Impala AI query profile analyzer

The Impala AI Query Profile Analyzer evaluates query performance in the Coordinator Web UI, surfacing bottlenecks and root causes such as join skew, memory pressure, and data spills.

The Impala AI Query Profile Analyzer processes query profiles directly in the Coordinator Web user interface (UI) to identify performance issues, summarize runtime data, and provide optimization suggestions. Impala query profiles track query planning, execution phases, and resource consumption. However, these profiles can grow very large and highly technical. Impala AI Query Profile Analyzer evaluates complex performance metrics such as scanner wait times, join skew, memory pressure, cardinality issues, or data spills.

To keep your experience smooth and secure, the system uses a multi-round tool-calling architecture consisting of the following components:

• Profile redaction component — Replaces sensitive identifiers, such as user names, IP addresses, hostnames, table names, and columns, with secure, consistent aliases before sending data to the large language model (LLM).
• Profile parsing and tool call interface — Breaks raw profile JSON into small segments, allowing the AI agent to request specific pieces, such as timeline details or scan node summaries, as needed.
• AI endpoint communication layer — Connects to your AI models using your existing Impala AI Functions setup. The selected LLM endpoint must support the OpenAI API format for requests and OpenAI-style tool calling. Before displaying the final analysis, this layer restores your original names and identifiers.