SillyTavern

AI

LLM Frontend for Power Users.

Latest 1.18.0 · by SillyTavernWebsiteSillyTavern/SillyTavern

Release activity

Release activity — 8 releases across 8 days in the last year. Each cell is one day; darker means more releases that day. Older weeks are hidden at this screen width.
MayJunJulAug
SundayNo releases on Apr 26, 20261 release on May 3, 2026No releases on May 10, 2026No releases on May 17, 2026No releases on May 24, 2026No releases on May 31, 2026No releases on Jun 7, 2026No releases on Jun 14, 2026No releases on Jun 21, 2026No releases on Jun 28, 2026No releases on Jul 5, 2026No releases on Jul 12, 2026No releases on Jul 19, 2026No releases on Jul 26, 2026No releases on Aug 2, 2026No releases on Aug 9, 2026
MondayNo releases on Apr 27, 2026No releases on May 4, 2026No releases on May 11, 2026No releases on May 18, 2026No releases on May 25, 2026No releases on Jun 1, 2026No releases on Jun 8, 2026No releases on Jun 15, 2026No releases on Jun 22, 2026No releases on Jun 29, 2026No releases on Jul 6, 2026No releases on Jul 13, 2026No releases on Jul 20, 2026No releases on Jul 27, 2026No releases on Aug 3, 2026No releases on Aug 10, 2026
TuesdayNo releases on Apr 28, 2026No releases on May 5, 2026No releases on May 12, 2026No releases on May 19, 2026No releases on May 26, 2026No releases on Jun 2, 2026No releases on Jun 9, 2026No releases on Jun 16, 2026No releases on Jun 23, 2026No releases on Jun 30, 2026No releases on Jul 7, 2026No releases on Jul 14, 2026No releases on Jul 21, 2026No releases on Jul 28, 2026No releases on Aug 4, 2026No releases on Aug 11, 2026
WednesdayNo releases on Apr 29, 2026No releases on May 6, 2026No releases on May 13, 2026No releases on May 20, 2026No releases on May 27, 2026No releases on Jun 3, 2026No releases on Jun 10, 2026No releases on Jun 17, 2026No releases on Jun 24, 2026No releases on Jul 1, 2026No releases on Jul 8, 2026No releases on Jul 15, 2026No releases on Jul 22, 2026No releases on Jul 29, 2026No releases on Aug 5, 2026No releases on Aug 12, 2026
ThursdayNo releases on Apr 30, 2026No releases on May 7, 2026No releases on May 14, 2026No releases on May 21, 2026No releases on May 28, 2026No releases on Jun 4, 2026No releases on Jun 11, 2026No releases on Jun 18, 2026No releases on Jun 25, 2026No releases on Jul 2, 2026No releases on Jul 9, 2026No releases on Jul 16, 2026No releases on Jul 23, 2026No releases on Jul 30, 2026No releases on Aug 6, 2026No releases on Aug 13, 2026
FridayNo releases on May 1, 2026No releases on May 8, 2026No releases on May 15, 2026No releases on May 22, 2026No releases on May 29, 2026No releases on Jun 5, 2026No releases on Jun 12, 2026No releases on Jun 19, 2026No releases on Jun 26, 2026No releases on Jul 3, 2026No releases on Jul 10, 2026No releases on Jul 17, 2026No releases on Jul 24, 2026No releases on Jul 31, 2026No releases on Aug 7, 2026No releases on Aug 14, 2026
SaturdayNo releases on May 2, 2026No releases on May 9, 2026No releases on May 16, 2026No releases on May 23, 2026No releases on May 30, 2026No releases on Jun 6, 2026No releases on Jun 13, 2026No releases on Jun 20, 2026No releases on Jun 27, 2026No releases on Jul 4, 2026No releases on Jul 11, 2026No releases on Jul 18, 2026No releases on Jul 25, 2026No releases on Aug 1, 2026No releases on Aug 8, 2026No releases on Aug 15, 2026

8 releases in the last year

Changelog

1.18.0

Latest
Added 14
  • Added Cloudflare Workers AI and MiniMax as Chat Completion sources
  • KoboldCpp added forwarding of reasoning effort when running as a Custom Chat Completion source
  • Tool Calling added configurable tool calling recursion limit and enabled interleaved thinking for Custom sources
  • Text Generation WebUI added Adaptive-P controls
  • NanoGPT added provider selection and model sorting
  • Added ability to view remaining balance for OpenRouter and NanoGPT
Changed 5
  • KoboldCpp grammar state will be preserved when using a Continue option
  • Text Completion impersonation requests use a Last User Message prefix at the end of the prompt if configured
  • Moved HTTP error pages and user.css file from /public to /data to support immutable setups
  • Disabled HTTP keep-alive by default to restore old Node 18 behavior, can be enabled with config
  • Increased the length of password reset code to 6 characters to guard against brute-force attacks
Removed 1
  • Removed post-install script, config migration is now handled by the app or a dedicated npm run init command

SillyTavern 1.18.0

Important news

Read the maintainers statement regarding a recent security incident involving the "Bot Browser" third-party extension and learn how to stay safe: https://github.com/SillyTavern/SillyTavern/discussions/5592

