@pheem49/mint 1.4.1 → 1.4.2

package/README.md

@@ -5,50 +5,48 @@
 </p>
 
 <p align="center">
-  <strong>Desktop assistant + CLI coding agent built with Electron, Node.js, and modern LLM providers.</strong>
+  <strong>The Unified AI Desktop Assistant & Agentic Coding CLI. Built for speed, power, and local control.</strong>
 </p>
 
 <p align="center">
   <img src="https://img.shields.io/badge/License-AGPL--3.0-blue?style=for-the-badge" alt="License">
   <img src="https://img.shields.io/badge/Node.js-LTS-green?style=for-the-badge&logo=node.js" alt="Node.js">
   <img src="https://img.shields.io/badge/Electron-40.x-47848F?style=for-the-badge&logo=electron" alt="Electron">
-  <img src="https://img.shields.io/badge/CLI-Agentic-orange?style=for-the-badge" alt="CLI Agentic">
+  <img src="https://img.shields.io/badge/CLI-Unified_Agent-orange?style=for-the-badge" alt="CLI Agentic">
 </p>
 
-Mint is an AI assistant that runs in two main surfaces:
+Mint is an advanced AI assistant designed to live in your workspace. It features a rich Electron desktop interface for day-to-day assistance and a powerful, unified CLI agent that seamlessly merges conversational chat with complex coding tasks.
 
-- A desktop Electron app for chat, screen-aware help, and system actions
-- A terminal-first CLI with a workspace-aware coding agent
+## 🌟 What's New in v1.4.1
 
-The project is built around practical local workflows: inspect code, edit files, run safe commands with approval, use project context, and fall back across multiple AI providers when needed.
+- **Unified Agent Loop:** No more switching modes. Every interaction in the CLI is now handled by a sophisticated agent that can think, plan, and execute tools autonomously.
+- **Agentic Web Search:** Mint can now search the internet in real-time to answer questions with the latest information using integrated web tools.
+- **Upgraded TUI:** A redesigned terminal interface featuring a Braille spinner, thinking timer, and a cleaner "✓ ActionName" logging style.
+- **Interactive "Ask User":** The agent can now pause and ask you for clarification or preferences during complex multi-step tasks.
+- **Enhanced System Control:** Reliable multi-fallback support for opening files, folders, and applications on Linux (Pop!_OS/Ubuntu), macOS, and Windows.
 
-## What Mint Can Do
+## 🚀 Key Features
 
-- Chat in Desktop or CLI mode
-- Route coding requests into Code Mode automatically from the CLI chat UI
-- Inspect a workspace before editing
-- Search code, read file ranges, and patch files
-- Run non-destructive shell commands with user approval
-- Keep lightweight per-workspace session memory
-- Perform a second-pass reviewer step in Code Mode
-- Execute structured actions such as opening apps, URLs, folders, and system tasks
-- Support multiple providers: Gemini, Anthropic, OpenAI, local OpenAI-compatible endpoints, Ollama, and Hugging Face
+### 💻 Unified CLI Agent
+Mint CLI is not just a chat wrapper; it's a full agentic workflow.
+- **Think & Plan:** Every response starts with a reasoning phase where Mint plans its next move.
+- **Autonomous Tools:** `web_search`, `list_files`, `read_file`, `search_code`, `run_shell`, `apply_patch`, `write_file`, `open_folder`, and more.
+- **User-in-the-Loop:** Safety first. Mint asks for your approval before running shell commands or making file edits.
+- **Workspace Aware:** Automatically understands your project structure, git status, and testing framework.
 
-## Current Agent Capabilities
+### 🖥️ Desktop Assistant
+- **Screen Vision:** Capture and analyze your screen for instant help with what you're looking at.
+- **Real-time Translation:** Instantly translate text from your screen into Thai or English.
+- **Proactive Engine:** Mint monitors your system events to provide helpful suggestions before you even ask.
+- **System Tray & Floating Widgets:** Quick access to Mint from anywhere on your desktop.
 
-Mint CLI is an agentic coding workflow, not just a chat wrapper.
+### 🛠️ Multi-Provider Support
+Mint supports the latest LLMs and local backends:
+- **Cloud:** Gemini 1.5/2.0 Pro & Flash, Anthropic Claude 3.5, OpenAI GPT-4o.
+- **Local:** Ollama, LM Studio, Hugging Face Inference API.
+- **MCP:** Full support for Model Context Protocol to extend Mint's capabilities with external tools.
 
