Select a conversation
Answer with this model (Live):
Import from the Android app
Paste the JSON you exported from the phone app
(openrouter.dialogs.v1 and/or
openrouter.batches.history.v1). Either paste each list
separately, or a whole object containing those keys — both work.
⚡ JSONL batch (OpenRouter async)
Paste a .jsonl file — works with OpenRouter/OpenAI,
Google Vertex AI ({"request":{"contents":[...]}}),
and AWS Bedrock ({"recordId":"…","modelInput":{…}}) line
formats. Each line is sent to the batch model; answers land in a new
conversation.
🔥 Cache keep-alive pings
Warm-up requests for dialogs where you pressed 🔥 Cache (one near-empty ping every 45 min). They never appear as messages — verify them here.
⚙ Provider API keys
Saved on the server (SQLite) and applied immediately — no restart needed. Leave a field blank to keep its current value.
Enables the "Web search" toggle in the chat (internet lookups).
Prompt cache
How long reused prompt text stays cached for OpenRouter/Anthropic calls.
Cache keep-alive
After the last chat, near-empty pings (every 45 min) refresh the 1-hour cache at ~10% of input cost, keeping replies cheap for longer.
Google Vertex AI (optional — not currently in use)
AWS Bedrock (optional — not currently in use)
Account & pairing. This instance has a portable identity — paste the pairing code on any new device (phone app → Sync card) and it pairs without the password and pulls the whole history. Works even after the server gets a new address.
Client accounts. Create an isolated account for another person/device: it gets its own pairing code and sees only its own dialogs — never yours and never the provider keys above. Logins must be unique and every client gets a mandatory password — they log in with Login + password (the master password never opens their account). Creating one requires the master password.
Server migration. Download every key above as one file before shutting down this server, then restore it on the new one — no need to retype anything.