Skip to content

Ayai GPT Nexus v1.0.0-alpha.8

Latest
Latest
Compare
Choose a tag to compare
@zeitlings zeitlings released this 17 Nov 22:56
v1.0.0-alpha.8
7dd7666
  • New
    • Added File Attachments
      • Document Attachments
        • Supporting: PDF, docx, plain text, source
        • Requires pandoc to extract text from docx. Otherwise a crude workaround will be used that loses formatting.
      • Image Attachments
        • Supporting: ChatGPT, Gemini, Claude
    • Added "Sticky Preview" ad hoc GUI applet (persistent floating window)
    • Migrated to Conversations (reworked chat history)
      • The chats (array of messages) will be migrated to a new conversation struct with richer metadata. Better safe than sorry: backup your archive.
      • Added conversation detail view
      • Added favorites, pins, indicators for attached files and images
      • Added !bang filters: !fav, !doc, !img, !key
      • Added keyword / tag system (access via !key)
        • Keywords are generated together with Summary Titles and can be manually configured via the conversation detail view.
    • Added automated summary title creation
      • Added configuration option "Create Title Summaries"
      • Added environment variable shared_chat_title_summary_override_service
      • Added environment variable shared_chat_title_summary_override_model
    • Added sound and HUD notification to inference actions completions that were copied to the clipboard
  • Improvements
    • Inference Actions now accept instructions or questions directly from the ScriptFilter, i.e. when no preset matches the user input.
      • If the query is suffixed with a question mark, the answer.question inference action will be used as template, otherwise follow.instruction.
    • Added configuration option to show model info in chat
    • Added the model to the accessibility description of the assistant icon, i.e. hover over the name with your cursor to see the model that was used.
    • Added token counting for Gemini
    • Added token counting for Claude
    • Added token counting for Proxy: Groq
    • Added token counting for Proxy: OpenRouter
    • Added token counting for Proxy: Fireworks.ai
    • Added token counting for Proxy: Together.ai
    • Added token counting for Proxy: DeepInfra
    • Added truncation to message preview slices in history when searching
    • Chat history filtering: purge all obvious markdown markup
    • Extensive internal refactoring
  • Updates
    • Rewritten conversation history
    • Migrated utility user actions into the main executable
    • Migrated archive handling into the main executable
    • Made all relevant external triggers available via URL handler
    • Models
      • Added Gemini 1.5 Flash 8B
      • Updated to claude-3-5-haiku-latest
      • Updated to claude-3-5-sonnet-latest
      • Updated to claude-3-opus-latest
      • Removed Gemini 1.0 Pro
  • Fixes
    • Adjusted HUD for macOS Sequoia

Important

Back up your customized Inference Actions before updating or your changes will be lost.

Assets 3
Loading