Fabric

Fabric — Open-source framework for augmenting humans with AI

Fabric is a practical, pattern-driven framework that treats prompts ("Patterns") as first-class, reusable units for solving real-world tasks with large language models and other AI providers. It aims to make AI integration easier by collecting high-quality, task-focused prompts, packaging them with metadata and helper tools, and exposing them through a CLI, REST API, and web interface for both personal and programmatic use.

Key concepts and features

• Patterns: Reusable prompt templates organized by task (summarization, code review, extract wisdom from video, etc.). Patterns are stored as markdown files and can be customized or extended. Patterns can be run locally or via the Fabric REST API.

• Multi-supplier model support: Fabric integrates with a variety of model vendors and providers (OpenAI, Anthropic, Gemini, Bedrock, and more), allowing per-pattern model mapping and vendor selection. It includes helpers for model configs, token settings, and vendor fallbacks.

• CLI-first design: The primary interface is a command-line tool (fabric) that supports streaming output, pattern variables, model selection, attachments, YouTube transcript extraction, and many convenience flags for developer workflows.

• REST API & Web UI: Fabric can run as a service (fabric --serve) exposing endpoints for chat completions, pattern management, session/context handling, and YouTube transcript endpoints. A web GUI provides a friendly way to browse and run patterns.

• Extensibility & helper apps: Supports extensions/plugins inside patterns, helper binaries (e.g., to_pdf, code_helper), Docker images, and custom pattern directories so users can maintain private patterns safe from updates.

• Integrations & workflows: Built-in support for YouTube transcript and comments extraction, image generation, TTS, model thinking/reasoning controls, OAuth for vendors, shell completions, and tooling for saving outputs to note systems (Obsidian), CI/CD, or notebooks.

• Internationalization & accessibility: Fabric has been extended with i18n support and language variants, making patterns and setup prompts available in multiple languages.

Typical use cases

• Quickly run a curated summarization or analysis workflow on an article, YouTube video, or repository.
• Maintain a personal library of proven prompts (Patterns) to ensure consistent outputs across tools and teammates.
• Use Fabric as a backend service to integrate pattern-driven AI capabilities into other applications via REST.
• Prototype multi-step AI workflows (stitches) that chain local models and cloud models for hybrid processing.

Installation & running

• One-line installer (curl | bash) or platform package managers (Homebrew, AUR, winget). Binaries and Docker images are provided for easy deployment.
• From source: go install github.com/danielmiessler/fabric/cmd/fabric@latest and run fabric --setup.
• Run patterns from stdin or with flags, e.g. pbpaste | fabric --pattern summarize or fetch YouTube transcript fabric -y <youtube-url> --pattern extract_wisdom.

Extensibility & customization

• Custom patterns: Keep personal patterns in a separate directory to avoid overwrite during updates; custom patterns take precedence over built-ins.
• Per-pattern model mapping: Configure specific vendors/models per pattern via environment variables.
• Extensions: Fabric supports internal plugins that can be invoked from patterns. Helper apps provide additional pre-/post-processing.

Community, license and project health

• Open-source MIT license.
• Active repository with a large star count and many contributors; the project is maintained publicly with a changelog and frequent releases.
• Primary maintainers include Daniel Miessler and several core contributors; the repo provides examples, docs, and video introductions to help new users onboard.

Notes

Fabric is designed less as a single LLM or model and more as an orchestration and organization layer for prompts and AI-assisted workflows — suitable for power users who want consistent, reproducible AI patterns across tools and teams.