What it does

Fabric is an open-source framework that organises AI prompts by real-world task. Each prompt is called a Pattern, and you run one from the terminal on whatever text you pipe in, for example to summarise content, pull the key points from a YouTube video or podcast, explain code or draft social media posts from a source. You can also write custom patterns of your own. The current version is written in Go. Fabric connects to AI providers that you configure during setup. The README lists native integrations such as OpenAI, Anthropic, Google Gemini, Ollama, Azure OpenAI, Amazon Bedrock and Vertex AI, and many OpenAI-compatible providers. Models can be mapped to individual patterns with environment variables. Beyond the command line, the README documents a REST API server, an Ollama compatibility mode, extensions, helper apps such as to_pdf, code2context and generate_changelog, and a web interface. It installs through a one-line script, Homebrew, AUR, Winget, Scoop, go install or Docker, and is released under the MIT License.

Capabilities

Library of task-specific prompts called Patterns, plus custom patterns of your own
Run patterns from the terminal on piped text, with streaming output
Connects to many AI providers, including local models through Ollama
REST API server and Ollama compatibility mode
YouTube transcript helper for feeding videos into patterns
Installers for macOS, Linux and Windows, plus Docker images

When to use it

You want a repeatable prompt for tasks such as summarising articles or extracting key points from videos
You work in a terminal and want to pipe text through different models
You want to keep your own prompts organised and reuse them across providers
You want to expose patterns to other tools through a local REST API