Mainframe AI Assistant

Multi-agent system powered by ERNIE 4.0 that parses mainframe JCL, diagnoses errors, and generates Zowe CLI commands to fix production issues automatically.

Comment

Inspiration

What it does

How we built it

Challenges we ran into## đź’ˇ Inspiration

Mainframe systems (IBM Z) power 70% of global financial transactions and run critical infrastructure for banks, governments, and Fortune 500 companies. When JCL (Job Control Language) jobs fail, engineers spend hours decoding cryptic error messages, searching through documentation, and manually constructing fix commands.

As a DevOps engineer working with enterprise systems, I've seen how a single JCL error can delay production deployments and cost organizations thousands of dollars per hour. I wanted to build an AI assistant that could instantly parse complex JCL, diagnose issues, and generate executable fix commands—turning hours of troubleshooting into seconds.

🛠️ What it does

Mainframe AI Assistant is a multi-agent system powered by ERNIE 4.0 that automates mainframe troubleshooting:

  1. Parser Agent: Extracts JCL structure—job names, step names, programs (PGM=), datasets (DD statements), DISP parameters, and return codes
  2. Analyzer Agent: Diagnoses issues like missing SYSIN datasets, invalid DISP allocations, space problems, and configuration risks
  3. Commander Agent: Generates complete Zowe CLI workflows to fix problems—download JCL, apply corrections, upload, submit jobs, and monitor status

Example workflow:

  • User pastes JCL with error RC=12
  • Parser identifies missing SYSIN dataset
  • Analyzer explains the root cause and suggests fix
  • Commander outputs: zowe files upload ftds fixed.jcl "USER.JCL(TESTJOB)"

🏗️ How we built it

Architecture:

  • Backend: Python FastAPI with /analyze REST endpoint
  • Agents: 3 sequential ERNIE 4.0 API calls using erniebot SDK
  • Frontend: Vanilla HTML/CSS/JavaScript (no frameworks)
  • Deployment: Local server + GitHub Pages for demo

Multi-agent design:

Accomplishments that we're proud of

What we learned

What's next for Mainframe AI Assistant

Built With

Share this project:

Updates