Releases · Lorydima/MindCLI

[2.0]

Migrated from GPT4All (GGUF) to Ollama for AI model management
Ollama model download directly from the application interface
Ollama model delete with automatic fallback to another model
Installed models list with file size display
Tavily web search integration for real-time web context
Agent Mode — create and edit files on your PC using AI
Memory Management — add, view, remove, and clear AI memories
Chat Management — list, open, and remove saved chat files
Chats saved in a dedicated chats/ directory instead of Desktop
Per-model configuration (base prompt, parameters, device) stored in separate JSON files
New configuration system with multiple files (paths.json, hardware.json, tavily_API.json, memory.json)
GPU detection via nvidia-smi and per-model device selection (CPU/GPU)
Support for additional file types: .docx, .xlsx, .md, .py, .cpp, images (.png, .jpg, .gif, etc.)
Rich typing effect with Live rendering for markdown and code blocks
Clipboard copy with error handling for Windows
Windows-style masked input for API keys
Signal handlers (SIGINT, SIGTERM, SIGHUP) for graceful Ollama shutdown
num_ctx parameter support alongside standard AI parameters
Model parameters normalization with backward compatibility for v1.0 key names (max_tokens → num_predict, temp → temperature)
Code reorganization into modular package structure under src/mindcli/

AI Engine: Replaced GPT4All with Ollama API
Configuration: Single config.json → multiple config files in configs/ directory
Project Structure: Reorganized from flat modules to a structured package with separate modules for state, utils, config, Ollama, file handling, web search, memory, chat, model management, and UI
Chat Save: Now saves to chats/ folder instead of Desktop
File Attachment: Extended from .txt/.pdf only to 20+ file types including Office documents and images
Typing Effect: Rewritten with Rich Live rendering for smoother markdown and code display
Command Interface: Expanded main menu with additional subcommands for memory, chats, models, and Tavily API management
Requirements: Replaced gpt4all dependency with ollama, added tavily-python, python-docx, openpyxl

Model file path resolution — v2.0 uses Ollama model names directly instead of GGUF file paths
Chat history management — history is now cleared when switching models
Ollama process lifecycle — proper start, detection, and kill across platforms
Better error handling for missing Ollama installation with download links displayed

Platform: Windows and Linux