Backends
  • Added Cloudflare Workers AI and MiniMax as Chat Completion sources.
  • KoboldCpp: Grammar state will be preserved when using a "Continue" option.
  • KoboldCpp: Added forwarding of reasoning effort when running as a Custom Chat Completion source.
  • Tool Calling: Added a configurable tool calling recursion limit; enabled interleaved thinking for Custom sources.
  • Text Completion: Impersonation requests use a "Last User Message" prefix at the end of the prompt (if configured).
  • Text Generation WebUI: Added Adaptive-P controls.
  • NanoGPT: Added provider selection and model sorting.
  • Added ability to view remaining balance for OpenRouter and NanoGPT.
  • Enhanced support for new models: DeepSeek v4, GPT 5.4 and 5.5, Gemma 4, GLM-5V-Turbo, Claude Opus 4.7.
Server & Security
  • Removed post-install script, config migration is now handled by the app or a dedicated npm run init command.
  • Added npm configuration to prevent execution of package scripts during installation.
  • Moved HTTP error pages and user.css file from /public to /data to support immutable setups.
  • Disabled HTTP keep-alive by default to restore old Node 18 behavior, can be enabled with config.
  • Added rate limiting to the basic authentication flow to mitigate brute-force attacks.
  • Added configuration options to choose which headers can be used for forwarded IP detection to prevent spoofing.
  • Added a private address whitelist to prevent SSRF attacks. See the documentation on how to enable and configure: Private Address Whitelist.
  • Added an IP whitelist for SSO trusted proxies to prevent authentication bypass.
  • Added invalidation of session cookies on password change to prevent session hijacking.
  • Increased the length of password reset code to 6 characters to guard against brute-force attacks.
  • Implemented PKCE challenge in OpenRouter OAuth flow for more secure key exchange.
UI/UX
  • Improved swipe picker: mobile requires a long press on swipe counter to open; added buttons to expand or copy the swipe text.
  • "Click to Edit" mode now also applied to reasoning blocks.
  • Welcome Screen: Number of recent chats can be configured.
  • Streamed requests now can show an error message in the console if the request fails.
STscript
  • Added commands for persona management: /persona-create, /persona-update, /persona-delete, /persona-duplicate, and /persona-get.
  • Added a command to force update the Prompt Manager's prompt list: /pm-render.
  • Added a command to get the state of the regex script: /regex-state.
  • Added a command to set fallback expression: /expression-fallback.
  • Added a command to generate a streamed response with a connection profile: /profile-genstream.
Extensions
  • Assets list now groups extensions by "Official" or "Community" categories.
  • Added an additional confirmation prompt when installing third-party extensions (can be disabled).
  • Supported extensions can use a secret-id from connection profiles when making an LLM request.
  • Extensions list now shows the extension's author name resolved from the git remote URL.
  • Vector Storage: Added Workers AI source; added a toggle to keep vectors for hidden messages; added retry logic to summary generation.
  • Image Generation: Added Workers AI source; generation can now be cancelled by pressing a button in the status toast.
  • Image Captioning: Added support for macros in the caption prompt.
  • TTS: "Skip code blocks" no longer ignores lines that start with 4 spaces (legacy code block syntax); "disabled" voice now shows a toast only once per character.
Bug Fixes
  • Fixed text edit flow in Firefox on mobile.
  • Fixed welcome screen chat pins not updating on chat renaming.
  • Fixed character list filters being stuck on app initialization.
  • Fixed application of instruct formatting to /genraw requests.
  • Fixed model routing to sd.cpp API in Image Generation logic.
  • Fixed validation of image URLs generated with Z.AI API.
  • Fixed vectors deletion for KoboldCpp when a message is deleted.
  • Fixed "Show More Messages" button triggering edit in "Click to Edit" mode.
  • Fixed max height of select-multiple elements in mobile layout.
  • Fixed server crash on empty messages when applying cache control parameters.
Community updates
New Contributors

Full Changelog: https://github.com/SillyTavern/SillyTavern/compare/1.17.0...1.18.0

View originalPermalink
How 1.18.0 went

1.17.0

Added 12
  • Claude backend now supports adaptive thinking mapped from Reasoning Effort with opt-in via config
  • OpenRouter backend added provider filtering based on selected model
  • OpenRouter added interleaved reasoning forwarding for tool-call continuations
  • SiliconFlow added API endpoint selection (Global / China)
  • Swipe Picker now provides a new interface to browse, jump, branch, and delete swipes
  • Backgrounds added virtual folders with grid view, breadcrumbs, and thumbnails
Changed 8
  • Requires Node.js 20 or higher
  • OpenRouter deactivated reasoning toggle with minimal reasoning effort now disables thinking (model-dependent)
  • Synchronized model lists for GPT, Claude, GLM, Gemini, and Grok Imagine
  • UI now displays a new splash screen design during app initialization
  • World Info renaming can optionally relink lorebooks across all characters
  • World Info unified chat, character and persona link buttons behavior with click to load and shift-click or long press to open a menu
  • Thumbnails improved APNG detection for thumbnail generation
  • Tokenization now uses byte length of a string for token estimation
Removed 1
  • xAI removed deprecated web search toggle

SillyTavern 1.17.0

Note: Requires Node.js 20 or higher. Need to update Node? See: https://docs.sillytavern.app/installation/updating/node/

