systemhealthai

AI SRE for system health triaging

As an SRE, Cloud engineer, you may have to constantly look at logs, metrics, traces to troubleshoot and triage issues to figure out why particular systems may be having issues. SystemHealthAI ( SHAI ) is an AI agent which will act as an AI SRE, to look at different data sources like prometheus, elasticsearch, cloudwatch, splunk and help triage issues and provide insights into why the system or systems might be acting up.

SHAI Architecture

!SHAI Architecture

Show Your Support ⭐

If you find SHAI useful, please consider giving it a STAR ! ⭐

Quick start

Pre-Reqs

• Install uv to run mcp servers
• OpenAI Api Key
• Datasource url for prometheus have a prometheus url ready to use
• pip or poetry

Using pip

pip install systemhealthai

From Source using poetry

git clone git@github.com:ajinkyakadam/systemhealthai.git
cd systemhealthai
poetry install -e . 

Setup

Update example_mcp_server_config.json

Ensure that the example_mcp_server_config.json file is updated with the correct configuration for your MCP servers. This assumes that the Prometheus MCP server code is already downloaded and a Prometheus endpoint is available.

1. Locate the example_mcp_server_config.json file in your project directory.
2. Open the file in a text editor of your choice.
3. Update the prometheus_endpoint field with the URL of your Prometheus server. For example:
4. Save the file and ensure the MCP server is configured to use this updated configuration.

Using SHAI

shai nodeA --model "openai:o4-mini"

The above command instructs shai to use the o4-mini model and triage the nodeA server. Please replace the nodeA with an actual hostname that you would like to find information for.

Roadmap

Datasource support

Data Source	Status	Description
Prometheus	✅	Find node metrics to correlate and triage health issues
Grafana Loki	🟡	search loki logs
Elasticsearch	🟡	search elasticsearch logs for system issues
Splunk	🟡	search splunk logs for system issues

LLM Provider Support

Provider	Status	Description
OpenAI	✅	Integrate with OpenAI models for advanced insights and triaging
Claude	🟡	Support for Claude models to assist in system health analysis
Hugging Face	🟡	Utilize Hugging Face models
Local LLMs	🟡	Deploy and use local LLMs for on-premise triaging solutions

How to Contribute

Contributions are welcome, be it bug reports, feature requests, or PRs!

• Open a github issue to report issues, or suggest features
• Open a pull request with improvements
• Share your experience and how it has been useful to you or your organization.

Citation

If you use shai in your work, blogs, projects, please do cite:

@software{systemhealthai,
  author = {Kadam, Ajinkya},
  title = {SHAI: An AI SRE for triaging system health issues},
  year = {2025},
  publisher = {GitHub},
  url = {https://github.com/ajinkyakadam/systemhealthai}
}

License

MIT