2.4.17

• Refactored kernel and render events.
• Added multiple formatted extra outputs from plugins.
• Added a 'Reset (clear)' option in the context list for clearing the current context.
• Improved agents steps output handling.
• Fixed the audio button.
• Improved the stop event.
• Other small fixes.

2.4.16

• Code Interpreter and Files I/O input/output is now displayed in chat window, using syntax highlight.
• Refactored plugins structure.
• Fix: command execution in Chat in Files mode when no index is selected.
• Fix: missing translations in Agent mode.

2.4.15

• Vision analysis added to all modes.
• Added commands for the Vision plugin: image analysis from an attachment, screenshot analysis, camera image analysis - performed in the background in every mode, without switching to vision mode.
• Improved execution of multiple commands at once.
• Improved integration with IPython and extended instructions for the model.
• Other fixes and improvements.

2.4.14

• Added a Loop / evaluate mode for Llama-index agents, allowing them to work autonomously in a loop and evaluates the correctness of received results using a percentage scale from 0% to 100%, and automatically continue and correct responses below expected value.
• Updated layout: mode and model lists replaced with comboboxes.

2.4.13

• Introduced Code Interpreter (v2) with IPython: Enables session state retention on a kernel and building on previous computations, allowing for a Jupyter-like workflow, which is useful for iterative development and data analysis. Supports magic commands like !pip install <package_name> for package installation within the session. (currently in beta)

2.4.12

• Added httpx-socks to the dependencies, enabling the use of SOCKS proxies.

2.4.11

• Added a new Llama-index agent: OpenAI Assistant.
• Added proxy settings to the configuration dialog.
• Added more models to the Agent (Llama-index) mode.
• Improved the agents/presets dialog window.
• Disabled the "OpenAI API KEY is not set" Llama-index error when using a local model in Chat with Files mode. You can now use local embeddings, such as Llama3 via Ollama, and use local models without any warnings.

2.4.10

• Added support for Llama Index agents. (beta)
• Introduced a new working mode: Agent (Llama Index).
• Added 3 Llama Index agents: OpenAI Agent, ReAct Agent, and Structured Planner Agent.
• Fixed: passing an embed model to vector stores on indexing.
• Fixed: python3-tk error in snap version.

2.4.9

• Mouse movement has been moved to the PyAutoGUI library for seamless operation.
• Bridge calls have been moved to an asynchronous thread.
• Fixed the DALL-E call in Image mode.
• Speed improvements and optimizations.

2.4.8

• Fix: restoring order of notepads.
• Fix: mouse move offset in Mouse and Keyboard plugin.