Privacy Policy
Memo AI is an internal tool built for Memo Fashion Limited staff. This policy explains what data is collected, how it is used and your rights under UK GDPR.
1. Who we are
Memo AI is operated by Memo Fashion Limited, a UK-registered company. The platform is engineered and maintained in-house.
2. Data we collect
- Account data: Your name, work email, role, and authentication credentials (managed by Supabase Auth).
- Chat history: Messages you send and AI responses, linked to your account so you can resume conversations.
- Memory facts: After every conversation, a small fast model extracts useful facts about you (your role, preferences, the people you work with) and stores up to 30 per user. You can view, edit and delete every memory at any time at /memo-ai/memories.
- Knowledge base documents: If you upload PDFs / Word / Excel / CSV to your personal knowledge base, the content is chunked, embedded (Gemini text-embedding-004, 768d) and stored in pgvector. Each document and chunk is scoped to your user id. Delete any document at /memo-ai/knowledge and its embeddings are cascade-deleted.
- Voice audio (transient): Voice-mode microphone audio is sent to Groq Whisper for transcription and immediately discarded — never stored. TTS audio is generated by Cloudflare MeloTTS and streamed to your browser, never persisted server-side.
- Uploaded files: Documents, images and spreadsheets you attach to a chat. Stored in Cloudflare R2 with per-user isolation. Delete the conversation to remove them.
- Usage metadata: Daily message counts (used for rate limiting) and `ai_events` audit rows for observability. No tracking pixels, no third-party analytics.
3. How we use your data
Your data is used solely to operate Memo AI for you and other authorised staff. It is never sold, shared with advertisers, or used to train external models. Chat history is stored so you can return to previous conversations.
4. Third-party AI providers
Memo AI routes inference through several AI providers via API. Each is contracted to not train on your data under their API privacy terms:
- Groq, SambaNova, Cerebras, OpenRouter, NVIDIA NIM, Together-style providers — chat inference. Zero data retention.
- Google Gemini API — Live web search grounding, embeddings (RAG), TTS, PDF reading. Google's API tier has zero retention and no training.
- Cloudflare Workers AI — image generation (FLUX.2) and voice synthesis (MeloTTS).
- Anthropic API — Claude Haiku used as the final OCR fallback for receipt scanning only. Anthropic's API tier has zero retention.
- Resend — outbound transactional email (password reset, weekly digest).
No data is sent to consumer-tier products (ChatGPT, Claude.ai, Gemini app, Microsoft Copilot).
5. Temporary chat mode
When you use Temporary chat, messages are not saved anywhere — not in the database, not in logs, not in backups. They exist only in your open browser session. Closing the chat erases it permanently.
6. Your rights under UK GDPR
You have the right to access, correct, export or delete your personal data at any time. Memory entries, knowledge-base documents and conversations can all be deleted directly from the in-app settings pages. For broader requests (export all data, full account deletion), use the contact form.
7. Retention
Chat history is retained for as long as your account is active, capped at 50 conversations per user (oldest auto-pruned). Memory facts are capped at 30 per user. Knowledge-base documents are capped at 50 per user. You can manually delete any item at any time and the deletion is permanent.
8. Security
All data is transmitted over HTTPS and stored encrypted at rest. Supabase Row-Level Security policies enforce per-user isolation. Per-user and per-IP rate limits protect against abuse. Authentication uses industry-standard secure cookies.
9. Changes to this policy
This policy may be updated periodically. Significant changes will be announced inside the app. The "last updated" date at the top reflects the most recent revision.