AI Integration - LLM API Integration Guide

Overview

Generates production-ready LLM API integration code. From basic chat completion to RAG pipelines, with cost tracking and error handling.

---

1. When to Apply

Trigger	Behavior
Adding AI/LLM features	Full integration setup
"OpenAI", "Anthropic", "RAG"	Provider-specific integration
AI feature architecture	Pattern recommendation

---

2. Provider Quick Start

OpenAI (TypeScript)

import OpenAI from 'openai'

const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY })

async function chat(prompt: string): Promise<string> {
  const response = await openai.chat.completions.create({
    model: 'gpt-4o',
    messages: [{ role: 'user', content: prompt }],
    max_tokens: 1000,
  })
  return response.choices[0].message.content ?? ''
}

Anthropic (TypeScript)

import Anthropic from '@anthropic-ai/sdk'

const anthropic = new Anthropic({ apiKey: process.env.ANTHROPIC_API_KEY })

async function chat(prompt: string): Promise<string> {
  const response = await anthropic.messages.create({
    model: 'claude-sonnet-4-20250514',
    max_tokens: 1024,
    messages: [{ role: 'user', content: prompt }],
  })
  return response.content[0].type === 'text' ? response.content[0].text : ''
}

Streaming Response (Next.js)

// app/api/chat/route.ts
import { streamText } from 'ai'
import { openai } from '@ai-sdk/openai'

export async function POST(req: Request) {
  const { messages } = await req.json()
  const result = streamText({
    model: openai('gpt-4o'),
    messages,
  })
  return result.toDataStreamResponse()
}

---

3. Common Patterns

Structured Output

import { z } from 'zod'

const schema = z.object({
  sentiment: z.enum(['positive', 'negative', 'neutral']),
  confidence: z.number().min(0).max(1),
  summary: z.string(),
})

const response = await openai.chat.completions.create({
  model: 'gpt-4o',
  messages: [{ role: 'user', content: text }],
  response_format: { type: 'json_object' },
})
const result = schema.parse(JSON.parse(response.choices[0].message.content!))

RAG Pipeline (Simplified)

1. Document ingestion → chunk → embed → store in vector DB
2. Query → embed query → similarity search → retrieve chunks
3. Augment prompt with retrieved context → generate response

Error Handling & Retry

async function callWithRetry<T>(
  fn: () => Promise<T>,
  maxRetries = 3,
  baseDelay = 1000
): Promise<T> {
  for (let i = 0; i <= maxRetries; i++) {
    try {
      return await fn()
    } catch (error: any) {
      if (i === maxRetries) throw error
      if (error.status === 429 || error.status >= 500) {
        await new Promise(r => setTimeout(r, baseDelay * Math.pow(2, i)))
        continue
      }
      throw error // Don't retry client errors
    }
  }
  throw new Error('Unreachable')
}

---

4. Cost Estimation

Model	Input (1K tokens)	Output (1K tokens)
GPT-4o	$0.0025	$0.01
GPT-4o-mini	$0.00015	$0.0006
Claude Sonnet	$0.003	$0.015
Claude Haiku	$0.0008	$0.004

Rule of thumb: Start with cheaper models, upgrade only when quality demands it.

---

5. Tool Coordination

Tool	Purpose
Write	Generate integration code
Read	Analyze existing codebase
Bash	Install SDKs (npm install openai)

---

6. Boundaries

Will:

• Generate LLM API integration code for major providers
• Include streaming, structured output, and error handling patterns
• Provide cost estimation and model selection guidance
• Set up basic RAG pipeline structure

Will Not:

• Train or fine-tune models
• Set up vector databases (mention options, don't configure)
• Build full AI products (provides building blocks)
• Guarantee prompt quality (use prompt-engineer agent)