v1.3.2

Michael-A-Kuykendall/shimmyv1.3.2Sep 14, 2025by Michael-A-Kuykendall

AI Summary

Enhances VSCode integration for Qwen models and adds support for custom model directories across multiple platforms.

Key Highlights

Fixed VSCode extension compatibility with Qwen models
Custom model directory support via environment variables
Multi-drive support for Ollama installations on Windows
Enhanced automatic template detection for Qwen models

New Features

VSCode extension compatibility fixes
SHIMMY_MODEL_PATHS and OLLAMA_MODELS environment variables
--model-dirs global CLI option
Multi-drive Windows search for Ollama

Full Release Notes

## [1.3.2] - 2025-09-12

### 🐛 Bug Fixes

**Issue #13: VSCode Integration with Qwen Models**
- Fixed VSCode extension compatibility with Qwen3-4B-Instruct and other Qwen models
- Enhanced automatic template detection for Qwen models (now uses ChatML template)
- Added better error logging for model loading failures in OpenAI-compatible API
- Improved error handling with detailed diagnostics for troubleshooting

**Issue #12: Custom Model Directory Detection**
- Added support for custom model directories via `SHIMMY_MODEL_PATHS` environment variable
- Added support for `OLLAMA_MODELS` environment variable for Ollama model directories
- Added `--model-dirs` global command-line option for specifying custom directories
- Enhanced Windows multi-drive search for Ollama installations (C:, D:, E:, F: drives)
- Improved model auto-discovery to handle Ollama installs on different drives

### ✨ Enhancements

- **Multi-Drive Support**: Automatic scanning of common Ollama paths across multiple Windows drives
- **Template Detection**: Enhanced model template inference for Qwen, ChatGLM, Llama models
- **CLI Improvements**: New global `--model-dirs` option works with all subcommands

### 📖 Documentation

- Created improved Homebrew formula using pre-built binaries (Issue #15)
- Enhanced Windows and macOS installation guidance