-In Code Mode it can:
-
-- Decide on the next step from the current observation
-- Use tools in a loop: `list_files`, `read_file`, `search_code`, `run_shell`, `apply_patch`, `write_file`
-- Observe tool output and continue iterating
-- Stop with a summary and verification result
-- Ask for approval before shell commands and file changes
-
-That makes Mint a practical CLI coding agent, while still keeping the user in control of risky actions.
-
-## Screenshots
+## 📸 Screenshots
 
 <p align="center">
   <img src="assets/Agent_Mint.png" alt="Mint Desktop UI" width="48%">
@@ -59,186 +57,97 @@ That makes Mint a practical CLI coding agent, while still keeping the user in co
   <img src="assets/CLI_Screen.png" alt="Mint CLI" width="100%">
 </p>
 
-## Installation
-
-### Global install
+## 📦 Installation
 
+### Global Install
 ```bash
 npm install -g @pheem49/mint@latest
 ```
 
-### Local development
-
+### Local Development
 ```bash
 git clone https://github.com/Pheem49/Mint.git
 cd Mint
 npm install
 ```
 
-## Quick Start
+## ⚡ Quick Start
 
-### Run the desktop app
+1. **Setup Mint:**
+   ```bash
+   mint onboard
+   ```
+2. **Start Chatting:**
+   ```bash
+   mint
+   ```
+3. **Run the Desktop App:**
+   ```bash
+   npm start
+   ```
 
-```bash
-npm start
-```
+## ⌨️ CLI Commands
 
-### Run the CLI
+- `mint` / `mint chat` : Start the unified interactive agent UI.
+- `mint code "<task>"` : Run a specific coding task in the current workspace.
+- `mint task "<task>"` : Queue a background task for the headless agent.
+- `mint mcp` : Manage Model Context Protocol (MCP) servers.
+- `mint list` : Display all available features and commands.
 
-```bash
-mint
-```
+## 🔌 MCP Management (Extensions)
 
-### First-time setup
+Mint supports the **Model Context Protocol (MCP)**, allowing you to extend its capabilities via the CLI without manual config editing.
 
+### Add a New Server
 ```bash
-mint onboard
-```
-
-## CLI Commands
-
-- `mint` or `mint chat`
-Start the interactive chat UI.
-
-- `mint code "<task>"`
-Run the workspace-aware coding agent in the current directory.
-
-- `mint task "<task>"`
-Queue a background task for the headless agent.
+# Template
+mint mcp add <name> <command> --args <args...> --env <KEY=VALUE>
 
-- `mint agent`
-Run the background headless agent.
+# Example: Google Search
+mint mcp add google-search npx --args -y @modelcontextprotocol/server-google-search --env GOOGLE_API_KEY=your_key GOOGLE_SEARCH_ENGINE_ID=your_id
 
-- `mint list`
-Show major features and commands.
-
-- `mint onboard`
-Configure API keys and local settings.
-
-## CLI Examples
-
-### Interactive chat
-
-```bash
-mint
+# Example: Filesystem Access
+mint mcp add my-files npx --args -y @modelcontextprotocol/server-filesystem /path/to/folder
 ```
 
-Then ask naturally:
-
-```text
-สำรวจโปรเจคนี้ให้หน่อย
-แก้บัคใน CLI ตัวนี้และรันเทสต์
-open github
+### List Configured Servers
+```bash
+mint mcp list
 ```
 
-### One-shot coding task
-
+### Remove a Server
 ```bash
-mint code "fix the failing tests and verify the result"
+mint mcp remove google-search
 ```
 
-### Background task
-
+### Clear All Servers
 ```bash
-mint task "inspect the repo and summarize the top 3 risks"
+mint mcp clear
 ```
 
-## Code Mode Workflow
-
-Mint Code Mode is designed around an inspect -> act -> verify loop.
-
-What it does well right now:
-
-- Understand the current workspace path
-- Read git status and diff summary
-- Suggest verification commands from `package.json`
-- Apply focused patches instead of blind rewrites
-- Keep a workspace session summary for future tasks
-
-What it deliberately does not do:
-
-- Run destructive commands like `rm -rf` or `git reset --hard`
-- Edit files outside the current workspace
-- Execute shell edits without approval
-
-## Desktop Features
-
-- Chat window with custom UI
-- Settings window for provider and behavior configuration
-- System tray support
-- Proactive suggestion loop
-- Silent screen capture for analysis
-- Screen translation support
-- Floating widget / overlay UI elements
-
-## AI Providers
-
-Mint supports multiple providers and local backends.
-
-- `gemini`
-- `anthropic`
-- `openai`
-- `local_openai`
-- `ollama`
-- `huggingface`
-
-For CLI Code Mode, Mint currently behaves best with:
-
-- `gemini`
-- `anthropic`
-- `openai`
-- `local_openai`
-
-## Project Structure
+## 🏗️ Project Structure
 
