Changelog - April 10, 2025

Feature Additions

• Introduced Model Validator to ensure compatibility and expected behavior before execution.
• Implemented dynamic tool invocation for supported models, enhancing modularity.
• Added support for remote Ollama servers, enabling distributed model execution.

Shell & Output Enhancements

• Cleaned piped output: system messages are now suppressed unless UI logging is enabled.
• Output validation added for generated shell commands to reduce execution errors.

Filtering & UI Improvements

• Enhanced filtering logic for more precise processing.
• Multiple UI responsiveness updates for smoother rendering and user interaction.

Bug Fixes

• Resolved issue where shell override logic bypassed full input parsing, acting only on hardcoded commands.
• Fixed shell hang behavior when incorrect password is entered.
• Addressed repeated LLM responses for previously handled inputs.

This update significantly improves execution reliability, expands deployment flexibility, and strengthens overall shell interaction stability.

Full Changelog: beta_v0.3.1...beta_v0.4

Changelog - March 18, 2025

Feature Additions

• Made shell output analysis optional.
• Shell now detects sudo in pipes.
• Persistence of shell session (for command such as !cd , !export and keeping sudo for the session without storing password in a variable).
• Queuing tasks. User may interact with a shell, add additional prompts without waiting for the stream to finish.

Task Manager Improvements

• Introduced a queue for calls to the OllamaClient.
• Optimization to ensure non-blocking behavior while keeping optimal performance.

UI Updates

• Implemented Rendering queue for improved responsiveness.
• Option to print logs in UI.
• Removed typing effect due to the rendering buffer overflow and inconsistent rendering performance.

Security

• No longer storing passwords for enhanced security.

Bug Fixes

• Fixed read_folder not processing all files.
• Avoided double folder indexing.
• Various UI and rendering fixes.

This update improves security, stability, and task management efficiency while enhancing UI responsiveness. 🚀

Full Changelog: beta_v0.2...beta_v0.3.1

Introducing mode override with context awareness for the shell output.

In default mode type:

@shell last 20 messages from kernel log

To trigger shell generator and execute command.

or use bypass to execute command directly:

!sudo dmesg | tail -n 20

After that one can prompt chatbot about the output.

Additionally one can use:

@code override to generate code snippets without leaving default mode.

Full Changelog: beta...beta_v0.2

Unveiling DeepShell, your intelligent terminal companion. Designed for seamless AI-powered command execution, file analysis, and interactive shell operations, it bridges intent and action with precision.

Key Features:

• Context-Aware Execution – Remembers files, conversations, and intent.
• Advanced Parsing – Understands natural language commands.
• Real-Time Markup Streaming – Delivers structured responses.
• Intelligent File Handling – Reads, analyzes, and processes seamlessly.
• Interactive Shell Mode – Executes commands dynamically.