All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Without the length-adjustment, the semantic cache is sometimes too aggressive. It's not disabled by default to avoid unexpected behavior.
- Added a launcher function
launch
to make it easier to launch the app. - Semantic caching provided by SemanticCaches.jl. You can change it by setting
cached=false
in thelaunch()
function. It's disabled by default.
- Fixed a bug when caching would error for certain types of HTTP body (eg,
IOBuffer
)
Chat tab
- Added support for Groq's speech-to-text model for much lower latency
Prompt Builder tab
- Added a "Detailed view" button to toggle between a compact view and a detailed view
- Allow user to choose between different Prompt Generators (see the "Prompt Builder Settings" at the top)
- Added an alternative template
:PromptGeneratorUpsampler
that generates a prompt template for chat without extensive examples (zero-shot prompt)
Configuration
- Changed default OpenAI model to GPT-4 Omni
- Improve grid layout and wider prompt textboxes
- Increased PromptingTools compat to 0.33
- Fixed a piracy warning in Aqua
- Improved subtle bugs in the Speech-to-text functionality on chat tab
Chat tab
- Added delete icon to the last message in the conversation (for easy deletion)
- Added a button in "Advanced Settings" to "Fork a conversation" (save to history for reference, but continue from a fresh copy)
- Added a focus when a template is selected (on expand, on template selection, etc)
- Added a little edit icon for messages (disables the q-popup-edit that was distracting and jumping out too often)
- Added speech-to-text for the chat (click Record/Stop -> it will paste the text into the chat input and copy it in your clipboard)
Meta-prompting tab
- Added an experimental meta-prompting experience based on arxiv. See the tab "Meta-Prompting" for more details.
Prompt Builder tab
- Added a tab to generate "prompt templates" that you can then "Apply In Chat" (it jumps to the chat tab and provides template variables to fill in)
- Allows selecting different models (eg, Opus or more powerful ones) and defining how many samples are provided to give you a choice
- The first iteration of the GUI released