CharacterVault 1.4.2
A patch on 1.4.1: Agent chat is easier to follow after a write, and Settings plus live streams leak less work when you leave.
Try it: https://vault.charactervault.app
Staging: https://staging.charactervault.app
Docs: https://vault.charactervault.app/docs/
Highlights
Agent chat
- Jump from a write. Click a successful write line to open that field, greeting, or lorebook entry in the workspace.
- Write-turn recap. Speech from a turn that also called tools stays on the message. If there was no speech, Applied N writes sits above the tool list.
- Native vs XML is a small header chip (hidden on the narrowest screens). Hover for whether this model is on native function calling or XML after a tools 400.
- Busy labels use English phrases (
Updating field) instead of snake_case tool names.
Settings and live streams
- NanoGPT sign-in stops if you close the popup or the settings panel. The button does not keep spinning.
- Model catalogs stay out of saved AI config. They are UI cache only, not written into IndexedDB or the editor
aiConfig. - Closing Settings aborts in-flight catalog and account-overview fetches, ignores stale draft loads, and drops document keydown / toast timers that used to outlive the panel.
- Orion still stores the full reply for history, but live UI only renders a capped tail so token-by-token updates do not keep unbounded strings in React.
- Agent Stop cancels the live speech/thinking flush timer. Tool fences are stripped once when a row commits, not on every chat flush. Native tool-call JSON is assembled with chunked strings instead of quadratic concatenation.
- Leaving chat (New chat / unmount) clears the ask-delay and code-copy timers so a finished wait cannot setState on a gone panel.
Upgrade
Client-only app. Open the hosted site or rebuild locally; no server migration.
bash
git clone https://github.com/spaceman2408/CharacterVault
cd CharacterVault
npm install
npm run devLocal data from 1.4.1 keeps working. Hard refresh if the UI looks stale after deploy. If a huge model catalog was saved into AI settings on an older build, Save Settings on this version strips it.