Using Anthropic Python SDK with AI-VERDE¶
Anthropic Python SDK is Anthropic's official Python SDK for interacting with Claude models. AI-VERDE can serve as the backend for the Anthropic Python SDK, routing requests through AI-VERDE's budget controls and guardrails to the Claude models available to your course or team.
Note
The Anthropic Python SDK uses AI-VERDE's Anthropic-style API at https://llm-api.cyverse.ai. It does not use the /bedrock endpoint, which is reserved for Claude Code, even when your Claude models are hosted on AWS Bedrock.
Prerequisites¶
- Obtain your AI-VERDE API key and note the AI-VERDE base URL. The base URL used in the script must have the
/v1suffix removed, if present. Instructions can be found here. - Note the Claude model ID you want to use. Instructions for listing available models can be found here. Model names in AI-VERDE are prefixed with an identifier for your course or team, separated by a dash or a slash (e.g.
my-team-claude-sonnet-5ormy-team/claude-sonnet-5). - Install the Anthropic Python SDK with
pip install anthropic. More information can be found at https://github.com/anthropics/anthropic-sdk-python.
Example Python script using Anthropic Python SDK with AI-VERDE¶
Export the following environment variable for your script.
export ANTHROPIC_API_KEY="<your-ai-verde-api-key>"
For example:
export ANTHROPIC_API_KEY="sk-xyz"
Example script:
import os
from anthropic import Anthropic
client = Anthropic(
base_url="https://llm-api.cyverse.ai", # Replace with the AI-VERDE base URL if different
api_key=os.environ.get("ANTHROPIC_API_KEY"),
)
message = client.messages.create(
max_tokens=128,
messages=[
{
"role": "user",
"content": "Hello, Claude",
}
],
model="<model name>", # Replace with a Claude model name from your model list
)
print(message.content[0].text)
Note
To obtain the AI-VERDE Anthropic Base URL, strip the /v1 suffix if present.