Error Boundaries

Detect, isolate, and recover from execution anomalies.

Overview

When Arivu hits an unrecoverable state (e.g., an LLM parsing failure, or unreachable database host), the pipeline delegates control to the error_boundary node, which logs precisely to this dashboard pane.

Stack Trace Preservation

Full Python stack traces mapped to the specific failing LangGraph node.

Anomaly Categorization

Differentiates between schema extraction issues, LLM rate limits, and semantic errors.

Common Error Types

SQL Syntax Failure

The LLM generated invalid SQL that exceeded the maximum retry limit (3 attempts). The trace shows all correction attempts before the final failure.

Rate Limiting (429)

Your LLM provider returned a rate-limit error. Consider switching to a provider with higher quotas or enabling semantic caching to reduce call volume.

Database Unreachable

The db_execution node could not connect to the target database. Verify your connection string and firewall rules.

Errors are logged with full context (question, generated SQL, node, and Python traceback) to enable quick root-cause analysis.