Pipelex provides an open-source language and framework for building and running repeatable AI workflows, enabling users to define business logic rather than complex API calls. It targets developers, researchers, and power users who need a structured, reliable, and human-readable way to orchestrate AI tasks, offering the robustness of traditional software development combined with the flexibility of AI. The primary benefit is simplifying the creation of complex AI applications by breaking them into manageable, validated steps.

How It Works

Pipelex utilizes a declarative language (.plx files) to define AI workflows as sequences or parallel executions of "pipes." Each pipe performs a specific transformation, processing information through "Concepts," which are typed data structures with defined meanings. This modular approach allows for clear separation of concerns, making workflows easier to understand, debug, and maintain compared to monolithic prompts. The language emphasizes human readability and structured validation, ensuring reliability in AI execution.

Quick Start & Requirements

• Installation: pip install pipelex
• AI Providers: Requires API keys for AI models. Options include a free Pipelex API key (obtained via Discord) or bringing your own keys from providers like OpenAI, Anthropic, Google, Mistral, or local models via Ollama, vLLM, LM Studio, etc.
• Workflow Generation: Use pipelex build pipe "description of workflow" --output workflow.plx to generate a .plx file.
• Workflow Execution: Run via CLI (pipelex run workflow.plx --inputs inputs.json) or Python API.
• Prerequisites: Python environment, API keys for chosen AI models.
• IDE Extension: Recommended for .plx file syntax highlighting and IntelliSense, available on Open VSX Registry.
• Documentation: Links to tutorials, guides, and examples are provided.

Highlighted Details

• AI-Assisted Development: pipelex kit rules enables natural language modification of pipelines using assistants like Cursor, Claude, and GitHub Copilot.
• Repeatable Business Logic: Focuses on abstracting AI model interactions into reusable workflow components.
• Structured Data Concepts: Ensures data integrity and clarity throughout the workflow.
• Demo Videos: Showcases rapid workflow creation from concept to execution.

Maintenance & Community

Pipelex fosters a community through its Discord server for support and API key requests. Bug reports and feature requests are managed via GitHub Issues, while general questions can be posted in GitHub Discussions. Contribution guidelines are available for developers interested in participating.

Licensing & Compatibility

The project is licensed under the MIT license, permitting broad use and modification. Runtime dependencies are managed via PyPI and are subject to their respective licenses. "Pipelex" is a trademark of Evotis S.A.S.

Limitations & Caveats

Pipelex collects optional, anonymous or identified usage data for product improvement, with options to disable telemetry via the DO_NOT_TRACK environment variable. Prompts, LLM responses, file paths, and URLs are redacted before transmission. The project is a trademark of Evotis S.A.S.