diff --git a/docs/docs/Support/macos-support-matrix.mdx b/docs/docs/Support/macos-support-matrix.mdx new file mode 100644 index 0000000000..670ef75301 --- /dev/null +++ b/docs/docs/Support/macos-support-matrix.mdx @@ -0,0 +1,104 @@ +--- +title: macOS Support +slug: /macos-support +--- + +Langflow supports both Apple Silicon (ARM64) and Intel (x86_64) Macs, but ML and AI features are only available on Apple Silicon due to hardware and dependency constraints. + +Intel Macs lack the Neural Engine, Metal 3 GPU acceleration, and unified memory architecture that Apple Silicon uses to run ML workloads efficiently. + +PyTorch, a popular AI/ML library, dropped support for macOS Intel (x86_64) starting with version 2.3.0 in April 2024. The last compatible release was PyTorch 2.2.2, which only supports Python versions 3.10–3.12. Many Langflow components depend on PyTorch, and are automatically excluded on Intel Macs to prevent installation failures. + +## Quick reference + +Core Langflow features and OCR work on both architectures. +ML/AI components and GPU acceleration are Apple Silicon only. + +| Feature category | Apple Silicon (M1/M2/M3) | Intel (x86_64) | +|-----------------|--------------------------|----------------| +| **Core Langflow** | Full support | Full support | +| **ML/AI Components** | Full support | Limited/Unavailable | +| **GPU acceleration** | Metal support | Not available | +| **Native OCR** | Full support | Full support | + +## Python version support + +All supported Python versions work on Apple Silicon. +On Intel Macs, Python 3.13 and later lose PyTorch-dependent features like ALTK, HuggingFace, EasyOCR, and Docling. + +| Python version | Apple Silicon | Intel Mac | +|---------------|---------------|-----------| +| 3.10 | Supported | Supported | +| 3.11 | Supported | Supported | +| 3.12 | Supported | Supported | +| 3.13 | Supported | Limited (no PyTorch) | +| 3.14 | Supported | Limited (no PyTorch) | + +## Langflow feature support + +Feature support varies by Mac architecture. The following sections describe what works on each. + +### Core functionality (both architectures) + +The following features work on **both Apple Silicon and Intel Macs**: + +- Flow builder and visual editor +- API server and endpoints +- Database operations (SQLite, PostgreSQL) +- Authentication and user management +- All non-ML components (text processing, API calls, data transformations) +- Native OCR via `ocrmac` (uses macOS Vision framework) +- Vector stores and embeddings (when using API-based providers) +- LangChain integrations (non-ML components) + +### ML/AI features (Apple Silicon only) + +The following features require **Apple Silicon (ARM64)** and are **not available on Intel Macs**: + + +- **ALTK (Agent Lifecycle Toolkit)** - Requires PyTorch +- **HuggingFace Transformers** - Requires PyTorch +- **Sentence Transformers** - Requires PyTorch +- **EasyOCR** - Requires PyTorch +- **Docling (binary processing)** - Requires PyTorch + - Note: `docling-core` (metadata-only) works on Intel +- **MLX and MLX-VLM** - Apple Silicon exclusive framework +- **Metal GPU acceleration** - Requires Apple Silicon hardware +- **CUGA** - Apple Silicon only + +If you need ML features on an Intel Mac, consider using API-based providers such as HuggingFace, or a remove inference server with GPU support. + +For some workloads, you can use Docker with ARM64 images with Rosetta 2: + +```bash +docker run --platform linux/arm64 langflowai/langflow:latest +``` + +## Installation recommendations + +Installation instructions vary depending on your Mac architecture. + +For Apple Silicon users, install Langflow normally with all features: + +```bash +pip install langflow +``` + +Or with specific ML extras: + +```bash +pip install langflow[altk,langchain-huggingface,easyocr] +``` + +For Intel Mac users, install the base package for core functionality. +Attempting to install ML-dependent extras on Intel Macs will be automatically skipped due to platform markers in the package configuration. + +```bash +pip install langflow +``` + +## See also + +For troubleshooting, see [Troubleshoot Intel Mac issues](./troubleshooting.mdx#intel-mac-issues). + +For more information, ask the Langflow community in [GitHub Discussions](https://github.com/langflow-ai/langflow/discussions) or the [Discord server](https://discord.gg/EqksyE2EX9). \ No newline at end of file diff --git a/docs/docs/Support/release-notes.mdx b/docs/docs/Support/release-notes.mdx index 380520d4bf..ffca065c82 100644 --- a/docs/docs/Support/release-notes.mdx +++ b/docs/docs/Support/release-notes.mdx @@ -74,6 +74,10 @@ For all changes, see the [Changelog](https://github.com/langflow-ai/langflow/rel For setup instructions and configuration details, see [Manage vector data](/knowledge). +- macOS support matrix + + A new [macOS support](./macos-support-matrix.mdx) page documents feature availability across Apple Silicon and Intel Macs. + ### Deprecations - Voice mode is removed diff --git a/docs/docs/Support/troubleshooting.mdx b/docs/docs/Support/troubleshooting.mdx index 3b0f6c40a9..3a147528a3 100644 --- a/docs/docs/Support/troubleshooting.mdx +++ b/docs/docs/Support/troubleshooting.mdx @@ -631,6 +631,41 @@ If you're using the `LANGFLOW_COMPONENTS_PATH` environment variable and componen If you continue to experience issues, please [report them on GitHub](https://github.com/langflow-ai/langflow/issues) with details about your directory structure and component setup. +## Intel Mac issues {#intel-mac-issues} + +The following issues can occur when running Langflow on Intel-based Macs. + +### "No matching distribution found" errors on Intel Mac + +If you see errors about missing packages after installing Langflow on an Intel Mac, such as: + +```text +ERROR: Could not find a version that satisfies the requirement torch +ERROR: No matching distribution found for torch +``` + +This is expected behavior. ML packages such as PyTorch are intentionally excluded on Intel Macs because PyTorch dropped support for macOS x86_64 starting with version 2.3.0. + +To verify your architecture and understand which features are available, run: + +```bash +python -c "import platform; print(platform.machine())" +# x86_64 = Intel, arm64 = Apple Silicon +``` + +Use core Langflow features or API-based alternatives instead of local ML models. +For details on which features are available and workarounds, see [macOS support](./macos-support-matrix.mdx). + +### Performance issues on Intel Mac + +If Langflow feels slow on an Intel Mac: + +1. Ensure you are not trying to run ML workloads locally, as these are not supported on Intel Macs. +2. Use API-based providers for embeddings and completions instead of local models. +3. Consider upgrading to Apple Silicon for better performance with ML features. + +For more information, see [Workarounds for Intel Mac users](./macos-support-matrix.mdx#workarounds-for-intel-mac-users). + ## See also - [Langflow GitHub Issues and Discussions](/contributing-github-issues) diff --git a/docs/sidebars.js b/docs/sidebars.js index 48cb2c8db4..5703cb3cf6 100644 --- a/docs/sidebars.js +++ b/docs/sidebars.js @@ -578,6 +578,11 @@ module.exports = { id: "Support/troubleshooting", label: "Troubleshoot", }, + { + type: "doc", + id: "Support/macos-support-matrix", + label: "macOS support", + }, { type: "doc", id: "Support/contributing-github-issues",