Backends
  • Claude: Added adaptive thinking support mapped from Reasoning Effort (opt-in via config)
  • OpenRouter: Added provider filtering based on selected model
  • OpenRouter: Deactivated reasoning toggle with minimal reasoning effort now disables thinking (model-dependent)
  • OpenRouter: Added interleaved reasoning forwarding for tool-call continuations
  • SiliconFlow: Added API endpoint selection (Global / China)
  • xAI: Removed deprecated web search toggle
  • Synchronized model lists (GPT, Claude, GLM, Gemini, Grok Imagine)
Improvements
  • Swipe Picker: New interface to browse, jump, branch, and delete swipes
  • Backgrounds: Added virtual folders with grid view, breadcrumbs, and thumbnails
  • UI: New splash screen design during app initialization
  • World Info: Renaming can optionally relink lorebooks across all characters
  • World Info: Unified chat, character and persona link buttons behavior (click to load, shift-click or long press to open a menu)
  • Tags: Added automatic cleanup of orphaned folder tags
  • Thumbnails: Improved APNG detection for thumbnail generation
  • Tokenization: Token estimation now uses byte length of a string
  • Gallery: Added a shortcut to extensions ("Magic Wand") menu.
  • a11y: Added support for system-wide reduced motion and high contrast preferences
  • Reasoning: Added a new "Think XML" template without newlines in tags
  • Server: Added isomorphic-git as an alternative backend for installing extensions
  • Server: Added optional gzip compression for large request payloads
  • Server: Added config for disallowing full data backups
  • Server: Clear error messaging for port conflicts during startup
  • Server: Auto-create data root in standalone mode if not exists
  • Build: Improved webpack cache management to prevent stale caches
Macros
  • New macro engine enabled by default for new installs
  • Added:
    • {{charFirstMessage}} / {{greeting}} (with optional indexing for alternate greetings)
    • {{maxContextTokens}}, {{maxResponseTokens}}
    • {{allChatRange}}: provides the range of the entire chat for commands that accept ranges
  • Fixed {{summary}} prioritization of UI memory content
STscript
New Commands
  • Character management: /char-create, /char-update, /char-get, /char-delete, /char-duplicate, /tag-import
  • Generation control: /swipe, /regenerate
  • Reasoning block control: /reasoning-collapse, /reasoning-expand, /reasoning-toggle
  • Array utilities: /array-wrap, /array-unwrap
  • Loader: /loader-wrap, /loader-show, /loader-hide, /loader-stop
Enhancements
  • Extended /input, /popup, /buttons with placeholders, tooltips, icons
  • /imagine now supports gallery argument
  • /bgcol rebuilt for full theme palette generation using color theory
Removals
  • Removed deprecated commands: /lock and /bind (aliased to /persona-lock)
  • Deprecated format parameter for message-sending commands (use return)
Extensions
For developers
  • Added extension lifecycle hooks via manifest declarations
  • Exported SlashCommandEnumValue and generateRawData
  • Extensions can now process streamed text during tool-call chains
  • Popup System:
    • Added textarea support
    • Added placeholders, tooltips, icons
  • Action Loader:
    • New reusable loading overlay system
    • Supports stacking, toasts, and STscript integration
  • Added events:
    • PERSONA_CHANGED
    • TTS_JOB_STARTED
    • TTS_AUDIO_READY
    • TTS_JOB_COMPLETE
Vector Storage
  • Ollama: Migrated to /api/embed with batch support
  • SiliconFlow: Added as embeddings provider
  • Google: Added gemini-embedding-2-preview
Image Generation
  • Overriden image dimensions via the slash command are preserved for swiped images
  • Manual prompt refinement: edit or discard saved negative prompt and resolution
Bug Fixes
  • Fixed sanitization of chat import names
  • Fixed secrets being present in data backups when not intended
  • Fixed user data paths validation in route handlers
  • Fixed Claude multipart text block parsing (non-streaming)
  • Fixed tokenizer requests requiring model name (llama.cpp)
  • Fixed embedding requests for custom API paths (llama.cpp, vLLM)
  • Fixed stop string message clean-up for Chat Completion
  • Fixed OpenRouter cache-at-depth logic to exclude system messages
  • Fixed OpenRouter not respecting enableThoughtSignatures setting
  • Fixed OpenRouter payload field name (input_video → video_url)
  • Fixed KoboldCpp DRY sequence stringification
  • Fixed KoboldCpp and Tabby tokenization counting a BOS token
  • Fixed Escape key closing non-dismissable popups (double tap Escape to override)
  • Fixed tag visibility persistence in Tag Management
  • Fixed priority of closing side panels with Escape keys
  • Fixed APNG image format detection in thumbnail handling
  • Fixed duplicate image-generation toasts
  • Fixed tool-call error handling and reporting
  • Fixed typo in KoboldAI instruct template (OUPUT → OUTPUT)
Community updates
New Contributors

Full Changelog: https://github.com/SillyTavern/SillyTavern/compare/1.16.0...1.17.0

View originalPermalink
How 1.17.0 went

1.16.0

Added 20
  • NanoGPT backend now supports tool calling and reasoning effort
  • OpenAI and compatible backends now support audio inlining
  • Adaptive-P sampler settings for supported Text Completion backends
  • Pollinations backend now requires an API key to use
  • Tag filters for group members list
  • Background images can now save additional metadata like aspect ratio and dominant color
Changed 7
  • Improved naming pattern of branched chat files
  • Enhanced world duplication to use the current world name as a base
  • Improved performance of message rendering in large chats
  • Improved performance of chat file management dialog
  • Pollinations backend updated to a new API
  • Moonshot thinking type now mapped to Request reasoning setting in the UI
  • Synchronized model lists for Claude and Z.AI
Fixed 13
  • Path traversal vulnerability in several server endpoints
  • Server CORS forwarding being available without authentication when CORS proxy is enabled
  • Asset downloading feature now requires host whitelist match to prevent SSRF vulnerabilities
  • Basic authentication password containing a colon character not working correctly
  • Experimental macro engine being case-sensitive when checking for macro names
  • Experimental macro engine compatibility with the STscript parser

SillyTavern 1.16.0

Note: The first-time startup on low-end devices may take longer due to the image metadata caching process.

Backends
  • NanoGPT: Enabled tool calling and reasoning effort support.
  • OpenAI (and compatible): Added audio inlining support.
  • Added Adaptive-P sampler settings for supported Text Completion backends.
  • Gemini: Thought signatures can be disabled with a config.yaml setting.
  • Pollinations: Updated to a new API; now requires an API key to use.
  • Moonshot: Mapped thinking type to "Request reasoning" setting in the UI.
  • Synchronized model lists for Claude and Z.AI.
Features
  • Improved naming pattern of branched chat files.
  • Enhanced world duplication to use the current world name as a base.
  • Improved performance of message rendering in large chats.
  • Improved performance of chat file management dialog.
  • Groups: Added tag filters to group members list.
  • Background images can now save additional metadata like aspect ratio, dominant color, etc.
  • Welcome Screen: Added the ability to pin recent chats to the top of the list.
  • Docker: Improved build process with support for non-root container users.
  • Server: Added CORS module configuration options to config.yaml.
Macros

Note: New features require "Experimental Macro Engine" to be enabled in user settings.

  • Added autocomplete support for macros in most text inputs (hint: press Ctrl+Space to trigger autocomplete).
  • Added a hint to enable the experimental macro engine if attempting to use new features with the legacy engine.
  • Added scoped macros syntax.
  • Added conditional if macro and preserve whitespace (#) flag.
  • Added variable shorthands, comparison and assignment operators.
  • Added {{hasExtension}} to check for active extensions.
STscript
  • Added /reroll-pick command to reroll {{pick}} macros in the current chat.
  • Added /beep command to play a message notification sound.
Extensions
  • Added the ability to quickly toggle all third-party extensions on or off in the Extensions Manager.
  • Image Generation:
    • Added image generation indicator toast and improved abort handling.
    • Added stable-diffusion.cpp backend support.
    • Added video generation for Z.AI backend.
    • Added reduced image prompt processing toggle.
    • Added the ability to rename styles and ComfyUI workflows.
  • Vector Storage:
    • Added slash commands for interacting with vector storage settings.
    • Added NanoGPT as an embeddings provider option.
  • TTS:
    • Added regex processing to remove unwanted parts from the input text.
    • Added Volcengine and GPT-SoVITS-adapter providers.
  • Image Captioning: Added a model name input for Custom (OpenAI-compatible) backend.
Bug Fixes
  • Fixed path traversal vulnerability in several server endpoints.
  • Fixed server CORS forwarding being available without authentication when CORS proxy is enabled.
  • Fixed asset downloading feature to require a host whitelist match to prevent SSRF vulnerabilities.
  • Fixed basic authentication password containing a colon character not working correctly.
  • Fixed experimental macro engine being case-sensitive when checking for macro names.
  • Fixed compatibility of the experimental macro engine with the STscript parser.
  • Fixed tool calling sending user input while processing the tool response.
  • Fixed logit bias calculation not using the "Best match" tokenizer.
  • Fixed app attribution for OpenRouter image generation requests.
  • Fixed itemized prompts not being updated when a message is deleted or moved.
  • Fixed error message when the application tab is unloaded in Firefox.
  • Fixed Google Translate bypassing the request proxy settings.
  • Fixed swipe synchronization overwriting unresolved macros in greetings.
Community updates
New Contributors

Full Changelog: https://github.com/SillyTavern/SillyTavern/compare/1.15.0...1.16.0

View originalPermalink
How 1.16.0 went

1.15.0

Added 9
  • Introduce Experimental Macro Engine with support for nested macros, stable evaluation order, and improved autocomplete
  • Add Chutes as a Chat Completion source
  • Add `/message-role` and `/message-name` STscript commands
  • Add Chutes, MistralAI, Z.AI, ElevenLabs, and Groq as Speech Recognition STT sources
  • Add Chutes, Z.AI, OpenRouter, and RunPod Comfy as Image Generation inference sources
  • Add Z.AI as a Web Search source
Changed 18
  • NanoGPT now exposes additional samplers to UI
  • llama.cpp now supports model selection and multi-swipe generation
  • Synchronize model lists for OpenAI, Google, Claude, and Z.AI
  • Electron Hub now supports caching for Claude models
  • OpenRouter now supports system prompt caching for Gemini and Claude models
  • Gemini now supports thought signatures for applicable models
Fixed 6
  • Fix resetting the context size when switching between Chat Completion sources
  • Fix arrow keys triggering swipes when focused into video elements
  • Fix server crash in Chat Completion generation when invalid endpoint URL passed
  • Fix pending file attachments not being preserved when using Attach a File button
  • Fix tool calling not working with deepseek-reasoner model
  • Fix image generation not using character prefixes for brush message action
Deprecated 1
  • Legacy macro substitution will eventually be removed in favor of Experimental Macro Engine

SillyTavern 1.15.0

Highlights

Introducing the first preview of Macros 2.0, a comprehensive overhaul of the macro system that enables nesting, stable evaluation order, and more. You are encouraged to try it out by enabling "Experimental Macro Engine" in User Settings -> Chat/Message Handling. Legacy macro substitution will not receive further updates and will eventually be removed.

Breaking Changes
  1. {{pick}} macros are not compatible between the legacy and new macro engines. Switching between them will change the existing pick macro results.
  2. Due to the change of group chat metadata files handling, existing group chat files will be migrated automatically. Upgraded group chats will not be compatible with previous versions.
Backends
  • Chutes: Added as a Chat Completion source.
  • NanoGPT: Exposed additional samplers to UI.
  • llama.cpp: Supports model selection and multi-swipe generation.
  • Synchronized model lists for OpenAI, Google, Claude, Z.AI.
  • Electron Hub: Supports caching for Claude models.
  • OpenRouter: Supports system prompt caching for Gemini and Claude models.
  • Gemini: Supports thought signatures for applicable models.
  • Ollama: Supports extracting reasoning content from replies.
Improvements
  • Experimental Macro Engine: Supports nested macros, stable evaluation order, and improved autocomplete.
  • Unified group chat metadata format with regular chats.
  • Added backups browser in "Manage chat files" dialog.
  • Prompt Manager: Main prompt can be set at an absolute position.
  • Collapsed three media inlining toggles into one setting.
  • Added verbosity control for supported Chat Completion sources.
  • Added image resolution and aspect ratio settings for Gemini sources.
  • Improved CharX assets extraction logic on character import.
  • Backgrounds: Added UI tabs and ability to upload chat backgrounds.
  • Reasoning blocks can be excluded from smooth streaming with a toggle.
  • start.sh script for Linux/MacOS no longer uses nvm to manage Node.js version.
STscript
  • Added /message-role and /message-name commands.
  • /api-url command supports VertexAI for setting the region.
Extensions
  • Speech Recognition: Added Chutes, MistralAI, Z.AI, ElevenLabs, Groq as STT sources.
  • Image Generation: Added Chutes, Z.AI, OpenRouter, RunPod Comfy as inference sources.
  • TTS: Unified API key handling for ElevenLabs with other sources.
  • Image Captioning: Supports Z.AI (common and coding) for captioning video files.
  • Web Search: Supports Z.AI as a search source.
  • Gallery: Now supports video uploads and playback.
Bug Fixes
  • Fixed resetting the context size when switching between Chat Completion sources.
  • Fixed arrow keys triggering swipes when focused into video elements.
  • Fixed server crash in Chat Completion generation when invalid endpoint URL passed.
  • Fixed pending file attachments not being preserved when using "Attach a File" button.
  • Fixed tool calling not working with deepseek-reasoner model.
  • Fixed image generation not using character prefixes for 'brush' message action.
Community Updates
New Contributors

