beddel

What This Skill Does

The Beddel skill provides a robust, declarative YAML-based engine for orchestrating complex AI agent workflows. It moves beyond simple prompt chains, allowing users to define multi-step pipelines that include branching logic, advanced guardrails for validation, automatic retry mechanisms, and native OpenTelemetry tracing. By using Beddel, you can treat your AI automation logic as code, ensuring version control and reproducibility across your projects. It supports multiple LLM providers via LiteLLM, making it highly flexible for switching between different models like Gemini, GPT-4, or open-source variants without rewriting your business logic.

Installation

To start using the Beddel skill within the OpenClaw environment, ensure you have an environment that meets the requirements, specifically Python 3.11+. Install the skill using the following command in your terminal:

clawhub install openclaw/skills/skills/botanarede/beddel

Once installed, you must configure your environment variables to point to your LLM provider. For example, to use Gemini, run export GEMINI_API_KEY="your-key" in your session. After configuration, you can immediately begin defining your .yaml workflows and executing them via the OpenClaw command interface.

Use Cases

Beddel is ideal for complex, multi-step LLM tasks that require more than a single inference call. Common use cases include:

• Building multi-agent orchestrators where one workflow calls another.
• Creating data validation pipelines where output from one LLM must pass specific guardrails before proceeding.
• Automating document processing tasks involving summarization, extraction, and formatting.
• Developing self-correcting agent chains that retry failed tasks automatically.

Example Prompts

1. "Run the workflow defined in process_docs.yaml and pass the input folder path as a parameter."
2. "Validate the syntax of my current workflow pipeline file to ensure all steps are correctly configured."
3. "Create a new multi-step LLM workflow that summarizes an article and then generates a JSON output for a database."

Tips & Limitations

• Always test your workflows using the beddel validate command to catch syntax errors early.
• Utilize the execution_strategy block to define how your pipeline should handle failures; silent failure is rarely desired in production.
• Remember that Beddel requires Python 3.11; ensure your system environment is not defaulting to an older 3.x version when executing scripts.
• Keep your configuration clean by separating credentials from your workflow YAML definitions using environment variables.