Built with support from Erutalia (erutalia.com) — a software company specializing in AI, Graph Systems, Custom Software Development, CRA solutions, and scalable backend architectures.
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.
The goal is to:
-
Fix the
/analyzeendpoint:- Ensure correct communication between Node.js and Python
- Return accurate JSON responses
-
Ensure the server runs correctly on port
3000 -
Implement a working rate limiter:
- Endpoint:
/ping - Limit: 5 requests per 10 seconds per client
- Exceeding limit should return HTTP
429
- Endpoint:
- 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
node_python_ai_task/ ├── task.toml ├── instruction.md ├── environment/ ├── solution/ └── tests/
harbor run -p "./node_python_ai_task" -a oracleAI Agent (Groq model)
harbor run -p "./node_python_ai_task" -a terminus-2 --model groq/moonshotai/kimi-k2-instruct-0905 -k 10 -n 10Functional correctness (API responses) Proper rate limiting behavior System stability Ability of AI agent to reason and fix issues
- 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.