Full Changelog: https://github.com/SillyTavern/SillyTavern/compare/1.14.0...1.15.0

View originalPermalink
How 1.15.0 went

1.14.0

Added 13
  • Z.AI API and SiliconFlow as Chat Completion sources
  • Ability to attach multiple files and images to a single message
  • Ability to attach audio files to messages
  • Prompt audio inlining support for Gemini and OpenRouter
  • Gallery and list style switching for messages with attached media
  • Start Reply With to master import/export in Advanced Formatting
Changed 10
  • Default presets for Text Completion and Kobold Classic updated, legacy presets removed
  • Model lists updated for Gemini, Claude, OpenAI and Moonshot
  • OpenRouter providers list synchronized
  • VertexAI-specific safety categories added with OFF values for Gemini models
  • Advanced Formatting settings that do not work for Chat Completion API are now grayed out
  • Alternate greetings can now be reordered
Fixed 10
  • Quick Replies not auto-executing on new group chats
  • NanoGPT and DeepSeek models not saving to presets
  • Message editor closing when ending IME composition with Esc key
  • Edge cases in JSON schema conversion for Gemini models
  • Image caching on avatar updates in Firefox
  • Horde shared keys showing incorrect kudos value

SillyTavern 1.14.0

Breaking changes

Due to changes in the handling of attached media, chat files containing such media will not be backward compatible with versions prior to 1.14.0.

Third-party extensions that read or write media to chat messages may require updates to be compatible with this version. Please contact the extension authors for more information.

Backends
  • Added Z.AI API and SiliconFlow as Chat Completion sources.
  • Updated default presets for Text Completion and Kobold Classic, legacy presets are removed.
  • Updated model lists for Gemini, Claude, OpenAI and Moonshot.
  • Added VertexAI-specific safety categories with OFF values for Gemini models.
  • Synchronized OpenRouter providers list.
Improvements
  • Added the ability to attach multiple files and images to a single message.
  • Added the ability to attach audio files to messages.
  • Added prompt audio inlining support for Gemini and OpenRouter.
  • Can switch between gallery and list styles in messages with attached media.
  • Advanced Formatting settings that do not work for Chat Completion API are now grayed out.
  • Added Start Reply With to master import/export in Advanced Formatting.
  • Alternate greetings can now be reordered.
  • Added per-chat overrides for example messages and system prompts.
  • Markdown: Block quotes can be rendered when "Show tags" setting is enabled.
  • Text Completion: Empty JSON schema input no longer sends empty object in the request.
  • Text Completion: Added an option to lock sampler selection per API type.
  • Server: Invalid IP whitelist entries are now skipped and logged on startup.
  • Tags: Can be sorted by most used in the Tag Manager dialog.
  • Tags: Increased maximum number of tags that can be imported from a character to 50.
  • UX: Multiline inputs in popups will input a linebreak instead of submitting the popup.
  • UX: Show a confirmation when reloading/leaving the page while editing a message.
  • More complete support for characters import from BYAF archives.
  • World Info: Added a height limit to entry keys inputs.
Extensions
  • Git operations with extensions now timeout after 5 minutes of inactivity.
  • Regex: Preset regex are now applied before scoped regex.
  • Image Captioning: Added captioning for video attachments (currently Gemini only).
  • Image Generation: Added Veo (Google) and Sora 2 (OpenAI) models for video generation.
  • Vector Storage: Added OpenRouter and Electron Hub as provider options.
Bug fixes
  • Fixed Quick Replies not auto-executing on new group chats.
  • Fixed NanoGPT and DeepSeek models not saving to presets.
  • Fixed message editor closing when ending IME composition with Esc key.
  • Fixed edge cases in JSON schema conversion for Gemini models.
  • Fixed image caching on avatar updates in Firefox.
  • Fixed Horde shared keys showing incorrect kudos value.
  • Fixed error notifications for Gemini in non-streaming mode.
Community updates
New Contributors

Full Changelog: https://github.com/SillyTavern/SillyTavern/compare/1.13.5...1.14.0

View originalPermalink
How 1.14.0 went

1.13.5

Added 14
  • NanoGPT: Added reasoning content display
  • Electron Hub: Added prompt cost display and model grouping
  • UI: Added a user setting to enable fade-in animation for streamed text
  • UX: Added drag-and-drop to the past chats menu and the ability to import multiple chats at once
  • UX: Added first/last-page buttons to the pagination controls
  • UX: Added the ability to change sampler settings while scrolling over focusable inputs
Changed 6
  • Synchronized model lists for Claude, Grok, AI Studio, and Vertex AI
  • UI: Updated the layout of the backgrounds menu
  • UI: Hid panel lock buttons in the mobile layout
  • Persona: The persona description textarea can be expanded
  • Persona: Changing a persona will update group chats that haven't been interacted with yet
  • STscript: /genraw now emits prompt-ready events and can be canceled by extensions

