Clear, concise answers about ByteQuant
Answers about how the tools work, privacy, output limitations, and site policies.
01Where does tool input go?+
Input is processed in the active browser tab and is not sent to ByteQuant servers. Tools do not store input in localStorage. Theme and tool-usage counts are separate local values with no personal content; the usage count only ranks personal home-page shortcuts.
02Are the tools really free?+
Yes. All core tools are free and require no account. Clearly labeled Google AdSense advertising may support the site and never influences tool output.
03Why are some scores estimates?+
Token and readability measures vary by model or language. Transparent estimates are intended for planning, not as an exact model tokenizer or expert linguistic assessment.
04Does the KVKK/GDPR masker guarantee compliance?+
No. It finds common patterns but cannot catch every name, rare event, or indirect identifier. Human review and, when appropriate, expert advice remain necessary.
05Does ByteQuant call a remote AI model?+
No. Tools and Local Agent call no remote LLM or AI inference API. Fast mode uses versioned semantic planning; the optional Qwen3 model runs on your device through WebGPU. Its open-source model files are downloaded only after you opt in, and chat content is not included in that download request. Plans and short conversation memory stay in the open tab's sessionStorage.
06Can I use very large files?+
The tools are designed for small and medium text and files. Tool pages disclose size, pixel, page, and count limits that protect browser memory. Audited desktop or command-line tools may be better for very large PDFs, images, or deeply nested data.
07Which browsers are supported?+
Current Chrome, Edge, Firefox, and Safari are recommended. Password, UUID, and SHA-256 features need Web Crypto in a secure context; image tools require Canvas and the relevant encoder. WebP/SVG details and available memory vary by browser.
08Why can Android show an old-version warning?+
ByteQuant is a PWA, not an APK. The browser creates the Android app wrapper. If you see the warning, do not accept it: update the browser, remove any previously installed ByteQuant shortcut, reload the page, and reinstall from the current browser menu.
09Is the new summarizer an AI model?+
No. The local summarizer is an explainable extractive method that scores source sentences from word frequency. It writes no new sentence, does not understand context like a person, and can miss critical detail; compare output with the source.
10Does a file hash or CSP audit provide security approval?+
No. SHA-256 compares byte integrity but does not prove safety or source identity. The CSP audit finds a bounded set of structural signals and still requires Report-Only testing on the real site.
11Are Community posts genuinely global?+
Yes—after you explicitly connect, posts, replies, and reactions are published to the open Nostr network through your selected relays. ByteQuant runs no account or relay server: your portable profile secret is encrypted on-device, while public content may be retained by relays and cannot be guaranteed deleted everywhere. Relay operators can see your IP while connected. Device filtering does not govern the whole network, so never publish sensitive data and independently verify untrusted content.
12How do I report a bug or correction?+
Email [email protected] with the page link and a non-sensitive example. X: @byte_quant; Instagram: @byte.quant. Use email for privacy requests.