Configure inference providers and map local models to hex agent tiers.
npx skills add https://github.com/gaberger/hex --skill hex-inference
Please help me install this Agent Skill. Skill: hex-inference Source: https://github.com/gaberger/hex/tree/main/.claude/skills/hex-inference Command: npx skills add https://github.com/gaberger/hex --skill hex-inference
Manage inference providers (Ollama, vLLM, OpenAI-compatible) and map local models to hex agent tiers to route tasks.
Register an Ollama provider, map a local model to a hex tier, and verify the configuration with a provider list.
High-intent search queries and answers about installing and using this skill.
To configure an Ollama provider for local inference, you discover and register the provider, map your local models to specific hex agent tiers, and verify the setup by testing connectivity and listing the provider configuration.
Yes, you can map local models to hex agent tiers to route tasks efficiently. This aligns models with the correct agent tiers, allowing appropriate task distribution across your configured inference providers.
Yes, it supports OpenAI-compatible and vLLM inference providers. You can discover, register, and test connectivity across these providers to ensure proper model mapping and tier-alignment for your agents.
The best way to test inference provider connectivity and model mappings is to validate the configuration across your registered providers. This verifies that local models are correctly aligned with hex agent tiers for routing.
Local models fail to route to the correct agent tiers if inference provider mappings are misconfigured. Adjust the model mappings and verify provider connectivity to ensure proper tier-alignment.
winsorllcConnect to local LLM endpoints with automatic cloud fallback.
mistakeknotBenchmark token efficiency of agent workflows by analyzing prompt and model interactions.
dejavoo21Coordinate multi-agent swarms for parallel tasks with dynamic topologies.
understudylabsCompare local MLX models against frontier gateways on latency, cost, and quality.
