Skip to content

v4.7.7 - Multimodal Vision, Bug Fixes & Code Quality

Choose a tag to compare

View all tags
@tvcnet tvcnet released this 20 Mar 00:17
· 7 commits to main since this release
v4.7.7
39c68c3

What's New in v4.7.7

Multimodal (Vision) AI Support

All four AI providers (Claude, OpenAI, Gemini, Ollama) now officially support sending uploaded images alongside the text prompt. When photos are uploaded, the prompt engine automatically injects visual analysis instructions - enabling more visually relevant, engaging posts.

Polishing UX Transparency

When a generated post exceeds a platform's character limit and requires a second API pass, the loading message now explicitly shows "Polishing text to fit platform limits..." instead of random loading phrases, giving users clear transparency about the delay.

Code Quality Improvements

  • Namespaced Utilities: Consolidated loose global functions (tstStripLinks, tstStripMarkdown, tstSanitize) into a modular TST_Utils namespace object.
    • Declarative Alpine.js Keyboard Shortcuts: Replaced imperative document.addEventListener bindings with idiomatic @keydown directives on the <body> element, fixing an invalid .cmd modifier.
    • Escape Key Enhancement: Escape now cascades - closing the Settings modal first, then the Help modal.

Bug Fixes

  • Critical: Prompt Length Instruction Bug - Fixed a bug where the AI was never receiving the "CONCISE", "BALANCED", or "DEEP DIVE" content length instruction due to destructuring length instead of contentLength.
    • Word Count Edge Case - Fixed wordCount calculation where empty/whitespace-only responses displayed "1 word" instead of "0 words".

Files Changed (8)

tvcnet-social-toolkit.html, social-toolkit.js, tst-utils.js, tst-services.js, tst-content.js, tst-config.js, CHANGELOG.md, README.md

Assets 2
Loading