Docs

Appearance

What's New ​

Quick overview of recent updates to CharacterVault.

October 2026 ​

1.8.4 ​

CharacterVault 1.8.4 lets you pin and rank OpenRouter hosts, keep prompts private or stick to free models with new OpenRouter Options, keep Exacto on for Agent tool calls, and merge a library lorebook into a character's own entries, plus a fix for typing spaces in book names.

  • OpenRouter host control: pin a host per model under Provider, set Settings → AI Config → OpenRouter Options → Host priority, and see which host answered in a reply's stats
  • OpenRouter privacy and model filters: No training on prompts, Zero data retention only, and Free models only in OpenRouter Options, all off by default
  • Exacto for the Agent: OpenRouter Options → Always use Exacto for the Agent keeps :exacto on Agent tool calls with any host priority
  • Merge a library book (Character → Lorebook → Attach): Merge adds the book's entries after the character's own and unlinks the book

Release 1.8.4 → · AI Setup → · Lorebook Vault →

1.8.3 ​

CharacterVault 1.8.3 adds clickable variable chips to every prompt template, keeps linked lorebooks in sync both ways, and makes Settings ask before throwing away unsaved changes, plus Paste from Clipboard on import and a round of Settings and lorebook fixes.

  • Prompt variable chips in Settings → Prompts and Settings → Creation Studio → Generation Prompts: click a chip to insert its variable at the cursor or find it if already used, with a hover explanation and red for a missing required one
  • Two-way lorebook sync (character → Lorebook → Attach): saved edits on a linked character update the library book and every other linked character, and Open in vault takes an Auto snapshot first
  • Settings that keep your changes: closing with unsaved edits asks Keep editing or Discard, and saving waits until settings load

Release 1.8.3 → · AI Creation Studio → · Lorebook Vault →

September 2026 ​

1.8.2 ​

CharacterVault 1.8.2 makes inline AI results easy to retry, time, and reuse, speeds up search, font size, image links, and name macros in the editor, and carries card lorebook options and JPEG or WebP images cleanly to SillyTavern, plus a bigger spellcheck dictionary and a round of editor fixes.

  • Inline AI results: ↻ Retry, Reject reopens Custom with your instruction, model and timing pills, a cut-off warning, Rewrite / Insert on Custom, and recent instructions with ↑
  • Faster editing: Ctrl+H replace and F3 / Ctrl+G in search, a live aA font slider with shortcuts, Ctrl+click image links, and Backspace to keep a plain char / user (Settings → Character Workspace → Name macros)
  • Clean round-trips to SillyTavern: card lorebook options go where SillyTavern reads them, older cards catch up when opened, and Export → PNG converts JPEG or WebP images

Release 1.8.2 → · Text Editor → · Import & Export →

1.8.1 ​

CharacterVault 1.8.1 brings snapshot diffs laid out like the Agent review, a one-click Orion / Agent switch with edit-and-resend in Orion, and a Creator Notes preview that can't quietly phone home, plus a large round of fixes to saving, import, export, and the lorebook editor.

  • Snapshot diffs in Snapshots: one card per change, side by side or inline, Restore per card, and one card per greeting and lorebook entry
  • One chat header: the chat bubble (Orion) and robot (Agent) at the left switch modes, and the pencil on your last Orion message edits and resends
  • Remote content: the Creator Notes preview loads only https images, fonts, and styles and warns when they come from another site (Settings → Character Workspace, on by default); images in chat replies are links

Release 1.8.1 → · Snapshots → · Creator Notes →

1.8.0 ​

CharacterVault 1.8.0 makes the AI Agent easier to steer, follow, and review: edit and resend your last message, @-mention fields and entries, see what each run is doing and what earlier runs changed, and check its edits in a clearer diff.

  • Steer the Agent in the chat: the pencil on your last message edits and resends, @ mentions a field, greeting, or entry, an empty chat suggests requests based on what the card is missing, and Continue picks up after the turn limit
  • Know what it is doing: the status line names the target, the Agent remembers what earlier runs changed and your review choices, and the tab title reads Agent finished when a run ends while you are away
  • Clearer review diffs with Settings → Character Workspace → Review agent edits before applying: side by side or inline, word highlights, and Show N unchanged lines folds

Release 1.8.0 → · AI Agent →

1.7.5 ​

Clearer agent-review greeting diffs and custom tags in Creation Studio gender rules.

