Use any LLM without handing it your data. Sensitive values are replaced on your machine before the prompt leaves, and restored in the reply.
A privacy-first chat for multiple LLMs: your own API keys, a local model, or a CLI subscription. Names, emails, IBANs, keys and the rest are redacted on-device, then restored in the answer. → openmasq/openmasq
The same redaction engine as a TypeScript library: detect, replace with realistic fakes, restore from a vault. Offline, Apache-2.0.
npm install @openmasq/redact