Build fast without breaking things.
Praevisio: Know your code’s consequences before you merge. Decision‑first pre‑commit checks that adapt to actual risk.
Before you merge, Praevisio tells you what can go wrong, how bad it could be, and the safest way to ship now—or wait 10 minutes for the right proof.
Silent landmines and shipping roulette
Silent Landmines
A tiny change leaks PII to logs, a model hallucinates pricing, or an OSS license trips legal. You don’t see it in code review; you feel it in incidents.
“Ship Now” Roulette
Teams guess whether to add more tests or push. Sometimes they over‑test; sometimes they ship blind. This slows down safe changes.
Unpriced Waivers
Exceptions happen, but without a clock, a cost, or accountability, they linger and bite later when they finally trigger an outage.
No Receipts
When you do make the right call, there’s nothing concrete to show auditors, customers, or your future self how the decision was made.
The Praevisio Workflow
Turns commits into actionable promises
user@devbox:~$ git commit -m "Refactor logger for new feature"
Praevisio Check Summary
2 Promises at riskNo PII in logs or metrics RED
Found potential PII (email pattern) in new log line: log.info(`User $${user.email} success`)
Hallucination rate ≤ X% on critical endpoints AMBER
Model path touched. 500 eval cases needed for confidence.
... 8 minutes later ...
Praevisio: All checks GREEN. Commit proceeded.
Decision receipt posted to PR #142 —View Proof & Signatures
The Decision Receipt is a signed note with links to evidence, providing an audit trail for every risky change.
Starter presets
Promises your code can keep
Pick from enforced promises or write your own in plain language. Praevisio checks only what matters for the code you touched.
Core Engineering
- No PII in logs or metrics.
- No secrets in repo, env, or prompts.
- SLO met: p95 latency ≤ target; error budget respected.
- Cost guardrail: max added € per 1k requests.
- License compliance: no copyleft in prod path.
GenAI / LLM Teams
- Hallucination rate ≤ X% on critical endpoints.
- Toxic/unsafe output rate ≤ Y% (red‑team pack).
- Prompt injection resilience: ≥ target score.
- Retrieval faithfulness: grounding score ≥ threshold.
- PII redaction in model outputs/logs.
Security & Compliance
- Dependency risk: no known critical CVEs in call path.
- Third‑party calls whitelisted; no shadow APIs.
- Data residency: EU‑only data paths for flagged routes.
- Audit‑ready receipts for every risky change.
- Waiver accountability with cleanup expiry.
What leaders get
Control, speed, and proof
Fewer surprise incidents
Address silent landmines before they become public post‑mortems and brand damage.
Faster merges, safer shipping
Stop blanket testing. Run targeted evidence checks to de‑risk only the changes that matter.
Safer exceptions (waivers)
Exceptions are time‑boxed, priced, and auto‑expire, ensuring that cleanup actually gets done.
Proof on tap
Every risky change gets a signed, audit‑ready decision receipt for compliance and customers.
Board‑level scorecard
0.4
Incidents / quarter (€ impact)
3.2 hrs
Time‑to‑merge (median)
92%
Waiver cleanup on time
98%
Receipts coverage (risky PRs)
“If Praevisio prevents one PII slip or one high‑severity GenAI incident this year, it’s paid for itself—everything else is upside.”
Why Praevisio is different
Ship day wins. Breach day doesn’t happen.
Promise‑first
Human‑readable commitments (Safety, Cost, Latency) instead of raw, unmaintainable test lists.
Targeted evidence
Runs the smallest check that actually de‑risks the specific change you made—nothing more.
Economics built‑in
Waivers are priced and time‑boxed so “temporary” doesn’t become “forever.” Exceptions with a seatbelt.