  • Agent review labels New alternate greeting / Deleted alternate greeting rows and shows Greetings-list rewrites as per-greeting blocks with counts and New / Removed badges; text diffs align by paragraph and line
  • Feeling Lucky draws your custom tags, and custom tags gray out by gender opposition like built-ins in Creation Studio → Tags Mode
  • Thinking and chat autoscroll follow streams reliably and release when you scroll up

Release 1.7.5 → · AI Agent → · AI Creation Studio →

1.7.0 ​

Customizable AI toolbar, roleplay highlight colors, and editor quality-of-life.

  • Reorder, hide, or re-add built-in toolbar buttons in Settings → Prompts → Toolbar Buttons; extras collapse into More
  • Create custom buttons with your own label, icon, color, prompt template, and model
  • "Dialogue", narration, and *actions* get distinct colors (on by default); recolor under Character Workspace → Roleplay colors and Name macro colors
  • Line highlight only with no selection, prefilled search + Ctrl+H, font-size shortcuts, Char/USER macro expansion, split-preview undo retention

Release 1.7.0 → · Editor →

1.6.0 ​

Full vault backup and restore, plus editor quality-of-life.

  • Library Backup ZIP now includes cards, standalone lorebooks, and settings; import the ZIP to restore (copies of content; settings replace on this device)
  • Settings → Backup for config-only export/import, with optional API keys
  • Header save status, live section counts, editable Extensions JSON, Ctrl+S / ⌘+S
  • Duplicate greetings (and reorder) and lorebook entries

Release 1.6.0 → · Import & Export → · Editor →

1.5.0 ​

AI Creation Studio overhaul and an optional Agent review diff.

  • Custom generation prompts, skip First Message / Examples, Dynamic and Kink & Fetish tags, hide NSFW, favorites / recents / custom tags
  • Go Back stays on screen during and after generation
  • Review agent edits stages writes into a word-level Original / Agent diff until you apply or discard
  • Settings split: Creation Studio vs Character Workspace; Settings from the vault header

Release 1.5.0 → · AI Creation Studio → · AI Agent →

August 2026 ​

1.4.5 ​

Orion and Agent chats stay with the card or lorebook you opened them on.

  • Separate saved threads for Orion and Agent, per character and per standalone lorebook
  • New chat confirms, then clears only that panel; leaving or closing the panel keeps the transcript
  • Load earlier messages pages older turns; chats are vault-local, not part of PNG/JSON export
  • Closing Ask AI unmounts character chat so a switch cannot leak the previous thread

Release 1.4.5 → · AI Assistant → · AI Agent →

1.4.2 ​

Agent write recap and jump-to-target, plus leaner Settings and live streams.

  • Click a successful write line to open that field, greeting, or lorebook entry
  • Speech stays on write turns; otherwise Applied N writes above the tool list
  • Header chip shows Native or XML; busy text is plain English
  • NanoGPT sign-in stops if you close the popup; model catalogs stay out of saved config
  • Closing Settings and leaving chat abort fetches and timers that used to linger

Release 1.4.2 → · AI Agent →

1.4.1 ​

Agent chat polish, a rewritten tutorial, and a way to try upcoming builds.

  • Composer stays usable during an Agent run; Agent writing is in the workspace header
  • Assistant text streams after thinking; retry clears error bubbles on the message
  • If a model rejects native tools, that model stays on XML for the rest of the session
  • First-run walkthrough matches the current product (replay from the vault ?)
  • On production, a dismissible header strip links to staging.charactervault.app

Release 1.4.1 → · AI Agent →

Agent (1.4.0) ​

A second chat next to Orion that fills and revises the open character or standalone lorebook.

