Launching today
OpticParse & PhishVision
Vision AI Scraping API & Zero-False-Positive Phishing Check
5 followers
Vision AI Scraping API & Zero-False-Positive Phishing Check
5 followers
Web scraping breaks when DOMs change. Phishing detectors fail on modern fakes. We fixed both by building an API that looks at websites exactly like humans do—using Vision AI. Extract structured JSON or detect phishing in 0.3 seconds without maintaining a single CSS selector.
It would be great if you could add a confidence score to the phishing detection responses so I can decide whether to auto-block, flag for review, or just log it. Also a quick dry-run mode that shows what the vision model actually "saw" on the page would help when I'm tuning thresholds.
@solaker70811 Thanks for the awesome feedback!
1. Confidence Score: We actually already return a `confidence_score_percentage` in the API response payload, so you can immediately use that to drive your auto-block/flagging logic! (I've just updated our API docs to make this clearer).
2. Dry Run Mode: We loved this idea so much that we just built and deployed it! You can now pass `"dry_run": true` in your JSON payload to our `/api/phish-detect` endpoint. It will bypass the AI vision model and instantly return the raw base64 screenshot and DOM telemetry without consuming any of your AI tokens. We also added a toggle for it directly in the developer dashboard so you can test it easily.
Let me know how the threshold tuning goes!