SillyTavern 1.13.5

Backends
  • Synchronized model lists for Claude, Grok, AI Studio, and Vertex AI.
  • NanoGPT: Added reasoning content display.
  • Electron Hub: Added prompt cost display and model grouping.
Improvements
  • UI: Updated the layout of the backgrounds menu.
  • UI: Hid panel lock buttons in the mobile layout.
  • UI: Added a user setting to enable fade-in animation for streamed text.
  • UX: Added drag-and-drop to the past chats menu and the ability to import multiple chats at once.
  • UX: Added first/last-page buttons to the pagination controls.
  • UX: Added the ability to change sampler settings while scrolling over focusable inputs.
  • World Info: Added a named outlet position for WI entries.
  • Import: Added the ability to replace or update characters via URL.
  • Secrets: Allowed saving empty secrets via the secret manager and the slash command.
  • Macros: Added the {{notChar}} macro to get a list of chat participants excluding {{char}}.
  • Persona: The persona description textarea can be expanded.
  • Persona: Changing a persona will update group chats that haven't been interacted with yet.
  • Server: Added support for Authentik SSO auto-login.
STscript
  • Allowed creating new world books via the /getpersonabook and /getcharbook commands.
  • /genraw now emits prompt-ready events and can be canceled by extensions.
Extensions
  • Assets: Added the extension author name to the assets list.
  • TTS: Added the Electron Hub provider.
  • Image Captioning: Renamed the Anthropic provider to Claude. Added a models refresh button.
  • Regex: Added the ability to save scripts to the current API settings preset.
Bug Fixes
  • Fixed server OOM crashes related to node-persist usage.
  • Fixed parsing of multiple tool calls in a single response on Google backends.
  • Fixed parsing of style tags in Creator notes in Firefox.
  • Fixed copying of non-Latin text from code blocks on iOS.
  • Fixed incorrect pitch values in the MiniMax TTS provider.
  • Fixed new group chats not respecting saved persona connections.
  • Fixed the user filler message logic when continuing in instruct mode.
Community updates
New Contributors

Full Changelog: https://github.com/SillyTavern/SillyTavern/compare/1.13.4...1.13.5

View originalPermalink
How 1.13.5 went

1.13.4

Added 13
  • Google backend now supports the gemini-2.5-flash-image (Nano Banana) model
  • DeepSeek backend can pass sampling parameters to the reasoner model
  • NanoGPT backend enabled prompt cache setting for Claude models
  • OpenRouter backend added image output parsing for models that support it
  • Chat Completion added Azure OpenAI and Electron Hub as sources
  • Server added validation of hostnames in requests for improved security (opt-in)
Changed 4
  • Chat Completion requests that fail with code 429 will no longer be silently retried
  • Reasoning auto-parsed reasoning blocks are now automatically removed from impersonation results
  • UI updated the layout of the background image settings menu
  • TTS improved handling of nested quotes when using the Narrate quotes option
Fixed 3
  • Fixed request streaming for the Vertex AI backend in Express mode
  • Fixed erroneous replacement of newlines with <br> tags inside HTML code blocks
  • Fixed custom toast positions not being applied to popups

SillyTavern 1.13.4

Backends
  • Google: Added support for the gemini-2.5-flash-image (Nano Banana) model.
  • DeepSeek: Sampling parameters can be passed to the reasoner model.
  • NanoGPT: Enabled prompt cache setting for Claude models.
  • OpenRouter: Added image output parsing for models that support it.
  • Chat Completion: Added Azure OpenAI and Electron Hub as sources.
Improvements
  • Server: Added validation of hostnames in requests for improved security (opt-in).
  • Server: Added support for SSL certificates with a passphrase when using HTTPS.
  • Chat Completion: Requests that fail with code 429 will no longer be silently retried.
  • Chat Completion: Inline Image Quality control is now available for all compatible sources.
  • Reasoning: Auto-parsed reasoning blocks are now automatically removed from impersonation results.
  • UI: Updated the layout of the background image settings menu.
  • UX: Ctrl+Enter will send a user message if the text input is not empty.
  • Added Thai locale and made various improvements to existing locales.
Extensions
  • Image Captioning: Added custom model input for Ollama, updated the list of Groq models, and added NanoGPT as a source.
  • Regex: Added a debug mode for regex visualization and the ability to save regex order and state as presets.
  • TTS: Improved handling of nested quotes when using the "Narrate quotes" option.
Bug Fixes
  • Fixed request streaming for the Vertex AI backend in Express mode.
  • Fixed erroneous replacement of newlines with <br> tags inside HTML code blocks.
  • Fixed custom toast positions not being applied to popups.
  • Fixed the depth of in-chat prompt injections when using the continue function with the Chat Completion API.
What's Changed
New Contributors

Full Changelog: https://github.com/SillyTavern/SillyTavern/compare/1.13.3...1.13.4

View originalPermalink
How 1.13.4 went

1.13.3

Added 9
  • Added Moonshot, Fireworks, and CometAPI Chat Completion sources
  • Context Template: Added {{anchorBefore}} and {{anchorAfter}} Story String placeholders
  • Advanced Formatting: Added the ability to place the Story String in-chat at depth
  • Advanced Formatting: Added OpenAI Harmony (gpt-oss) formatting templates
  • Chat Completion: Added an indication of model support for Image Inlining and Tool Calling options
  • World Info: Added a per-entry toggle to ignore budget constraints
Changed 9
  • Synchronized model lists for OpenAI, Claude, Cohere, and MistralAI
  • Synchronized the providers list for OpenRouter
  • Instruct Mode: Removed System Prompt wrapping sequences and added Story String wrapping sequences
  • Welcome Screen: The hint about setting an assistant will not be displayed for customized assistant greetings
  • Tokenizers: Downloadable tokenizer files now support GZIP compression
  • World Info: Updated the World Info editor toolbar layout and file selection dropdown
Removed 1
  • Removed the 01.AI Chat Completion source

SillyTavern 1.13.3

News

Most built-in formatting templates for Text Completion (instruct and context) have been updated to support proper Story String wrapping. To use the at-depth position and get a correctly formatted prompt:

  1. If you are using system-provided templates, restore your context and instruct templates to their default state.
  2. If you are using custom templates, update them manually by moving the wrapping to the Story String sequence settings.

See the documentation for more details.