 ```text
 Mint/
 ├── src/
-│   ├── AI_Brain/          # Provider integration, prompts, memory, orchestration
-│   ├── Automation_Layer/  # App/file/browser actions
-│   ├── CLI/               # Chat router, code agent, TUI support, onboarding
-│   ├── Plugins/           # Docker, Spotify, calendar, system monitor, MCP
-│   ├── System/            # Config, notifications, daemon, automation, task queue
-│   └── UI/                # Electron renderer files
-├── tests/                 # Jest tests
-├── mint-cli.js            # Main CLI entry
-├── mint-cli-logic.js      # CLI action executor
-├── main.js                # Electron main process
+│   ├── AI_Brain/          # Gemini API, Unified Agent Client, Memory Store
+│   ├── Automation_Layer/  # File Ops, Browser Automation, System Control
+│   ├── CLI/               # TUI logic, Chat Router, Code Agent
+│   ├── Plugins/           # MCP Manager, System Monitor, Third-party integrations
+│   ├── System/            # Config, Notifications, Daemons
+│   └── UI/                # Electron Renderer, Settings, Widgets
+├── mint-cli.js            # Main CLI entry point
 └── package.json
 ```
 
-## Development
-
-### Run tests
-
-```bash
-npm test -- --runInBand
-```
-
-### Watch tests
-
-```bash
-npm run test:watch
-```
-
-### Build Linux packages
-
-```bash
-npm run build:linux
-```
-
-See [BUILD_AND_RELEASE.md](/home/pheem49/vscode/Project/Mint/BUILD_AND_RELEASE.md) for Linux packaging and release notes workflow.
-
-## Security Notes
-
-- API keys are stored in local config, not in source files
-- Code Mode asks for approval before shell commands and file edits
-- Workspace path resolution blocks writes outside the active workspace
-- Several shell-based execution paths have been hardened to use argument-based process execution
-
-This is still an actively evolving project. Review permissions and local configuration before using Mint against sensitive files or production systems.
+## 🛡️ Security & Privacy
+- **Local Control:** Mint prioritizes local execution and user privacy.
+- **Approval System:** No destructive command or file change happens without your explicit `y/n` confirmation.
+- **Secure Config:** API keys are stored locally on your machine and never transmitted outside of the chosen AI provider.
 
-## License
+## 📜 License
+Mint is licensed under the **GNU Affero General Public License v3.0**.
+See the [LICENSE](LICENSE) file for more details.
 
-Mint is licensed under the GNU Affero General Public License v3.0.
-See [LICENSE](/home/pheem49/vscode/Project/Mint/LICENSE).
+---
+<p align="center">Made with 💚 by <a href="https://github.com/Pheem49">Pheem49</a></p>

package/mint-cli-logic.js

@@ -2,9 +2,6 @@
 const { openApp } = require('./src/Automation_Layer/open_app');
 const { openWebsite, openSearch } = require('./src/Automation_Layer/open_website');
 const { createFolder, openFile, deleteFile, findPath } = require('./src/Automation_Layer/file_operations');
-const { indexFile } = require('./src/AI_Brain/knowledge_base');
-const SystemAutomation = require('./src/System/system_automation');
-const pluginManager = require('./src/Plugins/plugin_manager');
 
 async function executeAction(action) {
     if (!action || action.type === 'none') return null;
@@ -24,11 +21,9 @@ async function executeAction(action) {
                 createFolder(action.target);
                 return `Created folder: ${action.target}`;
             case 'open_file':
-                await openFile(action.target);
-                return `Opening: ${action.target}`;
             case 'open_folder':
-                await openFile(action.target);
-                return `Opening folder: ${action.target}`;
+                const res = await openFile(action.target);
+                return res === true ? `Opening: ${action.target}` : res;
             case 'delete_file':
                 await deleteFile(action.target);
                 return `Deleted: ${action.target}`;