Privacy Policy

What we process

HappyToken processes the content you submit to its multimodal API, including images, video and audio files, related prompts, text-to-speech input, and embedding input. We also process account email addresses, hashed API keys, subscription information, and usage records needed to provide the service. Text-only chat is routed by the local HappyToken router directly to DeepSeek, using the key stored on your Mac, and does not pass through the HappyToken API.

Storage and retention

Images submitted as data URIs and uploaded video or audio files are stored on operator-controlled local storage and become eligible for deletion once they are 30 days old, with a 5 GiB raw-upload limit per user. Expired uploads and their database records are removed automatically when the API starts and during a daily cleanup. HappyToken v1 accepts vision images only as base64 data URIs; remote image URLs are rejected. HappyToken does not persist related prompts, text-to-speech input, or embedding input as raw content. Retained content is never used to train models.

Service providers and model processing

Stripe processes payments and subscription records. Resend delivers sign-in and service emails. Cloudflare protects and transports API traffic through its edge network and secure tunnel. Vercel hosts the HappyToken website and account interface. These providers process the information needed for their role under their own privacy terms.

Multimodal inference runs on operator-controlled local model workers. HappyToken does not send that inference content to an external model API. DeepSeek receives text directly using the key stored on your Mac and under its own terms. HappyToken does not sell personal data.