Cockpit watches every detection you run through withConflux: speed, errors, cost, how confident each verdict is and where detectors disagree. Use it in your dashboard, or put it inside your own admin with one script tag.
The trial needs a withConflux account. No card needed for the trial.
| Check | Provider | Detections | Errors | Trend |
|---|---|---|---|---|
| AI-generated · image | Sightengine | 4,210 | 0.3% | steady |
| AI-generated · video | BitMind | 1,380 | 5.9% | errors up |
| Moderation | OpenAI | 2,602 | 0.1% | steady |
Detections, speed, errors and cost for each detector, over the last day, week or month.
Every verdict's confidence on one chart, with the threshold drawn on it, so the borderline calls stand out.
When two detectors checked the same file, Cockpit shows how often they gave the same answer.
Email and webhook alerts when the flag rate jumps, errors climb or detectors start to disagree.
Forensic and re-upload checks are view only, on every plan.
Cockpit reads the checks your platform already sends through the gateway. Nothing to change in your code.
Health, alerts and a short weekly note in plain words, in your withConflux dashboard.
One script tag, your domains on the allow-list, and a 15-minute token from your server.
Made-up data on a test account. No real customer.
Leave your email and we send you the demo right away, with the steps to start the 14-day trial.
User uploads at scale: know which detector to trust.
Cockpit for platforms →Sellers' images and listings: fewer false alarms on real products.
Cockpit for marketplaces →Turn your reviewers' decisions into a score per detector.
Cockpit for moderation teams →Health and Accuracy only read the results of checks that already run through withConflux. Only the Control plan can change a threshold or switch a detector off, and only when someone on your team saves it with a reason; every change is logged and can be undone.
Yes. Paste one script tag, allow your domains and have your server ask us for a short-lived token. Your secret key never reaches the browser.
Cockpit can show them, never change them. Forensic paths are view only, on every plan.
Right versus wrong: Cockpit compares each verdict with your own moderation team's decisions, shows false alarms and misses per detector and suggests a better threshold. Your platform sends only the content hash and the decision, never the media.
No. Cockpit works on verdicts, scores, timings and file hashes. No image, video or audio is kept for it.