skills.rest
yo-steven
yo-stevenCOMMUNITY
Explore all skills by this author →
GitHub

nemoclaw-user-configure-inference

Configure NemoClaw to route agent inference to local model servers.

Updated May 23, 2026
Search Tags:#ollama#vllm#local inference#nemoclaw#inference routing#openai-compatible endpoint#tool calling reliability
One-click install
npx skills add https://github.com/yo-steven/skills-exploration-20260522 --skill nemoclaw-user-configure-inference-yo-steven
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: nemoclaw-user-configure-inference
Source: https://github.com/yo-steven/skills-exploration-20260522/tree/main/skills/NemoClaw/nemoclaw-user-configure-inference
Command: npx skills add https://github.com/yo-steven/skills-exploration-20260522 --skill nemoclaw-user-configure-inference-yo-steven

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps you configure NemoClaw to use a local inference server instead of a cloud API, so your agents can run with the model runtime you choose.

Core Features & Use Cases

  • Local inference routing with safe defaults: Connects NemoClaw to Ollama, or to OpenAI-compatible local endpoints (vLLM, TensorRT-LLM, llama.cpp, LocalAI, and similar) using NemoClaw’s inference.local routing.
  • Provider onboarding and endpoint validation: Guides you through the NemoClaw onboard flow and ensures the selected backend supports the required API path and tool-calling behavior (including streaming expectations).
  • Tool-calling reliability for local models: Addresses common Ollama tool-call leakage issues and directs you to vLLM-compatible setups with the correct tool-call parser when structured tool calls are required.
  • Runtime switching support: Works alongside the ability to switch inference models/providers at runtime without re-onboarding by re-probing and re-baking the correct routing when needed.

Use cases include setting up Ollama on macOS/Linux/WSL, pointing NemoClaw at a self-hosted OpenAI-compatible server, configuring local vLLM/NIM options, and troubleshooting tool-calling failures where the TUI shows raw JSON instead of dispatching tools.

Quick Start

Run the setup wizard and select Local Ollama: nemoclaw onboard

Frequently Asked Questions about nemoclaw-user-configure-inference

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I route my local inference server to handle agent tool calling instead of a cloud API?▼

To route local inference, you can configure your agent framework to use a local model server like Ollama or an OpenAI-compatible endpoint, ensuring the selected backend supports the required API path and streaming behavior for reliable tool calls.

Why does my local Ollama model return raw JSON instead of dispatching structured tool calls?▼

Ollama tool-call leakage often causes raw JSON output instead of structured tool dispatch; switching to a vLLM-compatible setup with the correct tool-call parser resolves this reliability issue for structured tool calling.

Can I use an OpenAI-compatible local endpoint like vLLM or llama.cpp for offline agent inference?▼

Yes, you can connect to OpenAI-compatible local endpoints including vLLM, llama.cpp, and TensorRT-LLM for offline deployments by setting the inference route to local and validating connectivity during onboarding.

What's the best way to automate local inference provisioning without interactive setup prompts?▼

You can automate local inference provisioning non-interactively by setting NEMOCLAW_* environment variables, which configure the inference routing and backend selection without requiring the interactive onboard wizard.

How do I switch local inference providers at runtime without repeating the onboarding process?▼

You can switch local inference providers at runtime by re-probing the endpoint and re-baking the routing configuration, which updates the inference path dynamically without requiring a full re-onboarding flow.

Related Skills

Explore More Software Engineering Skills→
gabergergaberger
4

hex-inference

Configure inference providers and map local models to hex agent tiers.

Community
Advanced
SkastVnTSkastVnT
7

core-chatbot-routing-audit

Audit chatbot request paths for Flask and FastAPI route parity.

Community
Advanced
kwcantrellkwcantrell

llama-server

Serve local GGUF LLMs with an OpenAI-compatible HTTP API.

Community
Advanced
venomez-vipervenomez-viper
1

foundation-models

Integrate Apple's Foundation Models for on-device AI text generation and structured output.

Community
Advanced