Chapter 1: Using LLM APIs

How Applications Connect to LLMs

When you use ChatGPT through the website, you're using a user interface. But how do apps, websites, and services integrate LLMs into their products? Through APIs.

---

What Is an API?

Simple Definition

An API (Application Programming Interface) is a way for different software applications to communicate with each other.

The Restaurant Analogy

Think of an API like ordering at a restaurant:

Restaurant	API
You (customer)	Your application
Menu	API documentation
Waiter	The API itself
Kitchen	The LLM service
Your order	Your API request
Your food	The LLM's response

You don't go into the kitchen—you communicate through the waiter (API), who takes your order (request) to the kitchen (LLM) and brings back your food (response).

---

ChatGPT vs. API

These are different ways to access the same technology:

Aspect	ChatGPT Website/App	OpenAI API
User	Everyday users	Developers
Interface	Chat window	Code
Pricing	Monthly subscription	Pay per use (tokens)
Customization	Limited	Extensive
Use Case	Personal productivity	Building products

---

How API Calls Work

A Simplified Flow

What Goes in a Request

1. Your API key: Identifies and authenticates you
2. The model: Which LLM to use (e.g., gpt-4, claude-3)
3. The prompt: What you want the LLM to do
4. Parameters: Settings like temperature, max tokens

What Comes Back

1. The generated text: The LLM's response
2. Usage info: How many tokens were used
3. Metadata: Model version, timing info

---

Simple API Example (Conceptual)

You don't need to code, but here's what an API call looks like:

// Conceptual example - not code to run
Request:
{
  "model": "gpt-4",
  "messages": [
    {"role": "system", "content": "You are a helpful assistant"},
    {"role": "user", "content": "What is the capital of France?"}
  ],
  "temperature": 0.7
}

Response:
{
  "choices": [
    {"message": {"content": "The capital of France is Paris."}}
  ],
  "usage": {
    "prompt_tokens": 25,
    "completion_tokens": 8,
    "total_tokens": 33
  }
}

---

Major API Providers

OpenAI API

• Models: GPT-4, GPT-3.5-turbo
• Strengths: Most popular, well-documented
• Access: api.openai.com

Anthropic API

• Models: Claude 3 (Opus, Sonnet, Haiku)
• Strengths: Long context, safety-focused
• Access: anthropic.com

Google AI

• Models: Gemini Pro, Gemini Ultra
• Strengths: Multimodal, Google integration
• Access: ai.google.dev

Azure OpenAI

• Models: Same as OpenAI, via Microsoft
• Strengths: Enterprise features, compliance
• Access: azure.microsoft.com

---

No-Code Options

You don't need to be a developer to use LLM APIs!

No-Code Tools

Tool	What It Does
Zapier	Connect LLMs to other apps without code
Make (Integromat)	Visual automation with AI
LangFlow	Visual programming for LLM apps
ChatGPT GPT Builder	Create custom GPTs without coding

Low-Code Platforms

Platform	Use Case
Microsoft Power Platform	Enterprise automation
Bubble	Web apps with AI
Retool	Internal tools

---

Key API Concepts

Authentication

Your API key is like a password—keep it secret!

Rate Limits

Limits on requests per minute to prevent overload:

• Free tiers: ~3 requests/minute
• Paid tiers: ~60+ requests/minute

Endpoints

Different URLs for different functions:

• /chat/completions - Chat responses
• /embeddings - Text to numbers
• /audio/transcriptions - Speech to text

Streaming

Receiving responses word-by-word as they're generated (like ChatGPT typing).

---

Common Use Cases

1. Customer Support Bots

• Integrate into website or app
• Handle common questions
• Escalate complex issues

2. Content Generation

• Email drafts
• Marketing copy
• Product descriptions

3. Data Processing

• Summarize documents
• Extract information
• Categorize content

4. Code Assistance

• Generate code snippets
• Explain code
• Fix bugs

---

Choosing Between Consumer and API

Choose Consumer (ChatGPT/Claude) When...	Choose API When...
Personal use	Building a product
Occasional use	Need automation
Exploring AI	High volume
Simple tasks	Custom integration

---

Key Takeaways

✅ APIs let applications communicate with LLM services

✅ ChatGPT is for consumers; APIs are for building products

✅ API calls include your key, model choice, prompt, and settings

✅ No-code tools make APIs accessible to non-developers

✅ Rate limits and authentication are important considerations

---

Think About It

1. What kind of application might you build if you had API access to an LLM?
2. Why would a company use the API instead of just telling employees to use ChatGPT?
3. What security concerns might arise from using API keys?

---

Next Steps

Now let's understand the costs involved in Chapter 2: Tokens and Pricing.