VOUQIS
AI Change Verification

Verify AI changes before they merge.

Vouqis runs your evals on every pull request and posts a clear verdict — BLOCK, WARN, or SAFE — before your team decides to merge. No guessing. No skipped evals. Evidence on every PR.

FreeOpen source (MIT)GitHub ActionPython CLI
vouqis verify · PR #47POSTED
## What Changed
prompts/ 4 files ████████
evals/ 2 files ████
config/ 1 file ██
eval score 0.71 / 1.00
threshold 0.80
BLOCK MERGE
eval score below threshold
0.71 < 0.80 · blocked
// without vouqis: this PR would have merged without evidence

The PR looked fine.

The eval was never run.

Teams merging AI changes without eval evidence is the default. Prompts shift, configs change, and nobody has a score to point at when production degrades.

PR status
Merged
Eval run?
No
Score before
Unknown

The code review said LGTM. The diff looked small. Nobody ran the eval. Vouqis makes it automatic.

How it works

Vouqis sits in your CI pipeline. On every PR it detects which AI files changed, runs your eval, and posts a verdict comment. Four steps. No SDK. No per-file wrappers.

Pull RequestVouqis VerifyYour EvalPR Comment
PR OPENED
on every push
DETECT
changed AI files
RUN EVAL
your eval command
POST VERDICT
BLOCK / WARN / SAFE
One CI step. Vouqis detects the changed files, runs your eval, and posts the verdict to the PR automatically.
vouqis.yml
# minimal config to get started
eval_command: "python evals/run.py"
threshold: 0.80
baseline: origin/main
ai_paths:
- prompts/
- evals/
- config/model.yml
ready — 3 paths monitored
What happens on every PR
01Add vouqis verify to your CI workflow
02On every PR, detect which AI files changed
03Run your eval command — any script, any framework
04Post a structured verdict to the PR as a comment

Every PR becomes a data point. Every merge becomes a decision with evidence.

Three verdicts. No ambiguity.

Every Vouqis report ends with one of three verdicts. The logic is deterministic — same eval result, same changed files, same verdict.

BLOCK MERGE

Your eval failed. The score fell below the threshold. This change is not ready.

MERGE WITH WARNING

Eval passed but AI files changed. Review carefully before you ship.

SAFE TO MERGE

Eval passed. No AI file changes detected. Ship it.

Every verdict includes
Verdict + ConfidenceWhat Changed (per path)Eval score + thresholdWhy (cited reasons)Feedback link

Every PR report shows exactly what moved.

Before the verdict, Vouqis shows a per-path breakdown of which AI files changed and how many — so reviewers know exactly what to look at.

Vouqis Verify Report · PR #47BLOCK MERGE
## What Changed
PathFilesTypeShare
prompts/4Prompt Files57%
evals/2Eval Scripts29%
config/1Model Config14%
Eval score0.71 / 1.00
Threshold0.80
Latency1,240 ms
BLOCK MERGE
eval score 0.71 is below threshold 0.80
4 prompt files changed — model behavior may have shifted
AI file paths monitored
prompts/457%
evals/229%
config/114%

7 AI files changed. One command to know if they're safe to ship.

Watch it verify one.

One command. Detects changed files, runs your eval, delivers a verdict. Works in CI or locally before you push.

vouqis — verify
without vouqis: PR merged without eval evidencewith vouqis: verdict posted to PR before merge
Early access

We're working with teams shipping AI changes.

We're working with a small number of teams merging AI-generated changes without eval coverage. If your team makes prompt or model decisions without evidence today, this is for you.

What you get
Direct access to the Vouqis team
Shape the eval integration before v1 ships
Weekly check-ins during onboarding
Your PR patterns define the roadmap
Apply now →