Compliance evidence collection with Jev
SOC 2 / ISO rows have owners, cadence, and pass/fail. Jev can flag “this evidence blob never mentions production.” The control owner still attests. Jev is not the auditor and not the GRC system of record.
This unofficial page is the evidence collection slice of the compliance evidence pre-score pack. Intent: apply the Jev (TypeSafe System One) decision model to compliance evidence pre-score evidence collection. Primary search language: Compliance Jev evidence collection. Confirm patterns on docs.typesafe.ai. This site does not sell, issue, or proxy TypeSafe keys. Use a credential you already have from the console or a documented gateway.
Independent angle (cover ≠ clone): Checklist is the control; Jev pre-scores whether evidence language looks complete. Owner still signs — not a GRC-clone or rival checklist-IA photocopy. Fan-out extra atoms on one request; open a second HTTP call only for a new artifact, not the same state.
Compliance use-case context
Evidence collection for compliance evidence pre-score happens before POST /v1/systemone. Jev does not browse your warehouse, retriever, or ESP. You gather the control prompt + evidence blob facts, filter them, then ask snap questions. This slice is where fan-out cost math belongs: batch questions, do not re-send state.
Hub: Use cases. Compare, when the other tool is the real job: compliance checklists.
Evidence Collection inputs
Collect:
- The control’s prompt text (do not hope the model memorized last year’s PDF)
- A bounded evidence excerpt
- The one policy sentence the Noul names
Never send:
- Asking Jev to sign the control
- Raw screenshot bytes
- Counsel memos you have not cleared to send
Shape the payload like this once the gather step finishes:
{
"control": { "id": "CC-6.1", "prompt": "Evidence must describe production access reviews this quarter." },
"evidence": { "title": "Access review export", "text": "We reviewed staging users in January." },
"policy": { "env": "Staging-only language is a gap for production controls." }
}Decision signals and actions
Each evidence field should change a named answer:
| Id | Type | Job |
|---|---|---|
mentions_prod |
Noul | Does evidence.text address production as required by control.prompt / policy.env? |
completeness |
Score | How complete does the blob look vs the control prompt? |
next |
Choice | ready_for_owner / gap / owner_review / other |
mentions_prod + completeness + next in one call. Extra Nouls (dates mentioned? sampling described?) are cheap vs another HTTP trip.
Do not treat a Noul of 0.5 as a “medium” compliance evidence pre-score score — it means yes and no are equally likely. Conjunctions stay in your code.
Guardrails and escalation
If the gather step fails (empty control prompt + evidence blob, redaction stripped everything, retriever empty), fail closed on marking a control passed or signing attestation. Do not invent evidence so Jev has something to say. TypeSafe’s confidence-gated examples use a lower bar for recoverable reads than for irreversible actions. Those numbers are illustrations. For compliance evidence pre-score, treat mark_control_passed as the high bar (marking a control passed or signing attestation). Tune on labels — see offline evaluation.
Evaluation and rollout notes
Your eval set should include thin-evidence cases, not only happy control prompt + evidence blobs. Label gap / ready / review gold from control owners, plus production-mention gold. Pin jev-1.13.0 (the versioned id) after you fit thresholds. jev-latest and the marketing line jev-1.13 can move. Log the response model. TypeSafe’s published list price for jev-1.13 is $0.042 per million input tokens (vendor claim — confirm on the models page); output tokens are free on that same page. Unused distractors still bill as input.
Official Python and JavaScript SDKs read TYPESAFE_API_KEY and retry documented 429/529. This site does not sell, issue, or proxy TypeSafe keys. Use a credential you already have from the console or a documented gateway.
Pack map
| Slice | Page |
|---|---|
| Graph and primitives | decision workflow |
What may enter state |
input contracts |
| What to gather first | you are here |
| Atomic rules | policy checks |
| Act / review / abstain | confidence thresholds |
| Reviewer payload | human handoff |
| What to persist | audit trail |
| How it breaks | failure modes |
| Labeled replay | evaluation |
| Shadow → canary | production rollout |
FAQ
Should evidence live in the question text?
Put facts in state and point instructions at control.prompt, evidence.text, policy.env. Criteria stay stable so you can replay.
When do I split calls? mentions_prod + completeness + next in one call. Extra Nouls (dates mentioned? sampling described?) are cheap vs another HTTP trip.
Where is the rest of the Compliance pack? Start with Compliance input contracts and Compliance decision workflow. Cluster hub: Use cases.
Can Jev be our auditor? No. It pre-scores language. Owners and auditors sign. See governance.
May we send screenshots? Not as images. Transcribe what the screenshot shows, then ask snap questions.
What this page does not claim
- Not an auditor, GRC, or certification.
- No readiness or pass-rate claims.
- Not official TypeSafe.
- Official TypeSafe status, or that jev.pro issues API keys.
- That a schema-constrained answer is automatically factually correct.
Disclaimer
This is an independent unofficial site and is not affiliated with TypeSafe AI; official documentation is available at https://docs.typesafe.ai.
Primary documentation: https://docs.typesafe.ai. Hub: Use cases.
Sources
Public TypeSafe or adjacent documentation only. No private claims.