RisuAI పాత వెర్షన్లు డౌన్‌లోడ్

- chore: update dependencies and update tests
- fix: update z-index for alert overlay in AlertComp component

- fix: remove unused local Python backend
- fix: stop Tauri streamed fetch on abort and replace the 240s limit with an idle timeout
- fix: remove password check from hex validation in isHex function
- fix: dbwrite permssion
- fix: handle low level access confirmation in readModule function

- fix: restore SafeMutationObserver payloads across the iframe bridge
- feat: magic link for node server

- fix: unload every V3 plugin and its replacers on plugin reload
- fix: show plugin name in install-via-plugin confirm and warn on replac

- fix: prevent setDatabase plugin round trip from clearing installed plugins

- feat: run plugin migration before plugin load

- fix: check undefined/falsy data in trimmarkdown

- ci: Add rust build cache and update actions in Tauri release workflow
- feat: add Gemini 3.6 Flash and 3.5 Flash models
- feat: add Gemini 3.7 Flash model
- feat: add local toggle
- feat: add new novelai model
- feat: Add Risuai.readInlay(id) to plugin API v3
- feat: add v3 chat output listener API
- feat: Add visual color palette selector
- feat: optimize streaming display rendering
- feat: PR check workflow with svelte-check and vitest
- fix: Character deletion reliability (chaId based lookup)
- fix: clean up leaked resources on V3 plugin unload
- fix: drag-and-drop reordering rework
- fix: evaluate CBS before counting lorebook tokens for cutoff
- fix: Fix module charx conversion: preserve moduleNamespace, hideChatIcon, and Replace Global Note
- fix: handle stream callback response via MessagePort in SandboxHost
- fix: include presetRegex in edittrans translation regex application
- fix: keep

- fix: fixed parser and typing issue

- fix: prompt template roles being inaccurately adjusted

- fix: decrypt local backup data before loading
- fix: avoid Tauri resource aborts in Auto Suggest

- feat: add a option to select export module type, only maintained until module charx is settled

- fix: partly revert params due to reverse proxy conflict

- feat: change coldstorage default option to enable only if there are no plugins
- this is due to some plugins without support of coldstorage makes some errors

- feat: coldstorage by default
- feat: add translation key for 'skipSavingAssetsOnWebSync'

- fix: converting character to module changes lorebooks in character

- fix: Reverted checkInput temp due to styling bug
- fix: Fixed character toggle causing updates too much

- chore: grant server.sh executable file permission
- chore: pin pnpm to latest v10
- feat: add Claude Opus 4.7 model with summarized thinking display
- feat: add Claude Opus 4.8 model
- feat: add configurable chat load counts
- feat: add configurable summary chunk separator for HypaV3
- feat: add current lorebook entries plugin api
- feat: add DeepSeek V4 model and fix reasoning content handling
- feat: add DeepSeek V4 tokenizer
- feat: add GLM-4.x(above 4.5) and GLM-5.x tokenizer support
- feat: add GPT-5.5 model support and reasoning / verbosity enhancement
- feat: add Ollama Cloud provider support
- feat: add openai flex processing toggle
- feat: add request / response body logging to Claude batch API
- feat: add setChatPanel API
- feat: add character to module conversion
- feat: add character toggle support
- feat: add charx import export on modules
- feat: add custom baseURL for OpenAI-compatible TTS providers
- feat: Improve self-host server setup docs
- feat: include environment metadata in error details
- feat: module to character conversion
- feat: plugin hooks for preprocessing and postprocessing
- feat: production-ready OpenAI Responses API support
- feat: Readd automatic clean chunks
- feat: refresh Related Links banner
- feat: Update zh-Hant.ts
- feat: using position on lorebook prompts
- Fix import path for characterCards module
- fix: Add Additional params settings for all models
- fix: add configurable trust proxy for reverse proxy deployments
- fix: Backup feature Chunk load error
- fix: cbs reverse, setdefaultvar
- fix: correct receiver allowedIPC check in postPluginChannelMessage
- fix: handle Tauri cold storage path and remote persistence
- fix: improve slider input accessibility
- fix: preserve error info from streamed_fetch parse failure
- fix: refactor additional parameter handling and add support for Anthropic
- fix: release doingChat lock in sendChat plugin API
- fix: remove duplicate anthropic beta declaration
- fix: remove duplicate OpenAI Responses API export
- fix: remove fork-only zh-Hant translations
- fix: resolve paramKey error, gate temperature by model, and add xhigh thinking effort
- fix: restore cold storage entries from local backups
- fix: support relaxed JSON for additional parameters
- fix: sync character grid search with simple list
- ref: migrate DisplaySettings to data-driven renderer
- style: Improved button visibility in light themes
- style: render checkbox inputs as toggle switches and optimize aesthetics