  • Toggle Agent in the Ask AI header (character workspace or lorebook vault), or set Settings → Studio → Chat panel to open there
  • Optional custom context as source notes; it reads fields and entries with tools
  • Snippet edits match existing text; tool calls show as short color-coded lines
  • Writes once when the run finishes (snapshot first). Stop, then empty Send, retries
  • Own model on Settings → Prompts → Agent (Default follows AI Config)
  • Greeting 1 is the first alternate, same as the editor
  • History stays lighter (metadata index, opened-card snapshot last); long chats keep the last 100 messages
  • NanoGPT cache routing — Cache-capable provider routing toggle (Settings → AI Config → NanoGPT Options) sends caching: true to route to a cache-capable provider for lower cost and latency (sticky by default; fails if none serves the model)

Release 1.4.0 → · AI Agent → · AI Setup → NanoGPT Options · Snapshots

Provider usage cards (1.3.1) ​

  • Synthetic — new preset. Model list prefers syn: aliases; Synthetic Usage shows five-hour request and weekly credit quotas
  • OpenRouter — dedicated adapter plus OpenRouter Usage: this key’s spend (today / week / month / all time), optional spending cap, and free-tier :free rate limits
  • API key field — masked text, not a password input, so browsers are less likely to offer to save Settings as a login

Release 1.3.1 → · AI Setup · Synthetic Usage · OpenRouter Usage

Lorebook Vault & recursion map (1.3.0) ​

  • Lorebook library on the home vault: create, search, import, export, duplicate, delete standalone World Info books
  • Full book workspace with shared lorebook editor, Orion chat, and snapshot history
  • Link library books to characters: Open in vault updates the library book from the character; edits in the Lorebooks workspace update every linked character. Export still uses the lorebook on the card.
  • ST field coverage for common world-info options (selective logic, recursion flags, depth/role, probability, book scan depth / token budget)
  • Content-first editor: keys, Enabled / Constant, and the content editor stay on screen. Options and Attach are header buttons that open over the editor (not stacked above it). Works on phones: list and entry swap, Attach stays reachable from the list
  • Recursion map rebuilt: fullscreen web view of every unlock path in the book, with pan/zoom, hover spotlight, a list fallback for big books, per-entry flag toggles, and staged bulk edits you review before applying
  • Field ? help aligned with SillyTavern World Info docs

Release 1.3.0 → · Lorebook Vault · Lorebook Editor · Recursion map · Map guide

CharacterVault 1.2.1 ​

  • Markdown image links — ![](https://…) highlighted in the editor; optional click-to-open with a safety warning (Settings → Studio → Editor links)
  • Name macros — and get distinct syntax colors
  • Spellcheck — skips Markdown image constructs; hyphenated compounds pass when each segment is valid

Release 1.2.1 → · Editor → Markdown image links · Spellcheck

CharacterVault 1.2.0 ​

  • Custom Context — free-text notes for the open character (AI Context panel), optional for Orion and the AI toolbar, vault-local only
  • Lorebook — Enable All / Disable All for context sits next to the entry list
  • Sampler — modern Creative / Balanced / Factual presets; calmer new-install defaults (streaming & reasoning on, 8K context)

Release 1.2.0 → · AI Context → Custom Context · Sampler Settings

Custom Context (AI Context panel) ​

Paste free-text notes for the current character and include them in AI work without putting them on the card.

  • AI Context → Custom — Add / edit in a modal, toggle include, or remove
  • Orion + AI toolbar — Same block feeds chat and editor ops (when enabled)
  • Vault-local only — Stored per character in the browser; not in PNG/JSON export or SillyTavern fields
  • Usage meter — Custom tokens count toward the panel estimate; soft warnings for large pastes

AI Context → Custom Context · Orion

July 2026 ​

Orion (1.1.0) ​

  • Smoother streaming on long replies; stop mid-reply keeps what Orion already wrote
  • Resizable desktop chat panel; scroll follows the stream until you scroll up
  • Delete a message (with confirm) and trim everything after it

AI Assistant → · Release 1.1.0 →

Per-Prompt Model Routing ​

Each AI toolbar prompt can use its own endpoint and model, not only the global AI Config choice.

  • Settings → Prompts — Expand Enhance, Fix, Rephrase, etc., and set Model for this prompt
  • Any preset — Nano-GPT, Synthetic, OpenRouter, Minimax, LM Studio, or a custom URL you already saved a key for
  • Default stays simple — Leave an op on Default (AI Config) to keep using your global model
  • Examples — Fast model for Fix, a stronger model for Rephrase, local LM Studio for experiments
  • Lorebook ✨ keys — Follow the Custom prompt mapping when set

Orion chat and AI Creation Studio still use the global AI Config model. The Agent has its own mapping on the Prompts tab. Sampler / streaming / reasoning stay global.

