PRIVACY BEFORE PROMPTING

People routinely paste emails, phone numbers, addresses, credentials, and wallet identifiers into AI tools. The useful context and the sensitive value often sit in the same sentence, making accidental disclosure easy.

Veilora places a review step before submission. It focuses on a small, understandable job: detect common patterns, let the user decide what to hide, and produce a safer prompt that can be copied elsewhere.

The application is a static client experience. Its scanner is a deterministic pattern engine running in the browser. Redaction decisions are represented in transient page state. Safe Prompt generation formats the reviewed text with explicit instructions not to reconstruct placeholders.

Integrity receipts are produced with the browser Web Crypto API. Wallet connection reads only the selected public address and chain ID. Optional proof anchoring is a separate zero-value digest submission with its own preview, confirmation, pending, confirmed, failed, and recovery states.

Original text is not submitted to a Veilora server. The default deployment has no original-text API and no cloud text database. Normal web hosting still receives standard request information such as IP address, user agent, and requested path.

Pattern detection is not a guarantee. Users should review both matches and unmatched text. A sophisticated secret, unusual address format, or context-dependent identifier may not match a rule.

Email, phone, street-address, API-key, and hexadecimal wallet patterns are checked locally. Matching positions are displayed so the user can understand what the rule found. Every finding begins selected for redaction and can be individually kept.

The Safe Prompt is built locally from the reviewed text. Veilora does not run a remote large language model in its default mode and does not present formatting as inference.

A local receipt hashes the exact reviewed text using SHA-256, records the creation time, and assigns a Veilora Proof ID. Changing even one character results in a different digest. Receipts are stored in local browser storage for convenience.

When a verified contract is configured, an explicit proof transaction can submit the digest with zero value. The history records the wallet, contract, transaction hash, block number, and block time only from real wallet and network responses. Local, pending, confirmed, rejected, wrong-network, failed, recovery, and unavailable states are not interchangeable.

Connecting a wallet requests only the selected public address and active chain identifier. Connection does not sign, approve, pay, or transact.

If no proof contract is configured, the interface remains in Local Demo / Unavailable mode and cannot open a proof transaction. Veilora does not include a generic payment, token claim, or points flow.

Regex-style detection can produce false positives and false negatives. Local storage can be cleared, copied, or inspected by someone with device access. A digest proves content consistency, not truth, authorship, ownership, or safe meaning.

Wallet software and networks can fail, reject requests, select an unsupported network, or remain pending. Users are responsible for reviewing wallet prompts and network fees before any configured transaction.

Near-term work may include more local detectors, custom redaction labels, receipt export, accessibility refinements, and optional verified anchor integrations. Any network capability will remain separate from local scanning and require explicit user action.

Roadmap items describe possibilities, not delivery commitments.

Veilora is an assistive privacy tool, not legal, security, compliance, investment, or financial advice. It does not guarantee that text is anonymous, safe to publish, or free of confidential information.

Never paste secrets you cannot risk exposing into any web page. Review output before sharing it with another service. Optional wallet and network features are used at your own discretion.