- feat: custom side bar
- fix: reset accumulated streaming reasoning
- feat: enhance popup editor with dynamic language support
- feat: rename coldstorage
- feat: enhance file removal API to support multiple file paths

- chore: upgrade Vite to v8 and modernize dependencies
- fix: strip blocks from SD submodel output
- fix: ensure system message comes first for WebLLM summarization models
- feat: Add NanoGPT provider integration
- feat: add coldstorage for characters
- feat: add coldstorage support for backups
- fix: Change font color to be seen even in the light theme
- refactor: extract Voyage Context 3 into ContextualEmbeddingProvider abstraction
- fix: add null guard for API response content across all providers
- fix: preserve callback identity across serialization for removeRisuScriptHandler/removeRisuReplacer
- fix: correct Kobold default URL and error display
- fix: close unclosed Thoughts tag in Anthropic response handler
- fix: wrap SafeElement[] returns with SafeClassArray for proper serialization
- fix: reject empty HypaV3 summary after stripping reasoning tags
- feat: always deobfuscate alert stack traces
- feat(translator): add LLM translation prompt presets with .risutl import/export
- perf: optimize chat hot paths
- fix: correct tokenizer dropdown values in Custom Models settings
- fix: show create character button in Beta Mobile GUI
- fix: [node] prevent /api/login crash from undefined writeFileSync
- fix: correct trimStartNewChat condition logic
- ci: upgrade GitHub Actions and Docker environment to Node.js 24
- chore(deps): bump express-rate-limit from 8.2.1 to 8.2.2
- chore(deps): bump dompurify from 3.3.1 to 3.3.2
- fix: remove hardcoded Windows target from .cargo/config.toml

Full Changelog: https://github.com/kwaroran/Risuai/compare/v2026.3.332...v2026.3.334

- feat: change plugin csp

- feat: added "periodically" option and added to some permssions.
- fix: permssion requirements spamming issue

- feat: tuned plugin handling
- docs: added documentation for missing plugin apis

- fix: seperate model option not setting up configs like URL
- fix: gemini flags set incorrectly

2026.2.280
- feat: add body interceptor for usage metadata and model status in Google request handling
- feat: enhance Content Security Policy with nonce for script execution in SandboxHost
- feat: added v2.1 plugin confirmation
- feat: add option to disable separate parameter change on preset change
- feat: added Easy Panel
2026.2.281
- fix: update CSP to allow inline styles in SandboxHost
- feat: add plugin channel functionality for message passing

- feat: now plugins models registered in v3 api will be also be added as seperate model
- feat: renamed old "Plugin" to "Plugin Legacy" model

- feat: Use Segmented Controls for Claude thinking parameters
- feat: Implement infinite scroll for inlay explorer
- fix: fixed addProvider hijacking vulnerability via mode. (v3 breaking)
- This will limit mode to v3
- The v2.1 api isn't yet blocked, considering its widely used and its scheduled for removal
- fix: fixed risuai server auth access vulnerability (v3 breaking)
- fix: apply Content-Type header on Google Cloud requests when streaming is disabled

- feat: Traditional Chinese translation update
- fix: fetchNative and globalFetch properly support plugin
- fix: prevent TDZ crash in sendChat error inlay (HypaV3 early failure)
- fix: Restore inlay explorer languages
- fix: Ignore audio.play() error when it is in fire-and-forget
- ref: Separate common functions out of requests.ts
- ref: Move parser.svelte under parser
- ref: fix typos in PR template
- feat: Prevent importing a HypaV3 preset as a module
- feat: Add accessibility option to move hamburger menu button
- feat: support claude adaptive thinking
- feat: Plugin V3 api for getting translation cache
- feat: add settings item type SegmentControl
- feat: add Claude Sonnet 4.6
- feat: add Gemini 3.1
- feat: add dynamic model registry
- feat: add replacer API for plugin v3
- feat: proxy AbortSignal across iframe boundary for V3 plugin provider callbacks

- feat: add GraphMem MCP for memory management using graph vector search
- feat: implement MCP module registration and management for plugins
- feat: add blockquote styling feature with customizable options
- feat: add dice MCP
- feat: add SafeLocalPluginStorage interface
- breaking: remove support for V2.0 plugins and notify users to update to API version 3.0