Model and provider pickers (AI Config and Prompts) open as mobile-friendly sheets so search is easy on phones.

AI Setup → Prompts Tab · Editor → model routing

AI Toolbar Ghost Preview ​

Toolbar AI edits (Enhance, Rephrase, Custom, and the polish tools) no longer dump a huge result panel over your writing. The suggestion appears in the editor as ghost text at the selection — what you see is what Accept inserts.

  • In-place ghost — Streams into the locked span with a soft glow; soft-pulses when ready
  • Compact chrome — Status, Accept / Reject, and optional thinking stay in a thin strip; thinking is collapsed by default
  • Clean decisions — Typing is paused while a suggestion is open so Accept is one undo step; saves wait until you Accept, Reject, or cancel
  • Toast on Accept — A short success toast confirms the edit (no jarring green highlight flash)
  • Shortcuts — Ctrl+Enter / ⌘+Enter to Accept, Escape to Reject or Stop

Text editor → AI toolbar

NanoGPT Balance & Subscription ​

When the Nano-GPT preset is selected, a NanoGPT Account card in Settings → AI Config shows:

  • Balance (USD and Nano)
  • Subscription status (active, grace, or not active)
  • Weekly input tokens and other quota windows when NanoGPT reports them

Refresh is rate-limited. Closing and reopening Settings soon after reuses the last result so the service isn’t hit again. The card keeps a stable size while loading so the rest of the settings panel doesn’t jump.

Subscription status

Official hosted app and localhost need no worker. Only self-hosted production may need a small proxy for weekly tokens / sub status — see NanoGPT Usage Proxy. Balance works everywhere without a proxy.

AI Setup → NanoGPT Account

Reasoning Effort Levels & Compatibility ​

Settings → AI Config → Advanced Options now offers a fuller Reasoning Effort ladder when Enable reasoning is on: Minimal, Low, Medium, High, Extra high, and Max.

  • GPT-style models often use Minimal through High, with Extra high as the peak
  • Many SOTA thinking models (DeepSeek V4, GLM-5.x, Kimi, and similar) mainly use High and Max
  • If a provider rejects an effort value or some optional sampler knobs, CharacterVault adjusts and retries instead of failing the whole request, and remembers what that model accepted for the session

Reasoning Effort guide →

Larger Context Windows & Smarter AI Budgeting ​

AI work with big cards is much more reliable. You can raise the context window far past the old 128K cap, and large lorebooks actually make it into the request.

  • Up to 1M context — Settings → Sampler → Context Length includes 256K, 512K, and 1M presets, plus Custom… for any value from 4K to 1M
  • Lorebook & greetings fill the window — Selected lorebook entries (and long greeting lists) are packed in piece by piece so you get as much context as the budget allows, instead of almost none
  • Safer requests — CharacterVault is more careful about staying under your Context Length and Max Tokens, which cuts down on “prompt too long” errors from providers

Sampler settings →
AI Context panel →

Vault Library Upgrades ​

The home library is built for bigger collections.

  • Backup — Download a ZIP of every character (PNG when there is an image, JSON otherwise)
  • Bulk import — Import many cards at once, including drag and drop onto the library
  • Sort — Switch between Name and Recent (preference is remembered)
  • Search — Find cards by name or tags
  • Token estimates — Each card shows active / total estimates; hover for exact counts
  • Quick export — PNG or JSON from the card actions without opening the editor

Vault organization →
Import & export →

Smaller Improvements ​

  • API key field — Less likely to trigger browser password-manager prompts when pasting a key in AI settings

June 2026 ​

In-Editor Spellcheck ​

A Hunspell-backed spellchecker — the same engine LibreOffice uses — now runs inside the character editor. Misspelled words get a wavy-red underline; hover one to see quick-fix suggestions, Ignore word, or Add to dictionary.

