Privacy Policy

1. Our Core Privacy Philosophy

At BadiZvezda.eu, privacy is the foundation of our engineering:

• Zero Model Training: We never train AI models on your prompts or responses, and we enforce Zero Data Retention (ZDR) across all upstream inference providers.
• Zero Tracking Cookies: We do not use advertising pixels, cross-site tracking, or behavioral profiling.
• Temporary Chat (Incognito) Mode: You can toggle Temporary Chat at any time to prevent conversations from ever being saved to our server database.

2. Data We Process and Why

3. Sub-Processors & Model Providers

Prompts and voice message audio transcriptions are routed securely via our gateway partner, OpenRouter Inc., to selected model providers (Anthropic, DeepSeek, Xiaomi, Mistral, OpenAI Whisper). Every single request (including speech-to-text audio transcriptions) forces:

• data_collection: "deny" - Rejects any provider endpoint that stores, logs, or trains on text or audio inputs.
• zdr: true - Enforces Zero Data Retention across all completions and voice transcriptions.
• allow_fallbacks: false - Prevents failover to unverified providers.

4. Cookies and Local Storage

We do not use advertising or analytical tracking cookies. We solely utilize essential browser storage (session cookies and local preferences) strictly necessary to authenticate your login and preserve your chosen theme.

5. Your Privacy Rights

You have complete control over your personal data:

• Access & Data Portability: Export your conversations in JSON or Markdown format with a single click.
• Rectification: Update your name and credentials via Account Settings.
• Erasure: Delete individual conversations or use the "Delete Account" button in Account Settings to permanently wipe all your data from our database.
• Support: Contact our team at any time with any questions regarding your personal information.

6. Discord Bot Privacy, Context & Voice Message Hashing

When using the official BadiZvezda Discord Bot (/ask command and voice transcription context menus), our privacy-first principles apply:

• Username & Display Name: Your Discord username and server display name are included in the prompt context strictly to allow the assistant to address you politely and naturally.
• Channel Dialogue History: To maintain continuous dialogue within a chat, the bot stores up to your last 5 messages and the AI's last 5 responses isolated per specific Discord channel. Conversations in one channel are never leaked or shared with another channel.
• Surrounding Channel Context: When you submit a question via /ask, the bot temporarily retrieves up to 5 recent public messages posted within the previous 7 days in that channel to understand what you are referring to. This context is used exclusively for that specific response.
• Voice Message Transcription & Cryptographic Hashing: When you trigger "Transcribe Voice Message" or "Transcribe for All" on an audio message, the audio file is downloaded temporarily to generate a speech-to-text transcript. To accelerate delivery and avoid redundant re-processing of identical audio clips, a fast cryptographic SHA-256 hash of the audio file is computed. The transcript text is stored alongside this hash in our Cloudflare D1 database. Raw audio files are discarded immediately and never permanently stored on our servers.
• Ephemeral Interactions: Ephemeral options (such as "Transcribe Voice Message" and /ask ephemeral: True) send responses visible only to you on Discord.
• Zero Model Training: Just like our web application, no Discord interactions or transcripts are ever used to train AI models.

7. Contact Our Privacy Team

For any privacy inquiries or account data requests, contact our privacy team directly: Email: privacy@badizvezda.eu