I wanted to move past the 'chatbot' phase of AI and build something that actually interacts with infrastructure. The biggest challenge wasn't the LLM—it was handling state and ensuring the AI didn't hallucinate math.
I ended up using LangGraph to manage the workflow, Ansible for the 'hands' (execution), and a local SQLite DB to calculate standard deviations for anomaly detection so the LLM stays out of the math business. Curious to hear how others are handling stateful orchestration for AIOps.
I wanted to move past the 'chatbot' phase of AI and build something that actually interacts with infrastructure. The biggest challenge wasn't the LLM—it was handling state and ensuring the AI didn't hallucinate math.
I ended up using LangGraph to manage the workflow, Ansible for the 'hands' (execution), and a local SQLite DB to calculate standard deviations for anomaly detection so the LLM stays out of the math business. Curious to hear how others are handling stateful orchestration for AIOps.