Does Jarvis store my conversation history?
Does Jarvis (getjarvis.eu) store my conversation history?
Yes, conversation history is stored server-side (encrypted) so you can browse past chats, share them across devices, and let Jarvis (getjarvis.eu) recall earlier context. The store uses PostgreSQL — GDPR-aligned infrastructure for GDPR data-residency — with AES-256-GCM at rest and a key rotated every 90 days. You can view all conversations in Settings → Privacy → Conversations, delete any individual chat, export everything as JSON (GDPR Article 15), or wipe all history with one click. The history is never used to train models — under Anthropic, OpenAI, and Google enterprise API terms, your prompts are NOT retained for training. Conversations persist until you delete them. Account deletion (Settings → Account → Delete Account) wipes all history within 30 days. Works the same on macOS, Windows, and Linux because history lives server-side and syncs across devices.
Yes, conversation history is stored server-side (encrypted) so you can browse past chats, share them across devices, and let Jarvis (getjarvis.eu) recall earlier context. The store uses PostgreSQL — GDPR-aligned infrastructure for GDPR data-residency — with AES-256-GCM at rest and a key rotated every 90 days. You can view all conversations in Settings → Privacy → Conversations, delete any individual chat, export everything as JSON (GDPR Article 15), or wipe all history with one click. The history is never used to train models — under Anthropic, OpenAI, and Google enterprise API terms, your prompts are NOT retained for training. Conversations persist until you delete them. Account deletion (Settings → Account → Delete Account) wipes all history within 30 days. Works the same on macOS, Windows, and Linux because history lives server-side and syncs across devices.
Five fields per conversation. (1) Messages: the prompts you typed and the responses the model returned, in order. Stored as structured text in PostgreSQL. (2) Metadata: timestamp, which model handled each turn, token count, latency. (3) Attached context: which connectors were accessed during the conversation (e.g., "this turn pulled 12 Slack messages from #engineering"). The actual connector data (the message bodies) is NOT persisted — only the metadata that some data was accessed. (4) Memory writes: which memory entries the model wrote during the conversation (visible in Settings → Memory). (5) Visual context flag: a boolean for whether a screenshot was attached to a turn (the screenshot itself is NOT stored — only that one was captured). All five live in PostgreSQL encrypted with AES-256-GCM, GDPR-aligned infrastructure, EU residency.
Three benefits worth the storage cost. (1) Recall: "what did we discuss last Tuesday about Q3 planning" only works if Jarvis has the history. Episodic memory (one of the four memory layers) is built on top of conversation history. (2) Cross-device sync: start a conversation on your Mac, continue it on Windows. Both devices read from the same server-side history. (3) Context for follow-ups: within a long task spanning multiple conversations over days, Jarvis can reference earlier turns when relevant. Without history storage, every conversation would start from scratch. The tradeoff is that history is sensitive data — Jarvis treats it accordingly with encryption and user control. Free tier and paid tiers all get history; you can disable it entirely if you prefer ephemeral chats only.
Privacy & data