- feat: Move aux models setting to models setting page
- feat: Improve OpenRouter experience
- feat: add update reminder feature with postpone options
- test: Inlay functions
- feat: add aws opus 4.5 and 4.6
- fix: make typecheck pass
- feat: Partial Edit (Block/Drag)
- feat: Corrupted Module Lorebook Reset option
- feat: add WaveSpeedAI integration for image generation
- feat: hide jailbreak toggle when no jailbreak prompt exists
- feat: add missing language keys
- ref: Move convertImage/checkImageType out of parser

- feat: make remote saving option until its safety is proven

- refactor: Type language objects
- fix: add crossorigin attribute to manifest link
- chore: Remove all Capacitor-related codes
- test: Add selIdState to stores.svelte mocks
- feat: Register #escape to CBS doc, add #escape::keep
- feat: Inlay asset explorer in Playground
- chore: Remove unused visual novel mode
- fix: cbs {{#when}} ... {{:else}} ... {{/when}} works in another pure block
- chore: Remove dead settings
- fix: Exclude account information from local backups
- feat: add default value support to alertInput and related components
- fix: Hyper Memory V3 'ot' & 'startsWith' error Bug Resolve
- l10n: Languages for inlay assets explorer
- test: #each, nested #each, nested #when
- feat: other minor updates

fix: v2.1 running on v2.0 in specific cases

- fix: generate unique asset paths in charx/charxJpeg export
- test: Test against #pure, #puredisplay, escaped chars
- Fix: Plugin V3 getCurrentCharacter, getItem Proxy Bug Fix
- feat: Save Partial Local Backup
- fix: Proper support of nested JSON objects in CBS
- fix: Change DBSaveError alertConfirm in saveDb to alertError
- feat: Add 'Scroll to Active Character' hotkey (Ctrl+G)
- perf: improve module import speed
- test: Test CBS string expressions
- test: MCP internal:risuai ModuleHandler reading tools
- test: Add newly found edge cases of #if
- feat: improve prompt diff replace-block alignment using DP + similarity scoring
- fix: prevent null parameters when separate params are enabled but unset
- refactor: Move parseChatML out of parser.svelte.ts
- chore: Delete promptTypeParser in parser.svelte.ts
- chore: Update @adobe/css-tools to 4.4.4
- fix: Partial backup asset path handling in Tauri
- feat: Add caption/textarea toggles
- fix: update missing inlay tag in imggen inlay mode
- Fix typo in README for 'Additional Assets'
- refactor: Import lodash functions from individual modules
- chore: Remove unused local model functions
- chore: Remove unused NAI login function
- refactor: Move chatvar functions out of parser.svelte.ts
- test: Add #when multiple operators + logical operator case
- feat: Enable F12 DevTools in Tauri builds
- refactor: Cache getAssetSrc/getEmoSrc results
- chore: Remove AI21 from LLMProvider
- chore: Remove unused prebuilt presets
- chore: Remove unused bot creator
- fix: resolve token count reactivity in CharConfig
- refactor: Playground lazyloading
- chore: Remove unused DB keys
- refactor: Separate TriggerV1List, lazily load it
- chore: Remove unused loggen
- fix: google backup auth flow error
- fix: RisuaiPluginAPI readImage and saveAsset security issue
- feat: add character and chat management methods to RisuaiPluginAPI
- fix: add error handling for plugin termination in unloadV3Plugin
- feat: update getDatabase method to include optional key filtering
- feat: add getRuntimeInfo method to RisuaiPluginAPI for platform and save method details

- update Tauri dependencies to latest versions in package.json and Cargo.toml
- this will fix infinite loading issue on local version

- fixed branch merge, pulled 04a3829

- temp fix: added risu cache fallback

- add: Show detailed error information on alert
- add: lua 'alertConfirm'
- improve: google blank response handling
- improve: google stable handling
- fix: Unable to import chat error
- fix: Resolve 403 Forbidden error in Node.js proxy

- fix: length 0 or non-array preset again

- Fix presets saved as non-presets

- Add: Lua sleep
- Fix: gemini null part causing crash
- Fix: cbs return causing crash

- Fix: character saving access deleting character

- add: GPT-5 series
- add: Claude Opus 4.1
- add: Support Persian (Farsi) in google translate
- fix: Cbs random, pick Bug where the first element cannot be picked
- fix: Stabilize Hub Proxy by resolving redirect and request body errors

