AI APIs & Tools

Modern AI systems are primarily accessed through APIs (Application Programming Interfaces). Understanding how to integrate AI into applications, manage tokens and costs, and use vector databases for enhanced retrieval is essential for building AI-powered products.

---

How AI APIs Work

AI API Request Flow:
+--------------+    HTTPS POST    +------------------+
| Your         |----------------->| AI Provider      |
| Application  |                  | (e.g. Anthropic) |
|              |<-----------------|                  |
|              |    JSON Response  |  +------------+ |
+--------------+                  |  | LLM Model  | |
                                  |  +------------+ |
                                  +------------------+

A typical call involves: authentication (API key in headers), a request body (model, messages, parameters), and a JSON response (generated text plus metadata).

Example API Call

import anthropic

client = anthropic.Anthropic(api_key="your-api-key-here")

message = client.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=1024,
    system="You are a helpful assistant that explains concepts clearly.",
    messages=[
        {"role": "user", "content": "What is photosynthesis?"}
    ]
)

print(message.content[0].text)

Common API Parameters

Parameter	Description	Typical Values
model	Which model to use	claude-sonnet-4-20250514, gpt-4o
messages	Conversation history	Array of role/content objects
system	System instructions	String defining behaviour
max_tokens	Max response tokens	256-4096
temperature	Randomness	0.0-1.0
top_p	Nucleus sampling	0.0-1.0
stream	Stream the response	true/false

---

Tokens and Tokenisation

LLMs process tokens, not raw text. One token is roughly three-quarters of a word (about 4 characters).

Tokenisation Example:
Text:     "Artificial intelligence is transforming the world."
Tokens:   ["Art", "ificial", " intelligence", " is",
           " transform", "ing", " the", " world", "."]

Token-Based Pricing

Model	Input (per 1M tokens)	Output (per 1M tokens)
Claude 3.5 Sonnet	$3.00	$15.00
GPT-4o	$2.50	$10.00
Claude Haiku	$0.25	$1.25
GPT-4o mini	$0.15	$0.60

Tip: Use the smallest model that meets your quality needs. Reserve frontier models for complex reasoning.

---

Embeddings and Vector Databases

Embeddings are dense vectors capturing semantic meaning. Similar texts have similar vectors.

Embedding Space (simplified to 2D):
                    ^
                    |   * "happy"
                    |     * "joyful"      * "delighted"
                    |
   "sad" *          |
     "unhappy" *    |                    * "programming"
   "miserable" *    |              * "coding"
                    |         * "software"
                    +----------------------------->
Words with similar meanings cluster together.