Backends
  • Chat Completion: Removed the 01.AI source. Added Moonshot, Fireworks, and CometAPI sources.
  • Synchronized model lists for OpenAI, Claude, Cohere, and MistralAI.
  • Synchronized the providers list for OpenRouter.
Improvements
  • Instruct Mode: Removed System Prompt wrapping sequences. Added Story String wrapping sequences.
  • Context Template: Added {{anchorBefore}} and {{anchorAfter}} Story String placeholders.
  • Advanced Formatting: Added the ability to place the Story String in-chat at depth.
  • Advanced Formatting: Added OpenAI Harmony (gpt-oss) formatting templates.
  • Welcome Screen: The hint about setting an assistant will not be displayed for customized assistant greetings.
  • Chat Completion: Added an indication of model support for Image Inlining and Tool Calling options.
  • Tokenizers: Downloadable tokenizer files now support GZIP compression.
  • World Info: Added a per-entry toggle to ignore budget constraints.
  • World Info: Updated the World Info editor toolbar layout and file selection dropdown.
  • Tags: Added an option to prune unused tags in the Tags Management dialog.
  • Tags: All tri-state tag filters now persist their state on reload.
  • UI: The Alternate Greeting editor textarea can be maximized.
  • UX: Auto-scrolling behavior can be deactivated and snapped back more reliably.
  • Reasoning: Added a button to close all currently open reasoning blocks.
Extensions
  • Extension manifests can now specify a minimal SillyTavern client version.
  • Regex: Added support for named capture groups in "Replace With".
  • Quick Replies: QR sets can be bound to characters (non-exportable).
  • Quick Replies: Added a "Before message generation" auto-execute option.
  • TTS: Added an option to split voice maps for quotes, asterisks, and other text.
  • TTS: Added the MiniMax provider. Added the gpt-4o-mini-tts model for the OpenAI provider.
  • Image Generation: Added a Variety Boost option for NovelAI image generation.
  • Image Captioning: Always load the external models list for OpenRouter, Pollinations, and AI/ML.
STscript
  • Added the trim argument to the /gen and /sysgen commands to trim the output by sentence boundary.
  • The name argument of the /gen command will now activate group members if used in groups.
Bug fixes
  • Fixed a server crash when trying to back up the settings of a deleted user.
  • Fixed the pre-allocation of injections in chat history for Text Completion.
  • Fixed an issue where the server would try to DNS resolve the localhost domain.
  • Fixed an auto-load issue when opening recent chats from the Welcome Screen.
  • Fixed the syntax of YAML placeholders in the Additional Parameters dialog.
  • Fixed model reasoning extraction for the MistralAI source.
  • Fixed the duplication of multi-line example message separators in Instruct Mode.
  • Fixed the initialization of UI elements in the QR set duplication logic.
  • Fixed an issue with Character Filters after World Info entry duplication.
  • Fixed the removal of a name prefix from the prompt upon continuation in Text Completion.
  • Fixed MovingUI behavior when the resized element overlaps with the top bar.
  • Fixed the activation of group members on quiet generation when the last message is hidden.
  • Fixed chat metadata cloning compatibility for some third-party extensions.
  • Fixed highlighting for quoted run shorthand syntax when used with QR names containing a space.
Community updates
New Contributors

Full Changelog: https://github.com/SillyTavern/SillyTavern/compare/1.13.2...1.13.3

View originalPermalink
How 1.13.3 went

1.13.2

Added 11
  • Added nsigma sampler controls to Text Generation WebUI
  • Added an optional Persona title field for cosmetic titles
  • Persona avatars can now be thumbnailed to reduce network load
  • Macros are now replaced in the Banned Strings list for Text Completion
  • Added generation type filters to injected prompts in Chat Completion
  • Added templates for Kimi K2 and Mistral Small 24B models to Advanced Formatting
Changed 7
  • Synchronized model lists for Perplexity, Groq, MistralAI, AI21, and xAI with their respective APIs
  • Gemini models on OpenRouter will now be passed the same safety settings as AI Studio/Vertex AI
  • The original aspect ratio is now preserved when 'Never resize avatars' is enabled
  • Redesigned the layouts of the character search bar and Creator's Notes display
  • Character tags filters list is now scrollable
  • Messages with image attachments can now be swiped to regenerate
  • 'Start New Chat' will now start a temporary chat only if you are already in one
Removed 3
  • Scale Spellbook and Window AI removed from Chat Completion sources as they are no longer in service
  • Removed Mirostat parameters from Ollama UI as they are not supported
  • Removed retired Claude 2 models from the Claude model list
Deprecated 1
  • The 01.AI (lingyiwanwu) Chat Completion source is pending deprecation due to underutilization and geographical restrictions

