faion-llm-integration（faion LLM 集成）

Entry point: /faion-net — invoke this skill for automatic routing to the appropriate domain.

LLM Integration Skill

Communication: User's language. Code: English.

Purpose

Handles direct integration with LLM APIs. Covers OpenAI, Claude, Gemini, local models, prompt engineering, and output structuring.

Context Discovery

Auto-Investigation

Check these project signals before asking questions:

Discovery Questions

question: "What LLM task are you building?"
header: "Task Type"
multiSelect: false
options:
  - label: "Chat/completion API"
    description: "Direct LLM API calls for text generation"
  - label: "Function calling / tool use"
    description: "LLM selects and executes functions"
  - label: "Structured output (JSON mode)"
    description: "Constrain LLM to return valid JSON/schemas"
  - label: "Prompt optimization"
    description: "Improve existing prompts (few-shot, CoT, templates)"

question: "Which LLM provider(s)?"
header: "Provider"
multiSelect: true
options:
  - label: "OpenAI (GPT-4o, o1)"
    description: "OpenAI API integration"
  - label: "Claude (Anthropic)"
    description: "Claude Opus/Sonnet via Anthropic API"
  - label: "Gemini (Google)"
    description: "Gemini Pro/Flash via Google AI"
  - label: "Local LLM (Ollama)"
    description: "Self-hosted models for privacy"

question: "Do you need safety/content moderation?"
header: "Guardrails"
multiSelect: false
options:
  - label: "Yes - content filtering/PII detection"
    description: "Implement guardrails for safety"
  - label: "No - internal use only"
    description: "Skip guardrails"

Scope

Quick Start

Methodologies (26)

OpenAI (5):

• openai-api-integration: API setup, authentication, models
• openai-chat-completions: Chat API, streaming, parameters
• openai-function-calling: Tool definitions, execution
• openai-embeddings: Text embeddings (moved to rag-engineer)
• openai-assistants: Assistant API, threads, tools

Claude (6):

• claude-api-basics: Anthropic API setup
• claude-messages-api: Messages API, streaming
• claude-tool-use: Tool definitions, structured output
• claude-advanced-features: Extended thinking, prompt caching
• claude-best-practices: Safety, context management
• claude-api-integration: SDK integration patterns

Gemini (4):

• gemini-basics: Google AI setup, models
• gemini-multimodal: Vision, audio, video inputs
• gemini-function-calling: Function declarations
• gemini-api-integration: SDK patterns

Prompt Engineering (6):

• prompt-basics: Structure, few-shot, roles
• prompt-techniques: Advanced patterns, templates
• cot-basics: Chain-of-thought fundamentals
• cot-techniques: Zero-shot CoT, reasoning chains
• structured-output-basics: JSON mode, schemas
• structured-output-patterns: Advanced structuring

Safety & Tools (4):

• guardrails-basics: Content safety, PII detection
• guardrails-implementation: Implementation patterns
• function-calling-patterns: Tool design, error handling
• tool-use-basics: Tool fundamentals

Local (1):

• local-llm-ollama: Ollama setup, model management

Code Examples

OpenAI Chat Completion

from openai import OpenAI

client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Explain quantum computing"}
    ]
)
print(response.choices[0].message.content)

Claude Messages

import anthropic

client = anthropic.Anthropic()
message = client.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Explain RAG systems"}]
)
print(message.content[0].text)

Function Calling

tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Get current weather",
        "parameters": {
            "type": "object",
            "properties": {
                "location": {"type": "string"}
            }
        }
    }
}]

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "What's the weather in SF?"}],
    tools=tools
)

Related Skills

*LLM Integration v1.0 | 26 methodologies*