Human-in-the-Loop as a discipline. Airlock sandbox, dependency triage, cross-machine replay, Zenodo reproducibility bundles with TIBET provenance.
Project description
tibet-triage
Human-in-the-Loop as a discipline. Not a rubber stamp.
Airlock sandbox execution, isomorphic mirror, risk-gated triage levels with full TIBET provenance. Every action is pre-computed, every change is diffed, every decision is signed.
The Problem
Compliance frameworks (SOC 2, ISO 27001, NIS2, DORA, PCI DSS, HIPAA) require human oversight. But nobody defines when, how, or with what evidence. The result: operators rubber-stamp everything because stopping the pipeline is worse than the risk.
Triage is not approval. Triage is assessment by someone who sees the difference.
How It Works
Process Request → AIRLOCK (sandbox) → RISK GATE → TRIAGE LEVEL → APPLY
- Airlock: Run any process in an isolated sandbox. Capture all file changes, side effects, and TIBET tokens.
- Risk Gate: Evaluate trigger rules against the airlock result. Automatically determine what level of oversight is needed.
- Triage: Present evidence to the right level of human reviewer.
- Apply: On approval, apply the pre-computed diff. Deterministic. No drift.
Triage Levels
| Level | Who | When |
|---|---|---|
| L0 AUTO | No human | Trust score high, no triggers fired |
| L1 OPERATOR | One operator, async | Moderate risk, standard review |
| L2 SENIOR | Senior + co-signer | System-critical, intent mismatch |
| L3 CEREMONY | Physical presence | Catastrophic risk, out-of-band |
Quick Start
pip install tibet-triage
# Run a command in the airlock
tibet-triage run python deploy.py --source ./app --intent "Deploy API v2.3"
# If triage is needed, review the evidence
tibet-triage pending
tibet-triage review triage-abc123
# Approve or reject
tibet-triage approve triage-abc123 --operator jasper@humotica.com
tibet-triage reject triage-abc123 --reason "Diff too large"
Not Just for AI
tibet-triage is a universal process integrity engine. The same airlock + mirror + risk gate pattern works for:
- CI/CD pipelines — pre-compute deployments, review before apply
- Scientific computing — reproducible experiments with evidence chains
- Supply chain — package build verification before publish
- Financial transactions — pre-compute + compliance check + execute
- Infrastructure — terraform plan is a primitive airlock; this is the full version
- Medical data — HIPAA minimum necessary, proven by diff
Part of the TIBET Ecosystem
- tibet-core — Token provenance
- tibet-pol — Process integrity checker
- tibet-forge — Code certification
- tibet-audit — Compliance reporting
- tibet-trail — Chain tracing
License
MIT — Humotica AI Lab
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file tibet_triage-0.2.0.tar.gz.
File metadata
- Download URL: tibet_triage-0.2.0.tar.gz
- Upload date:
- Size: 30.4 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.5
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
4c74d427946d8cef151240201f8644ef98e7f0ebe9e850f3726dec992f920a85
|
|
| MD5 |
a60563bcadef93b528f36e7904feac87
|
|
| BLAKE2b-256 |
1d22dd5607112449c7f8245a20620f0e144dfbc06ee0d28ed0325dfcc9d5291b
|
File details
Details for the file tibet_triage-0.2.0-py3-none-any.whl.
File metadata
- Download URL: tibet_triage-0.2.0-py3-none-any.whl
- Upload date:
- Size: 35.7 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.5
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
299edb61153cd6d956bb9d49f4aac83fa0da6754e32b93d18ede648679d8ba87
|
|
| MD5 |
cac20a63562686a65bde36b060e4c64b
|
|
| BLAKE2b-256 |
cec784f547e56a9d36492a95c8a4b4b9e241fce56eeb9bb26c32cf8d1b3d1f42
|