Skip to content

Latest commit

 

History

2 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

local/node_python_ai_task

Node + Python AI Debugging Task (Terminal Bench)


Built with support from Erutalia (erutalia.com) — a software company specializing in AI, Graph Systems, Custom Software Development, CRA solutions, and scalable backend architectures.


Overview

This task simulates a real-world DevOps / Backend engineering scenario where an AI agent must debug and fix a broken microservice.

The system consists of:

  • A Node.js API server
  • A Python-based processing module
  • Inter-process communication between Node and Python
  • A rate-limited endpoint

The application is intentionally broken and requires multi-step reasoning to fix.


Objective

The goal is to:

  1. Fix the /analyze endpoint:

    • Ensure correct communication between Node.js and Python
    • Return accurate JSON responses
  2. Ensure the server runs correctly on port 3000

  3. Implement a working rate limiter:

    • Endpoint: /ping
    • Limit: 5 requests per 10 seconds per client
    • Exceeding limit should return HTTP 429

Key Challenges

  • Debugging inter-process communication (Node <-> Python)
  • Fixing incorrect execution methods
  • Understanding middleware behavior in Express
  • Implementing and verifying rate limiting logic
  • Ensuring correct API responses under constraints

Project Structure

node_python_ai_task/ ├── task.toml ├── instruction.md ├── environment/ ├── solution/ └── tests/


How to Run

Oracle (solution validation)

harbor run -p "./node_python_ai_task" -a oracle

AI Agent (Groq model)

harbor run -p "./node_python_ai_task" -a terminus-2 --model groq/moonshotai/kimi-k2-instruct-0905 -k 10 -n 10

Evaluation Criteria

Functional correctness (API responses) Proper rate limiting behavior System stability Ability of AI agent to reason and fix issues

Notes

  • The environment does NOT contain the solution
  • Tests are hidden from the agent
  • The agent must infer and fix issues independently

Powered by Erutalia (erutalia.com) — building next-generation AI systems, graph-driven platforms, and scalable custom software solutions.

About

A collection of containerized tasks for benchmarking LLM agents in terminal environments.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages