🔧 Windows Cross-Platform Build Fix

Patch Release (v0.3.2)

This patch release fixes Windows cross-compilation build errors by implementing platform-specific process management.

🐛 Bug Fixes

Windows Build Compatibility

• Fixed syscall.Kill undefined error on Windows builds
• Fixed syscall.SysProcAttr.Setpgid undefined on Windows
• Added platform-specific files: server_manager_unix.go and server_manager_windows.go
• Abstracted process group management with cross-platform helper functions
• Removed unused syscall import from server_manager.go

🔧 Technical Changes

• Implemented setupProcessGroup() helper for Unix/Windows compatibility
• Implemented killProcessGroup() helper for Unix/Windows compatibility
• Used Go build tags (//go:build !windows and //go:build windows) for platform separation
• All 6 platform builds (macOS/Linux/Windows × ARM64/AMD64) now compile successfully

📦 Installation

curl -fsSL https://raw.githubusercontent.com/electroheadfx/efx-face-manager/main/install.sh | bash

---

Full Changelog: v0.3.1...v0.3.2

💬 Interactive Chat Interface with OpenCode Integration

Release v0.3.0

Title:

💬 Interactive Chat Interface with OpenCode Integration

---

Description:

New Features Release (v0.3.0)

This release introduces a powerful interactive chat interface and OpenCode web integration, transforming efx-face-manager into a complete MLX model interaction platform.

🚀 What's New

Interactive Chat Interface

• Built-in chat UI for direct model interaction within the TUI
• Conversation history management with persistent storage
• Real-time markdown rendering with syntax highlighting (Glamour/Dracula theme)
• Multi-exchange navigation and scrolling
• Conversation management: create new, rename, delete conversations
• Smart viewport with auto-scroll and exchange-level navigation
• Visual spinner for streaming responses

OpenCode Web Integration

• Launch browser-based chat interface (OpenCode) alongside TUI
• Automatic runner script generation for seamless server integration
• One-click launch with 'o' key from server manager
• Inline execution support with 'O' key (suspend/resume TUI)
• Clipboard copy for manual execution

Enhanced User Experience

• Context-aware ESC navigation (chat history ↔ chat view)
• Keyboard shortcuts: ctrl+n (new chat), ctrl+d (delete), ctrl+r (rename)
• Exchange-level browsing with e key for focused message review
• Auto-save conversations with model/port association
• Responsive layout with dynamic viewport sizing

🔧 Technical Improvements

• New chat package with Runner, Storage, and Conversation models
• SQLite-based conversation persistence
• Markdown renderer integration with word-wrap support
• Spinner component for loading states
• Text input with focus management
• Code cleanup: removed unused width parameter from renderMessages

📦 Installation

curl -fsSL https://raw.githubusercontent.com/electroheadfx/efx-face-manager/main/install.sh | bash

🎯 Usage

From the Server Manager with a running server:

• Press c to open interactive chat
• Press o to copy OpenCode launch command
• Press O to launch OpenCode inline (suspend TUI)
• Press h to view chat history and manage conversations

---

Full Changelog: v0.2.0...v0.3.0

🎨 Complete TUI Redesign with Multi-Server Architecture

Release v0.2.0

Title:

🎨 Complete TUI Redesign with Multi-Server Architecture

---

Description:

Major Breaking Release (v0.2.0)

A complete architectural overhaul of efx-face-manager, now built with Go + Bubbletea for a modern, responsive terminal interface.

