Overview
Use routerbase to plan model selection under a unified OpenAI-compatible integration surface. This skill turns workload requirements into a model candidate list, fallback chain, cost and latency strategy, and a pre-launch validation checklist.
When to use
Use this skill when the user requests the following:
- Select a RouterBase model for chat, code, reasoning, vision, or multimodal workloads
- Design a fallback chain for provider failures or quality degradation
- Make trade-offs among cost, latency, context length, output quality, and availability
- Write model routing rules for production applications or Agent workflows
- Create a pre-launch validation plan for model behavior
Do not use
This skill should not be used in the following scenarios:
- Assert exact prices, availability, or model performance without checking current source data
- Recommend only a single model without explaining assumptions and fallback behavior
- Hide routing changes without the knowledge of the application owner or production operations staff
Usage instructions
- Clarify the workload type, quality threshold, latency budget, context size, expected traffic, and failure tolerance.
- Organize candidate models by role: primary model, economical fallback, high-quality fallback, and specialized fallback.
- Clearly label routing assumptions, and separately flag data that requires checking the current catalog or prices.
- Recommend a small-step launch process: test prompts, golden cases, logging, retry strategy, rate-limit handling, and rollback conditions.
- Keep examples as OpenAI-compatible as possible so application integration remains portable.