Skip to main content
FlareDB Skills give AI coding agents the knowledge to build, test, debug, and run Apache Beam pipelines. Each skill focuses on a specific task, and agents can combine them when needed.

​
Available Skills

​
How They Work Together

The skills complement each other, allowing an agent to handle a task from pipeline development through execution and debugging. For example, if you ask an agent to analyze thousands of log files:
  1. beam-pipeline designs a Python Beam pipeline to process the files and produce a concise summary.
  2. beam-streaming provides the guidance on how to write a streaming pipeline correctly.
  3. flaredb-skills manages the local FlareDB environment and runs the pipeline after you approve the execution plan.
  4. beam-debugging helps diagnose failures or incorrect results if something goes wrong.

​
Example

Once the skills are installed, simply describe your task to your coding agent:
Analyze all the CSV files in this directory, identify schema inconsistencies, and report the affected files with a summary of the issues.
The agent can use beam-pipeline to construct the processing pipeline and flaredb-skills to run it on FlareDB, rather than reading every file into its context. The goal is simple: let AI agents process large amounts of data efficiently using Apache Beam and FlareDB, while keeping only the relevant results in their context.

​
Review

Before running a pipeline on FlareDB, the agent presents a pipeline brief for your review and approval. This lets you verify what the pipeline will do before the agent executes the pipeline on FlareDB Once approved, the agent can start FlareDB, execute the pipeline, inspect the job logs, and verify the results. Note: Sometimes you agent harness might prevent your agents from running CLI commands, in that case please run the commands for your agents.

​
Install the skills

Make sure Node.js and npx are available, then run:
The installer will guide you through these choices:
  1. Select skills — select the skills you want to install. You can install all four:
    • beam-pipeline
    • flaredb-skills
    • beam-debugging
    • beam-streaming
  2. Select coding agents — choose the agents you use, such as Codex, Cursor, Gemini CLI, GitHub Copilot or others shown by the installer.
  3. Choose installation scope — select Global to make the skills available across projects, or choose another scope if offered and preferred.
  4. Confirm installation — review the summary and approve the installation.
The installer should report that installation is complete.
Last modified on October 17, 2026