Poseidon: The AI-Powered Debugging Assistant

Poseidon is an innovative AI agent designed to streamline the investigation and resolution of production outages, aims to minimize service disruption and enhance operational efficiency.

Overview

This initial version (v0.0.1) is functional and demonstrates the core concept. The roadmap includes enhancements such as a more flexible API schema (MCP) and wider applicability based on user feedback, which is highly encouraged.

Key Features

• Flexible Integration: Poseidon can manage any format of API responses from internal data sources, allowing seamless compatibility regardless of how operation records, logs, or metrics are returned.

Current Capabilities

Poseidon currently analyzes logs and API responses to diagnose issues effectively. For instance, consider a scenario where a feature flag is activated for gender-based recommendations in an application, which inadvertently leads to API failures due to missing gender data:

2025-06-05 10:00:02.004 ERROR 12345 --- [nio-8080-exec-5] c.e.controller.UserController     : Failed to read `user`.`gender`: java.lang.NullPointerException

Upon processing the relevant data, Poseidon provides insightful diagnostics:

{
    "rootCause": {
        "change": "Feature flag enabled without validating data integrity",
        "featureFlag": "genderRecommendations",
        "issue": "NullPointerException when accessing user.gender"
    },
    "service": "user-service",
    "startTime": "2025-06-05T10:00:02Z",
    "summary": "A NullPointerException occurred in UserController immediately after enabling the 'genderRecommendations' feature flag. The error happens when attempting to access user.gender attribute, which appears to be null for some users. The issue began exactly when the feature flag was enabled for A/B testing, causing 500 errors for user API requests while other services remained unaffected."
}

Future Developments

Plans for enhancement include:

1. Codebase Search: Integration of tools like greptile for efficient codebase searching.
2. Database Querying: Implementing features to search databases directly for improved troubleshooting.

Contributions

Contributions are welcome! While a CONTRIBUTING.md file is being prepared, communication via email or LinkedIn is encouraged for inquiries and feedback.