LEGAL
Licenses & Algorithm Credits
Every open-source license and public algorithm source this product relies on. Referenced from the homepage “THE_ENGINE” and “TRUST” sections.
1. Why this page exists
This product runs on real, named algorithms and real open-source software — see "NOT A CHATBOT. AN ENGINE." on the homepage. Naming them honestly means being equally honest about what we owe each source: a license to comply with, or a credit line where one is required. Nothing below is boilerplate — it is a direct, current list of what this product actually uses.
2. Open-source libraries (MIT license)
The following are used under the MIT license, which permits commercial use, modification, and redistribution with attribution. Each link goes to the package's own registry page, which carries its full license text and copyright notice.
| Package | Used for | License source |
|---|---|---|
| React / React DOM | UI framework | npm ↗ |
| Next.js | App framework, static export | npm ↗ |
| @codemirror/* | The coding-assessment editor | npm ↗ |
| onnxruntime-web | Runs the on-device Silero VAD model | npm ↗ |
| @ricky0123/vad-web | Voice-activity detection, self-hosted for CSP | npm ↗ |
| Plyr | The product-tour video player | npm ↗ |
| jsPDF | Score-card PDF export | npm ↗ |
| @clerk/react | Authentication SDK | npm ↗ |
| stripe (Node SDK) | Payments server-side integration | npm ↗ |
| Recharts | Dashboard/analytics charts | npm ↗ |
| motion | Selected UI micro-interactions | npm ↗ |
| @lottiefiles/dotlottie-react | Lottie animation playback | npm ↗ |
| resend | Transactional email (server-side) | npm ↗ |
| jose | JWT verification in the reader worker (server-side) | npm ↗ |
3. Apache License 2.0 — MediaPipe
On-device face and object detection (used for presence coaching and proctoring) runs on Google's MediaPipe library and its published model files (FaceLandmarker, EfficientDet-Lite0), both licensed Apache-2.0 — permissive, commercial use allowed, no royalty.
- The model files are used unmodified, exactly as published.
- Apache-2.0 does not grant trademark rights: nothing on this site implies Google's endorsement, and we do not use the Google or MediaPipe logos.
- MediaPipe reports anonymous performance/usage telemetry to Google as part of normal operation — never your video or audio content (see /privacy Section 3 for the full on-device guarantee).
4. Mozilla Public License 2.0 — Pyodide
The coding assessment's Python execution runs on Pyodide, licensed MPL-2.0 (weak copyleft). It is loaded at runtime from a public CDN rather than bundled or modified by us, so no source-distribution obligation is currently triggered.
5. GSAP
All site animation uses GSAP, including plugins (SplitText, DrawSVG) that were paid add-ons before April 30, 2025 and are now free for any commercial use under GSAP's standard "No Charge" license. We do not strip the license banner comment from the shipped bundle.
6. Algorithm sources & required credit
Algorithms and methods are not protected by copyright — only specific code expressions are (Baker v. Selden, 1879; Google LLC v. Oracle America, 2021). The table below names where each real algorithm we run comes from, and the one credit we are contractually required to give.
| Algorithm | Where it runs | Source / credit |
|---|---|---|
| Elo rating system | Adaptive difficulty (lib/adaptive.ts) | Public method, not copyrightable (17 U.S.C. §102(b)) — named after its creator, Arpad Elo. No license required. |
| SM-2 (ease-factor update only) | Spaced repetition (lib/spaced-rep.ts) | Ease-factor formula from Algorithm SM-2, © SuperMemo World, 1991–2026, supermemo.com. Our interval rule is our own — this is an SM-2-derived implementation, not the canonical SuperMemo product. |
| Keyword-overlap coverage | Resume/ATS match (lib/ats.ts) | A standard, uncopyrightable method for computing text overlap. No license required. |
| Least-squares regression | Readiness forecast (lib/goal-engine.ts) | Standard statistical method (1800s, public domain). No license required. |
| Hamilton (largest-remainder) apportionment | Panel question scheduling (ai/panel.py) | 19th-century public-domain apportionment method. No license required. |
| ε-greedy bandit | Coding-topic selection (lib/coding.ts) | Standard reinforcement-learning technique, public domain. No license required. |
| EWMA / OLS momentum / stdev / coefficient of variation / TTR | Dashboard intelligence engine (lib/dashboard-intel.ts, lib/delivery.ts) | Standard statistical formulas, public domain. No license required. |
Required credit line: “Algorithm SM-2, © Copyright SuperMemo World, 1991.” — supermemo.com · supermemo.eu
7. Questions
If you maintain one of the libraries or algorithms listed above and believe something here is inaccurate, contact us at privacy@interviewforge.ai and we will correct it promptly.