v1.8.0 Released: Tools, Stronger Reliability & Better Multilingual Responses

Hi everyone,

v1.8.0 has been released:

New Features Settings Tab

A new Features tab has been added to the Settings menu, providing a clearer way to control which capabilities load at startup, remain available on demand, or stay disabled. Each feature shows its current configuration, running state, requirements, dependencies, and estimated resource usage.

A new Recommended Settings button considers your system specifications and selected app language to configure a suitable balance of compatibility, performance, and available features. This button can be found at the bottom of the new Features tab.

File tools remain restricted to the Bridge folder by default. Additional folders can be explicitly authorized from this tab. Restoring app defaults or using Reset All returns file access for Tool use to Bridge-only.

Smarter and More Reliable Tools

CielChan’s desktop tools have received a major reliability and multilingual update. Users can make naturally phrased requests to create or organize files within user specified directories and/or the Bridge folder, search the web, take screenshots, open applications, set reminders, check system information, and complete longer multi-step tasks. Tool requests are designed to work with the included default local model, while larger local and cloud models can provide more capable interpretation, planning, and execution for more complex requests.

Tool understanding, acknowledgements, results, recovery messages, reminders, generated skills, and goal handling have been improved across all 14 supported app languages.

CielChan can now plan longer goals, complete actions in sequence, verify results, and recover from failed steps. Tool availability, permissions, arguments, dependencies, and actions are also checked more carefully before running.

File tools are restricted to the Bridge folder by default. Additional folders can now be explicitly authorized through the new Features tab in Settings.

Web Search Improvements

Web searches now return shorter, cleaner, and more readable results.

Advertising links, tracking URLs, malformed addresses, and unnecessary search text are filtered from results. Useful source links are highlighted inside chat and can be selected to open the page using the computer’s default browser. CielChan now uses a separate concise version of search results for speech, preventing TTS from reading long URLs or excessive amounts of webpage content.

Chat Stop Button

The Chat Send button now changes into a Stop button while a manually requested response, tool task, or long-form writing job is running. Selecting Stop cleanly cancels the active request, including local or cloud generation, tool continuation, speech, and pending response activity. Any text already shown remains visible, and completed file changes or other finished actions are preserved.

Large Writing and File Creation

CielChan can now create longer stories, books, articles, reports, guides, scripts, and other written works directly as text files through the Tools flow. Previously available through the Book Reader, this extended writing capability is now also integrated into general tool requests. Longer works are created in connected sections and saved continuously as they are generated. Progress appears in chat, and completed sections remain safely stored if generation is interrupted.

Dynamic Skills and Workshop Improvements

Improved the Dynamic Skills system used to create, import, install, enable, disable, and share custom tool shortcuts. Generated skills now receive stronger validation for their target tool, arguments, triggers, dependencies, availability, and executable implementation before installation or Workshop publishing.

Trigger handling has been improved across all supported app languages.

Vision and Multilingual Companion Improvements

Observations, Watch Together, Game, and Computer interactions are now more consistent, natural, and grounded in what is currently visible for each respective language available within CielChan.

Responses should now better follow the selected app language more reliably, with less repetition, mixed-language text, malformed output, and unsupported visual details across compatible local and cloud vision models.

Observations and Watch Together Improvements

Observations, System Audio, Watch Together, Game, and Computer responses are now more concise, natural, and grounded in what is currently visible or audible.

Responses now should follow the selected app language more consistently across all 14 supported languages, with less repetition, mixed-language text, awkward phrasing, and malformed output. CielChan should be less likely to invent unsupported objects, actions, intentions, or outcomes, or describe unnecessary technical details.

Minigame Improvements

All existing minigames have received a reliability pass without changing their fundamental rules or gameplay.

AI choices, generated moves, scores, damage, indexes, health, suspicion, affection, and other game state are now validated more carefully. Invalid AI output uses safe fallbacks, duplicate submissions are prevented, and failed vision responses can be retried.

Delayed AI, vision, speech, and timer results are now tied to the correct game session so closing, restarting, or changing games cannot allow an older result to affect the new game.

Multiplayer and Ciel Social Improvements

Multiplayer rooms, peer-to-peer chat, automatic conversations, debates, avatar transfers, Fleet Duel, reconnect handling, and Ciel Social are now more reliable. Users should experience more consistent player identities, smoother reconnects, safer data sharing, and fewer synchronization issues during conversations, gameplay, profile and post updates, snapshots, and simultaneous avatar transfers.

Localization Improvements

Settings, tool acknowledgements, reminders, feature statuses, generated skills, ambient responses, and additional runtime interface elements have been improved across all 14 supported app languages.

Default Model, Live2D, and Sprite labels now update correctly when changing the app language. Additional UTF-8 and corrupted-text protections have also been added throughout chat, web results, memory, logs, and speech preparation.

Memory Improvements

Memory recall and storage now distinguish more reliably between requests to save information and requests to read existing memories.

Operational summaries, tool output, assistant-generated interpretations, and generic memory questions are less likely to be incorrectly stored as user facts. Explicit memory questions now require supporting saved evidence and should acknowledge when no matching memory can be found.

Startup and Performance Improvements

Optional systems are no longer unnecessarily activated or warmed during startup.

Features configured for on-demand use remain available without contributing to the initial loading sequence. Startup, cancellation, background work, response delivery, and feature-state synchronization have also received additional reliability improvements.

Interface and Avatar Fixes

Fixed persisted avatar sizing so BoZo, VRM, Live2D, and sprite avatars retain their selected scale after loading or changing models, settings localization refresh, feature-panel visibility, search behavior, dependency restoration, and several interface state inconsistencies.

Small Bug Fixes

Various additional tool routing, file handling, reminders, web search, memory, speech, vision, multiplayer, minigame, localization, startup, settings, avatar, encoding, recovery, and interface reliability improvements.

More to come soon.