SillyTavern 1.13.2

News
  • The 01.AI (lingyiwanwu) Chat Completion source is pending deprecation due to underutilization and geographical restrictions. Please reach out if you use it.
Backends
  • Chat Completion: Scale Spellbook and Window AI removed from sources as they are no longer in service.
  • Ollama: Removed Mirostat parameters from the UI as they are not supported.
  • Perplexity, Groq, MistralAI, AI21, xAI: Synchronized model lists with their respective APIs.
  • Claude: Removed retired Claude 2 models from the list.
  • Text Generation WebUI: Added nsigma sampler controls.
  • OpenRouter: Gemini models will now be passed the same safety settings as AI Studio/Vertex AI.
Improvements
  • Personas: Added an optional Persona title field for cosmetic titles.
  • Personas: Avatars can now be thumbnailed to reduce network load.
  • Personas: The original aspect ratio is now preserved when "Never resize avatars" is enabled.
  • Text Completion: Macros are now replaced in the Banned Strings list.
  • Chat Completion: Added generation type filters to injected prompts.
  • Advanced Formatting: Added templates for Kimi K2 and Mistral Small 24B models.
  • World Info: Added generation type filters to WI entries.
  • Import: Added the ability to import characters from Perchance AI.
  • Import: Added BYAF file import support.
  • UI: Redesigned the layouts of the character search bar and Creator's Notes display.
  • UI: A list of character tags filters is now scrollable.
  • UX: Messages with image attachments can now be swiped to regenerate.
  • UX: Added the ability to remove video attachments from messages.
  • Welcome Screen: "Start New Chat" will now start a temporary chat only if you are already in one.
  • Clean-Up: Added a cleanup scan for unused video attachments.
  • Server: Added a startup setting to use a global data path instead of the server data path.
  • Server: Increased request payload size limits (200 -> 500 Mb).
  • Server: Browser cache cleanup on server restart is now an optional setting.
  • Server: Console access log output is now controlled by the logging.enableAccessLog setting.
  • Added character tags as data attributes for rendered chat messages.
Extensions
  • Extensions can now save and load data from API setting presets.
  • Extensions can now use structured generation with a JSON schema.
  • Image Generation: Added support for video outputs from workflows.
  • TTS: Added Pollinations as a TTS source.
  • TTS: Added new models and speed control to the ElevenLabs TTS source.
  • Image Captioning: Added the 'Show captions in chat' setting.
  • Vectors: Added Google Vertex AI as a source.
STscript
  • /inject command: An ID will be automatically generated if not provided and will be returned as command output.
  • /genraw command: Added a prefill parameter.
  • {{setvar}}/{{setglobalvar}} macros: Now allow setting empty values.
Bug fixes
  • Fixed the uploading of MKV video attachments.
  • Fixed image models being displayed in the TogetherAI text model list.
  • Fixed being unable to search by model ID in OpenRouter for Text Completion.
  • Fixed checking for updates in extensions that are not Git repositories.
  • Fixed the Regex extension not loading if a script had an invalid placement array.
  • Fixed WI entries failing to load into the editor if they contained corrupted data.
  • Fixed thumbnails for backgrounds with names containing a single quote.
  • Fixed "Click to Edit" activating on copy from code blocks and while deleting messages.
  • Fixed not being able to assign additional WI connections during character creation.
  • Fixed the application of message CSS styling that uses pseudo-classes in selectors.
  • Fixed FAL.AI image models list loading.
  • Fixed {{getvar}} in slash commands if the macro name is not lowercase.
  • Fixed cutoff of hamburger and wand menus on height overflow.
  • Fixed prompts with inline videos when using Prompt Post-Processing.
  • Fixed non-streaming "Narrate by paragraph" to work regardless of the streaming setting.
Community updates
New Contributors

Full Changelog: https://github.com/SillyTavern/SillyTavern/compare/1.13.1...1.13.2

View originalPermalink
How 1.13.2 went

1.13.1

Added 20
  • Google Vertex AI (Full) support for accessing Gemini models with a service account
  • Google Vertex AI (Express) controls for Project ID and Region
  • New Gemini 2.5 Pro models in Google AI Studio with automatic API endpoint fallback for unlisted models
  • Cache TTL control for Claude in OpenRouter
  • New models to MistralAI list
  • Sampler controls in Pollinations

SillyTavern 1.13.1

News
  1. Node.js 18 has reached its EOL, please update Node runtime to the latest LTS version to continue receiving future updates.
  2. secrets.json file format has been updated and won't be compatible with previous SillyTavern versions.
Backends
  • Google Vertex AI (Full): Added support for accessing Gemini models with a service account.
  • Google Vertex AI (Express): Added controls for Project ID and Region.
  • Google AI Studio: Added new Gemini 2.5 Pro models. Models not in the list will be pulled from the API endpoint.
  • OpenRouter: Added cache TTL control for Claude; synchronized providers list.
  • MistralAI: Added new models to the list.
  • Pollinations: Added sampler controls, fixed reasoning tokens display.
  • xAI: Enabled backend web search capabilities.
  • DeepSeek: Added tool calls for reasoner model.
  • AI/ML API: Added as a Chat Completion source.
Improvements
  • Secrets: Added an ability to save multiple secret values per API type.
  • Welcome Page: Custom assistants will display their greeting message (if any).
  • Welcome Page: Added rename and delete buttons for recent chats.
  • Browser Launch (previously known as autorun): Added a config setting to choose the browser to launch.
  • Added a clean-up dialog to remove loose files and images from the data directory.
  • World Info: Budget cap max value increased to 64k tokens.
  • Backgrounds: Implemented lazy loading for backgrounds in the selection dialog.
  • Chat Completion: Added prompt post-processing types with tool calling support.
  • Added an ability to attach videos to messages (only supported by Gemini models).
  • Switched top drawer animations to use CSS transitions instead of JavaScript for better performance.
STscript
  • Added a setting to hide autocomplete suggestions in chat input.
  • Added a set of commands for managing secrets: /secret-id, /secret-write, etc.
  • Added access to WI entry character filters via /getwifield//setwifield commands.
Extensions
  • Extension manifest can now require other extensions presence to be loaded.
  • If any extensions failed to load, the reason will be displayed in the "Manage extensions" dialog.
  • Connection Profiles: Added Prompt Post-Processing and Secret ID to connection profiles.
  • Regex: Added bulk operations and multiple scripts export per file.
  • Image Generation: Added Google Imagen and AI/ML API as image generation sources. Added NovelAI V4.5 models.
  • TTS: Added Chatterbox, TTS WebUI and Google Gemini as TTS sources.
  • Gallery: Added delete functionality for gallery items.
  • Character Expressions: Added a switch between raw/full prompt building strategies for Main API classification.
  • Vector Storage: Allow chunk overlap when forced chunking on a custom delimiter.
Bug fixes
  • Fixed not being able to swipe right to generate if the first message was generated.
  • Fixed image prompt modified on image swipe not saving to the message title.
  • Fixed poor performance and memory leaks in the World Info editor.
  • Fixed personality/scenario missing in Chat Completion prompts if the respective utility prompt is empty.
  • Fixed parsing strings as numeric operands in STscript if command.
  • Fixed performance of "Back to parent chat" operation.
Community updates
New Contributors

Full Changelog: https://github.com/SillyTavern/SillyTavern/compare/1.13.0...1.13.1

View originalPermalink
How 1.13.1 went

More in AI

View all
codex

codex release notes.

codex
Ollama

Get up and running with large language models locally.

Ollama
openclaw

Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞

openclaw
tensorflow

An Open Source Machine Learning Framework for Everyone

tensorflow
superpowers

An agentic skills framework & software development methodology that works.

obra
ECC

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

affaan-m

Discussion

If you publish SillyTavern, you can claim this product by proving you administer its repository.