Velami: private chats with AI

Hides personal data before it goes to ChatGPT, Claude, Gemini or Perplexity and shows exactly what left. Works locally, no servers.

Short description: Hides personal data before it is sent to ChatGPT, Claude, Gemini and Perplexity and shows exactly what left. Works locally, no servers of its own. Description: Write to ChatGPT, Claude, Gemini and Perplexity as usual. Copilot, DeepSeek, Mistral Le Chat and Grok can be turned on in the settings. Before a message leaves your computer, Velami finds card numbers, IBANs, tax codes, ID documents, dates of birth, licence plates, phone numbers, e-mails and API keys, recognises names and addresses with a model that runs on your computer, and replaces them with labels like [[EMAIL_1]]. The model answers with labels, you see the answer with the real values. Text files you attach are cleaned the same way, PDF, Word, Excel and PowerPoint go as a text copy with labels, and the Copy button gives you the real values. You do not have to trust us. A panel next to the chat shows the exact text the model will get. The log keeps every request as it was sent, with its SHA-256, so you can compare it with DevTools. “Test it yourself” sends a sample to an independent echo server and shows what arrived. If a message ever leaves without passing through Velami, the button turns red. No servers, no accounts, no analytics. Everything stays on this computer. The source code is open (AGPL-3.0): github.com/viatech-labs/velami