SillyTavern 1.12.12

News

As Extras API is deprecated since April 2024, its support in the built-in extensions will be eventually removed starting from the next release.

The following extensions will be affected:

Character Expressions (including TalkingHead)
Image Captioning
Image Generation
Summarization
Vector Storage

NB: Only Extras API connectivity will be removed. The extensions will continue to work as usual. Please check the settings of affected extensions and migrate to other supported services.

Backends

Google AI Studio: Added new Gemini models.
Perplexity: Added new models, removed deprecated models.
Groq: Added new models, removed deprecated models.
OpenRouter: Added new providers and support for reasoning models.
OpenAI: Added o1/o3 models and reasoning effort control.
DeepSeek: Added reasoning model. Added tool calling support for chat model.
Ollama: Add num_batch config value.

Improvements

Improved compatibility with DeepSeek R1 and other reasoning models. Check the documentation for more details: Reasoning.
Most LocalStorage data is now stored in settings.json to prevent data loss.
Banned strings (Text Completion): Added a global non-preset list of banned strings and a killswitch to turn them off.
Server: Better compatibility with IPv6 networking, including auto-detection of supported interfaces. The server can now listen on any IP address.
Server: Added console logging level configuration.
Server: Added ability to customize HTTP error pages.
Server: CSRF protection token is now saved to session cookies.
Plugins: Added auto-update functionality for plugins with config for opt-out.
Groups: Added auxiliary field display to the members list.
Temporary Assistant chats can be exported as JSONL files.
Message image swipes can be removed one by one.

Extensions

Connection Profiles: Can now save and load custom stop strings.
WebSearch: Added an ability to search for images. Added support for Seprer API.
Image Generation: Added FAL.AI as a provider.
TTS: Added a toggle to narrate by paragraphs (non-streaming).
Regex: Added Reasoning blocks as a target for regex matching.
Typing Indicator: Moved to an official downloadable extension.

STscript

Added /stop-strings command to manage custom stop strings.
Added /member-get command to retrieve group member information.
Added /replace command to replace text in strings.
Added commands to manipulate reasoning blocks: (/reasoning-set, /reasoning-get, /reasoning-parse).

Fixes

Sampler seed values of -1 will not be sent to Text Completion backends anymore.
Imported characters will use a placeholder image if their original image is corrupted.
Browser cache for the app will be forced to clear on uploading an image or restarting the server.
Streamed generation will stop immediately when smooth streaming is used.
Character media alerts will be cleared from storage when the character is deleted.
Fixed background fitting class resetting after picking a new background image.
Fixed regex applying to prompts when "Alter Outgoing Prompt" is unchecked.
Fixed a looping condition when autoswipe is enabled and the last message is empty.
Fixed /lockbg and /unlockbg commands being non-functional.