- Fix: {{hidden_key::}}
- Fix: chat module menu not opening
- Improve: save block encoding

- Add simple encryption CBS
- {{crypt::A}} (caesar cipher, same function for encryption and decryption)
- {{xorencrypt::A}}
- {{xordecrypt::A}}

- Fix lorebook import
- Fix img detection

- add: addtional asset html <img> tag mapping
- add: insert asset prompt option
- add: asset style option
- add: {{chardisplayasset}}, {{//@customimageinstruction}}
- improve: move additional asset to display tab
- fix: auto translation with 'cached message only' option

- add: Pass chat index to editDisplay
- add: Upsert local lore books from Lua
- add: timing metadata to chat request information tab
- fix: Correct Function Calling Context and Enable Parallel Tool Use
- fix: Add google search MCP module
- fix: Moved the body.tool removal to fix GoogleAI error.
- fix: Variety+ option in NAI Image generation
- fix: Playground role name color & name update
- fix: enhance memory retrieval
- fix: Patch SSRF Vulnerability
- fix: Correct CBS parsing for Array in v2trigger
- fix: retranslate trigger and translation loading spinner visibility
- fix: prevent infinite backup accumulation on Tauri/Nod
- fix: handle empty response on google request
- Improve: Google/OpenAI request experience with tool use, including response streaming
- Improve: Enhance v2Trigger Features
- Improve: support NAI v4.5 vibe transfer
- Improve: gemini 2.5 pro stable
- Improve: use pixel-based resizing on writing inlay image
- Improve: Use module lorebook when importing charx

- Fixed styling issue

Add: MCP modules
Add: MCP support for google cloud
Add: MCP support for claude
Add: MCP support for openai compatible
Add: internal MCP: File System Access MCP
Add: internal MCP: Risuai Access MCP
Add: internal MCP: AI Access Client
Add: <tool call> tag
Add: drag-and-drop reordering for prompt presets
fix: Make sure to update whole chat when necessary
fix: reroll icon showing

- Add: Claude 1 Hour caching
- Add: New Lua functions for retrieving last character and user messages
- Add: Docker multi-stage build support
- Add: note to persona
- Add: NAI v4.5 support and fix minor issues in NovelAI image generation
- Add: delete and reorder functionality to Custom Model accordion
- Add: Limited MCP support
- Improve: Chat Loading
- Improve: Trigger V2 context menu positioning and code organization
- Improve: Chat Copy Feature with Modern HTML Template
- Improve: Syntax highlighted code blocks
- Fix: API key priority for Custom Models with Claude format
- Fix: incorrect references on clipboard paste in Trigger v2
- Fix: Apply indentation to all lines for Trigger V2 effect description
- Fix: v2EndIndent positioning in nested if/else statements in Trigger v2
- Fix: fetchLog index bug in response streaming (Tauri)
- Fix: swapped span classes for single and double quotes
- fix: add null check for addChat in scripting API

- Added lorebook folders

- fix: Undefined property access in AlertComp
- fix: Divider toggles should reference within group

- Added {{metadata::A}} CBS
- Added {{iserror::A}} CBS
- Added {{displayescapedcurlybracketopen}}, {{displayescapedcurlybracketclose}}, {{doubledisplayescapedcurlybracketopen}}, {{doubledisplayescapedcurlybracketclose}} CBS for displaying curly brackets.
- short alias: {{decbo}}, {{decbc}}, {{bo}}, {{bc}}
- Added @@keep_activate_after_match decorator
- This will force activation after the match
- Added @@dont_activate_after_match decorator
- This will force deactivation after the match
- Added @@recursive and @@unrecursive decorator
- Although it was already doable by combining other decorators, this will make it more readable and performant.
- Added @@no_recursive_search decorator
- This will make the lorebook entry unsearchable via other lorebook entries' content
- Added non-underbar syntax for all functions with an underscore

- Reverted to old loading temporarily
- removed duplicated NAI settings input
- improved HypaV3 Modal
- Updated tauri and its plugins to 2.5

- 2nd Anniversary!

