Making a request

Setup Steps

• Step 1: Install packages in Jupyter notebook using %pip install anthropic python-dotenv
• Ignore red syntax errors from the percent sign in VS Code
• Step 2: Store API key securely using . env file
• Create a . env file in the same directory as your notebook
• Format: ANTHROPIC_API_KEY="your-key-here"
• Add . env to version control ignore file (e. g. , . gitignore) to prevent accidental commits
• Step 3: Create API client and declare model variable
• Use the anthropic package to instantiate a client
• Store model name (e. g. , "claude-sonnet") in a variable
• Step 4: Make your first request using the client

The create() Function

The core function for accessing Claude is client. messages. create() with three required keyword arguments:

• model: Name of the model to run (string variable defined earlier)
• max_tokens: Maximum number of tokens Claude can generate
• Acts as a safety mechanism to prevent excessive output
• Claude doesn't target this number; it generates appropriate-length responses
• Generation stops automatically if limit is reached
• messages: List of message objects representing the conversation

Message Types

• User messages: Contain text authored by a person (developer or end user)
• Format: dictionary with role: "user" and content: "your text"
• Assistant messages: Contain text generated by Claude
• Returned in API responses

Accessing the Response

• The response object contains nested data; use message. content[0]. text to extract just the generated text
• Ignore other properties in the response object unless specifically needed

Takeaways

• Install the anthropic and python-dotenv packages, store your API key in a . env file, and create a client to make requests
• Use client. messages. create() with three required arguments: model, max_tokens (safety limit for output length), and messages (list of message objects)
• Format messages as dictionaries with role ("user" or "assistant") and content fields; extract the generated text from the response using message. content[0]. text