🚀 What's New

Multi-Server Management

• Run multiple models simultaneously on different ports
• Real-time scrollable logs for each server
• Quick server switching (1-9 keys) and graceful shutdown
• Independent server lifecycle management

Modern TUI Experience

• Built from scratch with Bubbletea & Lipgloss
• Live search with Hugging Face API integration
• Intuitive keyboard navigation with history stack (ESC)
• Interactive configuration panels with real-time command preview
• Pagination for 3000+ models across mlx-community & lmstudio-community

Dynamic Template System

• User-editable YAML configuration (~/.config/efx-face-manager/templates.yaml)
• 6 model type presets: lm, multimodal, image-generation, image-edit, embeddings, whisper
• Pre-configured templates for popular models (GLM-4, Qwen3-Coder, Qwen3-VL, PaddleOCR-VL, NVIDIA-Nemotron)
• Port auto-increment for conflict-free multi-server launches

Enhanced Workflows

• Single-page configuration with tab navigation
• External drive auto-detection for model storage
• Improved menu layout and action bar UI
• Better error handling and status indicators

🔧 Technical Changes

• Language: Complete rewrite from Bash to Go
• Dependencies: Bubbletea, Lipgloss, Cobra CLI framework
• Architecture: Modular codebase with separate TUI, server, model, and config packages
• Build: Multi-platform support (macOS, Linux, Windows × ARM64/AMD64)

📦 Installation

curl -fsSL https://raw.githubusercontent.com/electroheadfx/efx-face-manager/main/install.sh | bash

⚠️ Breaking Changes

This release is incompatible with v0.1.x configuration files. Previous shell-based scripts have been replaced with native Go binaries.

---

Full Changelog: v0.1.11...v0.2.0

v0.1.11

Release Notes - v0.1.11

🎯 Overview

This release introduces the GLM-4.7-Flash-8bit template, enhances installation documentation, and improves the developer experience with better environment management.

🚀 New Features

GLM-4.7-Flash-8bit Template

• New fast-launch template for the GLM-4.7-Flash-8bit language model
• Optimized configuration with vendor-recommended settings:
  • --reasoning-parser glm47_flash
  • --tool-call-parser glm4_moe
  • --message-converter glm4_moe
  • --debug mode enabled
  • Standard server settings (port 8000, host 0.0.0.0)

📦 Installation Improvements

Shell Integration

• Directory-preserving function that maintains working directory context
• Automatic environment activation without manual steps
• Seamless command execution with all arguments passed through

🛠 Technical Enhancements

User Experience

• Fixed directory switching issues - commands now return to original working directory
• Enhanced documentation with clearer installation steps
• Better error handling and troubleshooting guidance

📝 Documentation Updates

New Sections Added

• Template Models documentation with usage instructions
• Detailed installation procedures for both pipx and uv methods
• Enhanced main menu organization information

Version Tracking

• Consistent versioning across application and documentation
• Clear changelog with feature additions and fixes

🔧 Bug Fixes

v0.1.11 Specific Fixes

• Added glm47_flash reasoning parser support for GLM-4.7-Flash-8bit template
• Resolved directory context issues with mlx-openai-server commands
• Fixed alias/function implementation for better reliability

🎯 Compatibility

System Requirements

• macOS Apple Silicon (M1/M2/M3/M4)
• Python 3.11 or 3.12 (3.13+ not yet supported)
• pyenv recommended for Python version management
• uv package manager for dependency handling

Dependencies

• mlx-openai-server installed via GitHub repository
• gum, jq, huggingface-cli via Homebrew

🚀 Upgrade Instructions

For New Users

1. Follow the uv-based installation in the README
2. Add the shell function to your .zshrc
3. Install models using the new template system

For Existing Users

1. Backup your current configuration
2. Follow the new installation procedure
3. Migrate existing models to the new environment
4. Update shell configuration with the provided function

📊 Version History

v0.1.11 (Current)

• Added GLM-4.7-Flash-8bit template
• Updated installation documentation
• Improved environment management
• Fixed directory context issues

Previous Versions

• v0.1.10 - Added template menu system
• v0.1.8 - UI improvements and configuration fixes
• v0.1.5 - Bug fixes and symlink improvements

---

Release Date: January 20, 2026
Author: Laurent Marques
Repository: https://github.com/electroheadfx/efx-face-manager