@vintcessun: A pure Go LLM privacy gateway that redacts PII/secrets in milliseconds—already in production. Two-layer detection: structured PII via regex (email, phone, ID, bank card), secrets/credentials via gitleaks rules plus Shannan entropy as fallback. Key design: no NER (avoids latency), only irreversible redaction; Go's native regexp is RE2, linear time with no backtracking risk.
Summary
Introduces a pure Go LLM privacy gateway that uses two-layer detection (regex and gitleaks rules) to redact PII and secrets in milliseconds without additional models/GPU, already in production.
View Cached Full Text
Cached at: 06/22/26, 07:40 AM
A pure Go LLM privacy gateway, millisecond-level PII/secret sanitization — already in production.
Two-layer detection: structured PII via regex (email, phone, ID number, bank card), secrets/credentials using gitleaks rules plus Shannan entropy as a fallback.
Key design: no NER (to avoid latency), only irreversible masking; Go’s native regexp is RE2 — linear time with no backtracking risk.
If you’re integrating with an LLM gateway and want to avoid introducing extra models/GPUs, this architecture is worth a look.
Similar Articles
@vintcessun: An 8-stage vulnerability discovery agent that runs on a Claude subscription, essentially a reproduction of the Cloudflare Project Glasswing paper. Multiple narrow agents + intentional disagreement verification + reachability gating, breaking down 'which piece of code can actually be exploited by an attacker' into 8 precise steps. Recon uses Opus to decompose tasks, H…
Introduces an 8-stage vulnerability discovery agent based on Claude subscription, reproducing the core ideas of the Cloudflare Project Glasswing paper. It uses multiple narrow agents, intentional disagreement verification, and reachability gating to break down exploit analysis into precise steps, without the need for API keys.
@geekbb: Burn after reading. Send an encrypted text that self-destructs after one view — the server itself cannot see the content. A zero-knowledge encrypted Pastebin that uses the Web Crypto API for AES-256-GCM encryption in the frontend; only ciphertext is uploaded, the key stays in the URL's # fragment and is never sent to the server.
binthere is a zero-knowledge, end-to-end encrypted pastebin that self-destructs after one read. It uses AES-256-GCM encryption in the browser, stores only ciphertext on Cloudflare Workers, and includes a CLI client.
@realfxw: When away from home, the most annoying things about trying to connect to your home NAS, adjust Ollama on your desktop, or start Moonlight streaming are: no public IPv4, complex large NAT environments, and fear of being constantly scanned by scripts when enabling router port mappings. Try this lightweight private networking tool written in Rust…
Lantunnel is a lightweight private networking tool written in Rust that allows users to securely access home NAS, remotely invoke Ollama, or stream games in environments without public IPv4 or complex NAT, using P2P direct connections and encrypted relays to avoid port exposure.
@XAMTO_AI: Damn, just discovered a site-recon tool called Web-Check, and it's completely free! Just drop any URL in, and it strips the target site down to the bone—unsettlingly like doxxing: DNS records all laid out, server architecture exposed, frameworks, CMS, plugins—nothing stays hidden, open ports, historical snapshots…
Web-Check is a free, open-source website reconnaissance tool. Enter a URL and it extracts DNS records, server architecture, frameworks, CMS, open ports, subdomains, and more—ideal for developers and security researchers.
@GoogleCloudTech: An LLM's job is reasoning, not security. Relying solely on built-in model guardrails leaves you open to advanced attack…
Google Cloud Tech session on securing multi-agent LLM systems using defense-in-depth strategies, including Sensitive Data Protection and Model Armor to prevent prompt injections and data leaks.