Can I use this for local model deployment?

Yes, it includes specific configurations for Ollama setup, model selection based on task, and GPU optimization for environments like Apple Silicon.

When should I use fine-tuning according to this skill?

The skill recommends fine-tuning only after exhausting prompt engineering and RAG, specifically when you have 1000+ quality examples for style or domain adaptation.

How does this skill handle function calling?

It provides patterns for defining tools using strict-mode schemas and managing parallel execution with robust input validation and error handling.

What streaming protocols are supported?

It focuses on Server-Sent Events (SSE) for web applications and includes patterns for handling backpressure and partial JSON parsing during tool calls.

LLM Integration Patterns

Name: LLM Integration Patterns
Author: yonatangross

byyonatangross

•

116

•

データサイエンスとML

Implements production-ready LLM patterns including tool use, streaming responses, local inference with Ollama, and parameter-efficient fine-tuning.

This skill provides a comprehensive toolkit for integrating Large Language Models into production environments using best-in-class implementation patterns. It covers critical developer workflows such as function calling with strict schemas, Server-Sent Events (SSE) for real-time streaming, and local deployment via Ollama to reduce cloud costs and enhance privacy. Additionally, the skill offers specialized guidance for LoRA/QLoRA fine-tuning, context window optimization, and rigorous LLM-as-judge evaluation metrics, enabling developers to build scalable, high-performance AI applications directly within the Claude Code environment.

主な機能

01Advanced prompt engineering techniques including ReAct, CoT, and DSPy

02Strict-mode function calling and parallel tool execution patterns

03Real-time streaming via SSE with backpressure and partial JSON handling

04Local inference optimization for Ollama and Apple Silicon hardware

05116 GitHub stars

06Fine-tuning workflows using LoRA/QLoRA and synthetic data generation

ユースケース

01Building autonomous agents that reliably call external APIs using structured tool definitions

02Reducing API costs by migrating development and production workloads to local models via Ollama

03Fine-tuning specialized models for niche domains when standard prompt engineering reaches its limits

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yonatangross/orchestkit llm-integration

For use in Claude.ai and ChatGPT

Download Skill