Skip to main content

ModeKeeper: verify-first, read-only assessment for Kubernetes/GPU cost and risk

Project description

ModeKeeper

CI PyPI

ModeKeeper is a verify-first, read-only assessment product for Kubernetes/GPU cost and risk, with customer-managed execution and a change-ready handoff pack for enterprise review. Apply/implementation is a separate licensed and gated path, not part of the public assessment path.

Start here (10 minutes)

Use these buyer materials. Canonical workflow: observe -> plan -> verify -> export. Procurement pack language below refers to ./bin/mk-procurement-pack outputs (report/procurement_pack/**), which are distinct from mk export handoff-pack artifacts.

Proof snippet (synthetic replay, read-only):

  • Run: ./scripts/e2e-synthetic-proofs-replay.sh
  • Inspect the multi-signal “combo” summary (example):
# Scenario: combo
- trace: docs/evidence/mk060/observe_raw.jsonl
- active_signals: burst, drift
- k8s_plan_items: 4

What to expect:

  • signals: drift + burst; k8s_plan_items: 4; dry-run only (no apply)
./bin/mk-procurement-pack
cd report/procurement_pack && sha256sum -c checksums.sha256

Verify-first means every deliverable is deterministic, accompanied by SHA256 manifests, and backed by reproducible transcripts. Validation should start from hashes and transcript evidence before any business review.

PRO note: private releases are notes-only. Deliverables are distributed via vendor-provided stamp + transcripts + SHA256.

Release channels and versioning

  • Public (OSS / read-only): this repository + PyPI package modekeeper (wheel-only). GitHub Releases here track the public package.
  • Enterprise PRO (licensed / gated apply): delivered separately as notes-only release metadata + vault-only deliverables (no GitHub assets attached).
  • Why numbers can differ: public and PRO ship independently; the canonical PRO deliverable is always recorded in:
    • docs/SNAPSHOT.md (entrypoint)
    • docs/STATUS.md (latest release + vault + SHA256)

Contact

Documentation index: docs/INDEX.md

  • Questions / feedback: GitHub Issues (preferred) and Discussions.
  • Security issues: please use GitHub Security Advisories (private disclosure). See .github/SECURITY.md.

Reference docs

60-second quickstart

python3 -m pip install -U modekeeper
mk --help
mk observe --source synthetic --duration 10s --out report/quickstart/observe
mk closed-loop run --scenario drift --observe-source synthetic --observe-duration 10s --dry-run --out report/quickstart/plan
PLAN="$(python3 -c 'import json; print(json.load(open(\"report/quickstart/plan/closed_loop_latest.json\", encoding=\"utf-8\"))[\"k8s_plan_path\"])')"
mk k8s verify --plan "$PLAN" --out report/quickstart/verify
mk export handoff-pack --in report/quickstart --out report/quickstart/handoff

# quickstart artifacts
ls report/quickstart
ls report/quickstart/observe
ls report/quickstart/plan
ls report/quickstart/verify
ls report/quickstart/handoff

Expected artifact roots:

  • report/quickstart/observe (read-only telemetry capture)
  • report/quickstart/plan (dry-run planning outputs)
  • report/quickstart/verify (verify outputs)
  • report/quickstart/handoff (handoff-pack export outputs)

Safety gates

Apply/mutate paths are blocked unless all required gates pass:

  • verify_ok=true from verify artifacts
  • kill-switch is absolute (MODEKEEPER_KILL_SWITCH=1 blocks apply)
  • valid license and apply entitlement

Details and command contracts:

Public vs Pro

Public GitHub + PyPI (modekeeper) provides a verify-first, read-only assessment path (observe -> plan -> verify -> export) for Kubernetes/GPU cost and risk. ROI is supporting evidence with explicit assumptions, not a required stage in the canonical path. Outputs are designed as a change-ready handoff pack for enterprise review and customer-managed execution. Apply/mutate capabilities are disabled by default in public and reserved for licensed distribution; see boundary and release rules in docs/DISTRIBUTION_POLICY.md.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distributions

No source distribution files available for this release.See tutorial on generating distribution archives.

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

modekeeper-0.1.33-py3-none-any.whl (110.5 kB view details)

Uploaded Python 3

File details

Details for the file modekeeper-0.1.33-py3-none-any.whl.

File metadata

  • Download URL: modekeeper-0.1.33-py3-none-any.whl
  • Upload date:
  • Size: 110.5 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for modekeeper-0.1.33-py3-none-any.whl
Algorithm Hash digest
SHA256 aaf2910e2cc67de6243dc52d9c845ce70433f8e98790581734ba315354e018c8
MD5 e52358c3fcf5b0ad4a881cba735119b2
BLAKE2b-256 672ca05a48abd7476c958a2788c7e353e027676db752ef3431b2582edaf9d0af

See more details on using hashes here.

Provenance

The following attestation bundles were made for modekeeper-0.1.33-py3-none-any.whl:

Publisher: release-pypi.yml on abcexpert/modekeeper

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page