Community updates

Automatically detect what IP versions are available and use them by @BPplays in #3226
adds checking for if localhost resolves + JSDoc additions by @BPplays in #3283
add presets for DeepSeek R1 models by @kallewoof in #3329
exporting openGroupId from group-chats.js by @qvink in #3327
fix: corrects the preset name and adds deepseek v2.5 hash. by @kallewoof in #3333
Add new Gemini thinking model and its alias by @sirius422 in #3332
adding event after loading more messages in chat by @qvink in #3331
Adding reverse proxy support to DeepSeek chat completion by @subzero5544 in #3328
Update zh-CN translations by @Tosd0 in #3338
config.yaml: Group extension settings into one section by @Cohee1207 in #3315
New exported methods: getCharacters(), uuidv4(), humanizedDateTime() by @bmen25124 in #3340
Validate avatar_url field with a middleware by @Cohee1207 in #3314
Allow customisation of the 403 page by @Spappz in #3342
Replace CSRF middleware by @Cohee1207 in #3351
Add char version to group panel member names by @Wolfsblvt in #3360
Use default avatar if imported image is corrupted by @Cohee1207 in #3361
Update docker command in README to actually run the container by @witten in #3367
Better reasoning by @Cohee1207 in #3345
Fix tts.skip_tags's regex to match newlines by @braxton-p in #3373
Add command and profile for custom stop strings by @Cohee1207 in #3368
Decrease TTS generation delay by splitting a message on a new line by @fearmear in #3371
Fix /note-role argument name by @Finadil in #3379
Staging into reasoning-regex by @Cohee1207 in #3382
Regex: fix "Alter Outgoing Prompt" applying when unchecked by @Cohee1207 in #3380
Sets italics inside quotes to parent color. by @BooleanWhale in #3378
Add regex processing for reasoning blocks by @Cohee1207 in #3370
Autoswipe: Fix endless loop if blacklist empty by @Cohee1207 in #3388
fix: resolve URL duplication in AllTalk TTS audio output on V1 AllTalk by @erew123 in #3392
Update loader being removed correctly on manually forced transition duration by @IceFog72 in #3384
Reasoning blocks auto-parsing by @Cohee1207 in #3386
Add new OpenRouter providers by @pcpthm in #3397
Always add reasoning for continue by @pcpthm in #3395
Support reasoning for OpenRouter Text Completion by @pcpthm in #3396
Adds /replace by @Succubyss in #3377
Adds /member-get via enhancing findGroupMemberId by @Succubyss in #3409
Update Groq models by @conornash in #3408
Added option to disable the chat logs in terminal v2 by @Eradev in #3307
New exported methods: loadWorldInfo(), saveWorldInfo(), updateWorldIn… by @bmen25124 in #3410
Thinking is stylish - if you are not cool, I don't know how to help you by @Wolfsblvt in #3413
Add missing o3-mini model selector labels by @dvejmz in #3415
Fix reasoning is duplicated when a continue generation is interrupted by @pcpthm in #3418
OpenAI "Reasoning Effort" for o1/o3-mini + oai models restructured by @Wolfsblvt in #3424
Optimize zh-CN Translations by @Tosd0 in #3425
WI dry run checks delay + logging by @Wolfsblvt in #3423
await GENERATE_BEFORE_COMBINE_PROMPTS by @qvink in #3430
Better IPv6 support by @Cohee1207 in #3431
Add new gemini models by @Wolfsblvt in #3437
Add new gemini models and update the safety settings by @sirius422 in #3434
fix: adjust safetySettings handling for legacy Gemini models by @sirius422 in #3439
Enable streaming for gpt-o1 and gpt-o3 by @Wolfsblvt in #3441
Smaller improvements and fixes to the reasoning UI by @Wolfsblvt in #3442
Update layout of Reasoning settings block by @Cohee1207 in #3448
Fix top_k text value not being able to go up to 500, like the slider does by @subzero5544 in #3450
Allow mobile to access the "Raw Prompt" in Prompt Itemization by @Wolfsblvt in #3456
Remove reasoning from queit prompt results if auto-parse is enabled by @Cohee1207 in #3465
Add settings.json-backed KV string storage by @Cohee1207 in #3463
Fix kr i18n by @Eradev in #3468
Track reasoning duration for models with hidden reasoning by @Wolfsblvt in #3444
Add o1 to vision-supported models by @Wolfsblvt in #3470
Feature: FAL.AI Integration by @Dakraid in #3476
Ollama: Add num_batch config value by @Cohee1207 in #3475
Feature: Allow user to configure an address to listen to by @Dakraid in #3478
Fix auto expand on thinking during new message by @Wolfsblvt in #3480
Add cache buster middleware to clear browser cache on server restart by @Cohee1207 in #3481
Text Completion: Add global banned strings list by @Cohee1207 in #3477
Fix auto expand on thinking during new message (v2.1) by @Wolfsblvt in #3482
Set safety_tolerance for Flux on FAL by @Dakraid in #3490
Plugins: Add auto-update functionality by @Cohee1207 in #3487
Staging by @Cohee1207 in #3493

New Contributors

@qvink made their first contribution in #3327
@subzero5544 made their first contribution in #3328
@bmen25124 made their first contribution in #3340
@Spappz made their first contribution in #3342
@witten made their first contribution in #3367
@braxton-p made their first contribution in #3373
@fearmear made their first contribution in #3371
@BooleanWhale made their first contribution in #3378
@pcpthm made their first contribution in #3397

Full Changelog: 1.12.11...1.12.12

SillyTavern/SillyTavern 1.12.12 on GitHub