@theaiinc/torch-darwin-arm64

This npm package provides the Torch CLI binary for macOS ARM64 (Apple Silicon).

Installation

npm install -g @theaiinc/torch-darwin-arm64

How It Works

• The binary is not embedded in the npm package (to avoid npm size limits).
• On install, a postinstall.js script downloads the correct binary from the theaiinc/torch-binaries GitHub Releases.
• The binary is placed in bin/torch and made executable.

Manual Release Flow

1. Build the binary for your platform (e.g., torch-darwin-arm64).
2. Go to theaiinc/torch-binaries Releases.
3. Create a new release with tag vX.Y.Z (matching the npm package version).
4. Upload the binary as an asset named torch-darwin-arm64.
5. Publish the release.
6. Publish the npm package (which will now download the binary on install).

Troubleshooting

• If the binary cannot be downloaded, check that the release exists and the asset is named correctly.
• The install script will fail if the binary is missing or the network is unavailable.

Reporting Issues

For all Torch CLI issues (binary problems, features, bugs, installation, configuration, etc.), please report them in the torch repository.

License

MIT

Release Notes (2024-07-04)

• All CLI options are now visible and functional
• Image reference handling supports multiple images and multimodal models, with BLIP captioning
• Unit tests for image captioning, multimodal logic, pip install utility, and code parser edge cases
• All unit tests run in venv
• Semantic (vector) code search tool exposed for agent, using VFS
• Code parsing and chunking work on agent’s virtual filesystem (VFS)
• Dynamic, on-the-fly tree-sitter grammar installation with progress
• Reusable pip install utility and CLI
• Comprehensive edge case coverage for semantic search, VFS, and image handling
• All edge case tests pass