Show HN: Argus – Self-hosted observability that replaces dashboards with a chat (github.com)

Argus, a new open-source observability platform, has been launched, combining the capabilities of traditional monitoring tools like Datadog with an AI-driven chat interface. This innovative platform allows users to interact with their observability data through natural language questions instead of relying on complex query languages or customized dashboards. Argus autonomously investigates anomalies using a ReAct reasoning loop with access to over 18 tools, proposing fixes that require user approval before execution. This human-in-the-loop approach helps ensure that critical changes are supervised while providing detailed insights into system performance. The significance of Argus for the AI/ML community lies in its seamless integration of full-stack observability—covering logs, metrics, traces, errors, and security—within a single platform. Its LLM-agnostic design allows users to switch between different large language model providers, such as OpenAI and Google Vertex AI, simply by changing configuration settings. Additionally, Argus implements cost controls with token budgets to cap spending on AI processing, making it an attractive option for organizations looking to harness AI without the risk of unlimited expenses. This development could reshape how DevOps teams monitor infrastructure, promoting a more efficient, conversational approach to observability.