System Schema - Configuration and AI Management

Access: Admin and service roles only

`prompt_templates`

Stores LLM prompt text with placeholders for different processing steps
Enables prompt versioning, A/B testing, and updates without code changes
Links to workflow steps to determine which prompt to use when

`extraction_schemas`

Defines expected JSON output structure for LLM responses
Supports nested/hierarchical data extraction with validation rules
Allows complex structured data extraction from documents

`prompt_executions`

Logs every prompt execution with input context, output, and performance metrics
Tracks prompt effectiveness and enables optimization over time
Links to specific jobs and document chunks for complete traceability

`error_logs`

Master error table with centralized error tracking across all system components
Stores error details, severity levels, source schema/table/operation information
Includes resolution tracking with status, assignment, and resolution notes

`processing_errors`

Pipeline-specific errors linked to processing_jobs for full context
Captures retry attempts, failure patterns, and external API failures
Tracks worker identification and performance correlation data

`user_errors`

Frontend and user-facing issues with session and action context
Captures RPC function failures, browser compatibility issues
Links to user sessions and specific user actions for debugging