Your first chat¶
After the wizard, you land in a chat with Pluma — the built-in character. She knows she's the software you're running and is the safest starting point: ask her how anything works.
Sending a message¶
Type into the composer at the bottom and press Enter. The model streams its reply back; click Stop to abort generation mid-stream.
Shift-Enter adds a newline without sending.
Switching characters¶
Open the drawer (☰ on mobile, sidebar on desktop) → Characters tab. Tap a card to resume your most recent chat with that character; tap the + to start a fresh one.
Drop your own Tavern PNGs into <datadir>/characters/ and they show up automatically. See Characters & cards.
Starting a fresh chat¶
- With the same character: the + icon on the character row.
- Direct (no character, raw model): the home logo at the top of the sidebar. The assistant's name in chat reads as the active model id (e.g.
qwen2.5-32b-instruct) so you can tell it apart from a character chat.
Where the chat lives¶
Each conversation persists to <datadir>/conversations/<id>.json, encrypted at rest with AES-256-GCM. The key lives in your OS keyring (Keychain on macOS, libsecret on Linux, Credential Manager on Windows). Reload the page and conversations come back; restart the binary, same.
Per-conversation persona¶
If you've set up a persona (Settings → Personas), pick it from the as: … dropdown at the top of the chat. The persona's name fills {{user}} in card prose and lands explicitly in the system prompt so the assistant addresses you by name. See Personas.
What to try next¶
- Set a per-character sampler — different models like different temperatures. Samplers & presets.
- Generate an image — wand button in the composer. Image generation.
- Speak the reply aloud — Speak button in the message action menu. Voice & TTS.
- Reach Pluma from your phone — Multi-device access (Tailscale).