158.2.0
- Fixed abortsignal error

- Fixed fallback model duplication

- Added fallback options
- Moved some options to correct place
- Fixed translation alert sound playing before it translated in translation via LLM

- Added Hotkey settings
- Added new hotkeys
- Changed focus input hotkey to space
- Seperated share menu

- Added image output for gemini exp
- Added image output option
- Added {{inlayeddata::ID}} for model image output
- Added {{assetprompt::<asset name>}} for putting assets as prompts.
- use icon as <asset name> for bot icon
- Added {{module_assetlist}} #786
- Fixed lua functions #782
- Fixed translation request #783

- Added show folder name option
- Added automatic cache point option (experimental)
- Added chat compression option (experimental)
- Changed select/text toggles to fixed size
- Fixed cache point causing infinite loop

- Fixed typo
- Fixed toggle binding
- Fixed slot
- Fixed reasoning effort

- Added select type toggle syntax <key>=<name>=select=<option1>,<option2>,<option3>...
- Added text type toggle syntax <key>=<name>=text
- Added {{bgm::<asset name>}}, which doesn't refresh
- Added caching point prompt item
- Added missing providers #780
- Fixed {{solt::content}} to {{slot::content}}
- Fixed incorrect message caching, resulting weird flashing and wrong output
- Fixed folders #779
- Fixed unused assets not clearing on local version

- Fixed scroll bug
- Added GPT-4.5

- feat: add support for OpenAI-compatible embedding API by @Bo26fhmC5M in https://github.com/kwaroran/RisuAI/pull/776
- feat: Chat Folder by @HopeJourney in https://github.com/kwaroran/RisuAI/pull/766
- feat: add local activation for lorebook by @bangonicdd2 in https://github.com/kwaroran/RisuAI/pull/767
- feat: aws sonnet 3.7 + legacy by @bangonicdd2 in https://github.com/kwaroran/RisuAI/pull/772
- Optimize chat rendering and performance with caching and lazy loading by @hashcoko in https://github.com/kwaroran/RisuAI/pull/773
- Fix Docker image build error by @broot5 in https://github.com/kwaroran/RisuAI/pull/774

- Fixed claude streaming

- Added Claude Sonnet 3.7
- Improved translation playground
- Fixed model desync #770

- Added TriggerV2 hotkeys
- Added TriggerV2 context menu
- Added missing triggers from V1 in TriggerV2
- Added advanced if trigger
- Added show deprecated v1 trigger option
- Added show translation loading option #761
- Improved TriggerV2 GUI
- Fixed asset loading #760

- Added request related triggers

- Fixed some v2 triggers not showing/acting properly

- Fixed display triggers returning null

- Added progress dialog
- Added new gemini models
- Added Charx-Jpeg
- Added legacy translation support for translate only cached option #742
- Added lua button trigger #743
- Updated Traditional Chinese translation #741

- Added Charx-JPEG format
- Added Reasoning effort parameter
- Added paste image functionality
- Improved HypaV3 #737 #732

- Added inlayed error response option

- Added Gemini Models
- Gemini Flash 2.0
- Gemini Pro 2.0 Exp 0128
- Gemini Flash 2.0 Thinking 0121
- Added Deepseek model options
- Deepseek Chat
- Deepseek Reasoner
- Added Deepinfra support and its models
- deepseek-ai/DeepSeek-R1
- deepseek-ai/DeepSeek-R1-Distill-Llama-70B
- deepseek-ai/DeepSeek-V3
- meta-llama/Llama-3.3-70B-Instruct-Turbo
- meta-llama/Llama-3.3-70B-Instruct
- microsoft/phi-4
- meta-llama/Meta-Llama-3.1-70B-Instruct
- meta-llama/Meta-Llama-3.1-8B-Instruct
- meta-llama/Meta-Llama-3.1-405B-Instruct
- Qwen/QwQ-32B-Preview
- meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo
- meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo
- Qwen/Qwen2.5-Coder-32B-Instruct
- nvidia/Llama-3.1-Nemotron-70B-Instruct
- Qwen/Qwen2.5-72B-Instruct
- meta-llama/Llama-3.2-90B-Vision-Instruct
- meta-llama/Llama-3.2-11B-Vision-Instruct
- microsoft/WizardLM-2-8x22B
- 01-ai/Yi-34B-Chat
- Austism/chronos-hermes-13b-v2
- Gryphe/MythoMax-L2-13b
- Gryphe/MythoMax-L2-13b-turbo
- Sao10K/L3.3-70B-Euryale-v2.3
- Sao10K/L3.1-70B-Euryale-v2.2
- Sao10K/L3-70B-Euryale-v2.1
- google/gemma-2-27b-it
- google/gemma-2-9b-it
- Fixed HypaV3 scrolling ` #731
- Added setArg() plugin API #729