • Windows recovery now uses extended-length paths for deterministic quarantine
    destinations beyond the legacy MAX_PATH boundary.
  • The reference controller uses exactly two manager replicas and one
    pre-created namespaced Lease. Its separate leader-election Role is restricted
    to get and update on that exact Lease.
  • Deterministic fake-client fault injection covers create success,
    AlreadyExists, committed-response loss, and status-conflict convergence.
  • The pinned single-node kind gate separately covers one live-start takeover
    and one terminal-resume/pending-status takeover. It requires the same
    measured Job and Pod identities, the pre-existing passive manager acquiring
    exactly the next Lease transition without a restart, restored business RBAC,
    and one final Job and Pod for each of start, resume, and verify.

Exact public evidence:

The wheel, source distribution, and distribution checksum are the unchanged
bytes built and tested by the annotated-tag CI. The schema-2 takeover record
and its distinct checksum are the privacy-gated bytes from the annotated-tag
real-kind run. No controller image, Helm chart, package-registry publication,
production support, independent review, external use, or adoption is claimed.

This remains a maintainer-controlled, synthetic, fixed single-node experiment
with the API server and storage available. It does not establish strict
fencing, network-partition safety, arbitrary Pod recovery, multi-node or
multi-cluster behavior, exactly-once execution, production high availability,
performance, external adoption, or recruiting outcomes.

Development and documentation were AI-assisted. The claims above are limited
to the linked source revision, automated checks, released artifacts, and stated
boundaries.

Assets 7
Loading

BenchHandoff v0.4.0

BenchHandoff v0.4.0

Choose a tag to compare

View all tags
@tiramitree tiramitree released this 29 Jul 12:52
v0.4.0
c893f63

BenchHandoff v0.4.0 introduces an experimental Kubernetes AgentRun control plane.

Highlights:

  • control.benchhandoff.dev/v1alpha1 AgentRun CRD
  • immutable execution specifications and write-once resume approval
  • a Go controller binding start, resume, and verify Jobs to registered identities
  • a bounded, path-free termination-message protocol
  • a distinct fresh verification Job before success
  • hardened non-root runner Job defaults
  • pinned disposable kind coverage for failure, approval, resume, verification, restart/adoption, drift, duplicate-Pod, and race cases

The controller remains experimental. It does not provide high availability, leader election, storage fencing, exactly-once execution, a production operator, or a supported deployment package.

Rebuilt release provenance

The wheel, sdist, and SHA256SUMS attached here were downloaded without rebuilding from the successful workflow_dispatch CI run executed at the privacy-rewritten v0.4.0 tag.

Evidence boundary

All validation, including the single-node kind exercise, is maintainer-operated synthetic evidence. It is not production validation, independent reproduction, third-party review, performance evidence, external adoption, or a support commitment. There are 0 accepted external-evidence records. This GitHub Release is not a PyPI publication.

Loading

BenchHandoff v0.3.0

BenchHandoff v0.3.0

Choose a tag to compare

View all tags
@tiramitree tiramitree released this 29 Jul 12:52
v0.3.0
e85c565

BenchHandoff v0.3.0 adds bounded closed-world workspace integrity checks while retaining version 1 and version 2 compatibility.

Highlights:

  • suite and evidence schema version 3
  • one dedicated workspace.root and a separately reviewed manifest
  • snapshot-workspace and inspect-workspace
  • bounded, double-scanned topology and ordinary-file byte observations
  • before, after, and recovered workspace bindings
  • workspace-root writer locking
  • atomic no-replace quarantine moves on supported platforms

Workspace checks occur only at registered protocol boundaries. They are not continuous monitoring, a sandbox, or a hostile-writer guarantee.

Rebuilt release provenance

The wheel, sdist, and SHA256SUMS attached here were downloaded without rebuilding from the successful workflow_dispatch CI run executed at the privacy-rewritten v0.3.0 tag.

Evidence boundary

All validation is maintainer-operated synthetic evidence. It is not production validation, independent reproduction, third-party review, hostile-writer safety, external adoption, or a support commitment. There are 0 accepted external-evidence records. This GitHub Release is not a PyPI publication.

Loading

BenchHandoff v0.2.0

BenchHandoff v0.2.0

Choose a tag to compare

View all tags
@tiramitree tiramitree released this 29 Jul 12:52
v0.2.0
c466ea3

BenchHandoff v0.2.0 adds stricter execution-context and process-lifecycle controls while retaining version 1 compatibility.

Highlights:

  • suite and evidence schema version 2
  • byte-verified execution-context descriptors declared as seed inputs
  • executable-content and normalized-path identity binding
  • a minimal, non-inheriting launch environment
  • Windows Job Object and Linux process-group lifecycle controls
  • synthetic child-to-grandchild cleanup and residual-process checks

These controls are not a sandbox, VM snapshot, package inventory, or remote-attestation mechanism.

Rebuilt release provenance

The wheel, sdist, and SHA256SUMS attached here were downloaded without rebuilding from the successful workflow_dispatch CI run executed at the privacy-rewritten v0.2.0 tag.

Evidence boundary

All validation is maintainer-operated synthetic evidence. It is not production reliability, independent reproduction, external use, third-party review, or a support commitment. There are 0 accepted external-evidence records. This GitHub Release is not a PyPI publication.

Loading

BenchHandoff v0.1.0

BenchHandoff v0.1.0

Choose a tag to compare

View all tags
@tiramitree tiramitree released this 29 Jul 12:52
v0.1.0
98e7a9e

BenchHandoff v0.1.0 introduces a local, fail-closed harness for sequential task suites.

Highlights:

  • start, resume, inspect, and verify CLI workflows
  • versioned plans, state, event chains, and final evidence bundles
  • partial-output quarantine and suffix-only recovery
  • evidence-bound resume decisions and cooperative local writer locking
  • canonical synthetic recovery examples and machine-readable evidence
  • Apache-2.0 licensed Python package

Rebuilt release provenance

The wheel, sdist, and SHA256SUMS attached here were downloaded without rebuilding from the successful workflow_dispatch CI run executed at the privacy-rewritten v0.1.0 tag.

Evidence boundary

All validation is maintainer-operated synthetic evidence. It is not production validation, independent reproduction, third-party review, external adoption, or a support commitment. There are 0 accepted external-evidence records. This GitHub Release is not a PyPI publication.

Loading