v0.2.0: Complete Redesign of Internals, Better, Faster and Cheaper UX

This Update Brings:

1. Complete OverHaul of Unnatural agent loop with Instructor module.

   • Before I had been using a More stricter agent loop, where Agent can only communicate with tool calls, and strict Json schemas. This caused slow response times, and high token usage. Now I am using the agent loop, which is more familiar to the LLM, does make the whole experience faster.
   • TLDR: More Faster experience, Less Token Usage

2. Better Input Handling and Multiline Support:

   • Before, I had been using the rich module's prompt function, which hadnt have multiline support(Ctrl+Enter or Ctrl+J).
   • Now, With help of prompt_toolkit library, A reliable, and cross-platform way to capture input, the Esc key, and the Ctrl+C signals.

3. Dropped sd and added gron to the available tools: gron is much more helpful than the unused sd.

4. From a Request from a user, Now, the recordings are standalone folders, which can be used anywhere(cross-platform too), with diffrent names generated by the Vision LLM. No more overwriting the existing recordings.

5. Logs and historys are stored in ~/.automatiq/{logs/, history/} better formatting, better error handling where user only see the summarized version of issue, and full traceback are stored in logs.

6. Fixed preloading. The banner isnt just a timepassing animation, It preloads, downloads the required executables in the background.

v0.1.3: Fixed slow startups, and Fixed buggy binary handling

In this release:

• The slow startup has be managed, by loading under banner time.
• fixed buggy scenarios of Shims installation of binary executables
• Updated Readme to remove repeated sections and added more information for local models

v0.1.2 : Decoupled UI and Core, Improved Cross Platform Support

Things done in this Update:

1. Automatiq Version is now at v0.1.2
2. UI is completely de-coupled from the core with blinker, helping out any future adaptation to a better UI like Textual, or a website.
3. Added Tests for ipython_sandbox such as timing benchmarks, execution, execution cancellation. Also re-implemented, sandbox with subprocess.Popen instead of subprocess.run
4. Also added tests for Blinker events, basic scenario tests for agent, and tests for Core Schema used in Instructor modeled prompts
5. Added CI for the added tests, as well as added badges for testing and liniting on ReadMe
6. Added Infographics of the process in ReadMe