  • Affix-aware — running, ran, and runs all pass without you whitelisting anything
  • Smart ignores — Code fences, placeholders, numbers, URLs/emails, and ALL-CAPS acronyms are skipped automatically
  • Personal dictionary & per-user ignore list — Add good-but-uncommon words (character names, neologisms) once and they stick across all your cards
  • Toggle in Settings — Disable in Settings → Studio → Spellcheck if you don't want it
  • Offline-friendly — The English dictionary is fetched once on first use and cached locally so it works without a connection

Performance

Spellchecking is debounced and viewport-scoped, so even large lorebooks stay responsive.

Learn more →

NanoGPT Sign-In (PKCE) with Auto Model Fetch ​

You can now sign in to NanoGPT directly from Settings → AI Config — no need to copy an API key from your dashboard. The flow uses OAuth with PKCE (Proof Key for Code Exchange), so your password never leaves NanoGPT's site and no client secret is stored in CharacterVault.

  • One-click sign-in — A new Sign in with NanoGPT button appears next to the API Key field when the Nano-GPT preset is selected.
  • Auto-fetch models — On successful sign-in, your available models are fetched automatically and the model dropdown is populated for you.
  • Per-URL caching — The auto-fetch uses the same per-base-URL cache as manual fetches, so reopening the settings panel loads instantly.

Browser popups must be allowed

The flow opens a new window/tab to NanoGPT. Allow popups for this site if your browser blocks them.

Mobile

Works on Chrome for Android. Other mobile browsers may not relay the auth code back to the app — if sign-in doesn't complete, paste an API key manually.

Learn more →

Customizable Section Tabs ​

The character editor has 18 section tabs by default — more than most people need. A new Sections tab in Settings lets you hide the tabs you don't use, reorder the rest however you like, and reset to the default order at any time.

  • Show or hide — Toggle any section tab on or off with the eye icon
  • Reorder — Move tabs up and down with arrow buttons
  • Reset to Defaults — Restore the original layout in one click
  • Mobile-friendly — Arrow buttons work great on touchscreens; the mobile dropdown respects your custom order and visibility

Learn more →

May 2026 ​

AI Creation Studio ​

The new AI Creation Studio lets you generate complete character cards from scratch using AI. Describe your concept or select from curated tags, and the AI creates a name, description, first message, and example dialogue in one flow.

Features:

  • Write Mode — Free-form concept descriptions
  • Tags Mode — Visual tag selection across 6 categories (Identity, Personality, Role, Genre, Tone, Appearance)
  • Feeling Lucky — Random tag generation with visual vortex animation
  • Live Preview — Edit generated fields in real-time
  • Field Regeneration — Retry or regenerate individual fields

Learn more →

Model Caching in Settings ​

The Character Settings panel now caches available AI models and providers, so the model list loads instantly on repeat visits. No more waiting for your model provider to respond every time you switch back.

Configure your AI provider →

Response Performance Stats ​

AI responses now show performance metrics when complete — look for TTFT (time to first token) and T/S (tokens per second) in the result header. Available in both the AI Assistant chat and the AI Toolbar result panel.

AI Assistant →
Text Editor →

Mobile Copy Buttons ​

The copy button on chat messages now appears on mobile devices, so you can quickly grab AI-generated text from your phone or tablet.

Adjusted Sampler Ranges ​

The character sampler parameter sliders (temperature, top-p, etc.) now have a wider and more sensible range, giving you finer control over your generations.

Sampler Settings →

April 2026 ​

Lorebook Import & Export ​

You can now import and export lorebooks independently from the Lorebook Editor.

  • Export — Save your character's lorebook as a JSON file for sharing or backup
  • Import — Bring in lorebook data from JSON files exported by SillyTavern and other tools

Learn more →

Multi-line Input in AI Panels ​

The AI chat input and toolbar custom instructions now support multi-line text. Press Shift+Enter to insert line breaks when writing longer prompts or detailed instructions.

  • AI Assistant chat — Multi-line input for complex queries to Orion
  • AI Toolbar — Multi-line custom instructions for text operations

AI Assistant →
Text Editor →

Smaller Improvements ​

  • Lorebook editor — Empty character books are filtered out during export
  • History modal — Prevented content flash during modal entrance
  • Character images — Implemented content-addressed storage for deduplication
  • Lorebook UI — Swapped name and comment fields to match SillyTavern convention

See Also ​

Privacy

Copyright © 2026 spaceman2408