Show HN: Pipelex – Declarative language for repeatable AI workflows
6 months ago
- #AI workflows
- #automation
- #open-source
- Pipelex is developing an open standard for repeatable AI workflows.
- Users can write business logic instead of API calls.
- Install Pipelex via pip and initialize with `pipelex init`.
- API key options include a free Pipelex key, bringing your own keys (OpenAI, Anthropic, Google, Mistral), or using local AI solutions like Ollama or vLLM.
- Create AI workflows with commands like `pipelex build pipe` to generate production-ready `.plx` files.
- Example workflow: Analyze CV-job match and generate interview questions.
- Workflows are structured with domains, concepts, and pipes for clear transformations.
- Run workflows via CLI or Python, with inputs provided in JSON format.
- Install AI assistant rules to modify pipelines with natural language.
- Pipelex is open-source, with IDE extensions available for better development experience.
- Supports additional features like Anthropic, Google, Mistral AI, Amazon Bedrock, and image generation with FAL.
- Optional telemetry collects anonymous usage data to improve the product.
- Community contributions are welcome, with guidelines provided.
- Join the Discord community for support and discussions.
- Project is licensed under MIT, with runtime dependencies under their own licenses.