model-selection

Automated dispatching of AI agent tasks uses centralized routing logic to classify jobs and assign them to a multi-provider model inventory. It enforces tier-based constraints and fallback policies to ensure consistent fleet operations.

Model tiering categorizes a model inventory into mentat, mentor, minion, and myrmidon levels to balance performance and cost. High-priority tasks route to advanced tiers, while routine jobs dispatch to cost-effective models.

Fallback policies for fleet model routing are configured by integrating provider-specific capabilities with centralized routing logic. This approach enforces tier-based constraints to handle job failures or model unavailability.

You can normalize existing job tiers by running a model-selection migration script. This script standardizes job classifications to ensure consistent routing across the entire fleet inventory.

Routing high-priority tasks to specific reasoning capabilities requires the automated dispatching system to match job requirements with the mentor tier. This ensures complex jobs bypass cost-effective minion models for appropriate processing.

Centralized routing logic limitations depend on the accurate definition of provider-specific capabilities and tier-based constraints. Ineffective tier categorization or missing fallback policies can cause improper job dispatch across the fleet.