What problem does it solve?
The mdes-ollama skill solves the problem of unreliable AI model availability by automatically testing models on the MDES Ollama endpoint and selecting a working option with safe fallbacks.
Core Features & Use Cases
- Central Model Orchestration: Runs a full health-check and routing flow so other skills can reliably obtain a working model.
- Automatic Fallback Chain: Falls back from mdes.ollama models to codex, GPT Pro, and GitHub Copilot when Ollama models fail.
- Token Usage Tracking & Guardrails: Monitors input/output tokens, computes usage %, and blocks models when usage exceeds critical thresholds.
- Auto-Dev Loop Support: Enables automated testing and periodic health checks to keep the model pool ready for development workflows.
Use case example: When a multi-agent workflow starts (e.g., /nemotron or /gang), mdes-ollama ensures the best available model is selected based on success rate and token usage, then hands off execution to the chosen runtime.
Quick Start
Ask an agent to run: start model orchestration and pick the best working MDES Ollama model with fallback by issuing the command "/mdes-ollama".