Skip to content
Osarai
Get started

DocsSDKs and frameworks

Anthropic SDK

Use the Anthropic SDKs for Python and TypeScript with your Osarai instance through its Messages API.

The Anthropic SDKs call Osarai's Messages API. Give them the site address https://osarai.xyz (they add /v1/messages) and your qck_… key. The model is the open model on your instance.

Install

Terminal
pip install anthropic

Configure

Both SDKs read these environment variables:

Terminal
export ANTHROPIC_BASE_URL="https://osarai.xyz"
export ANTHROPIC_API_KEY="$OSARAI_API_KEY"

Or pass them explicitly:

Python
import anthropic, os

client = anthropic.Anthropic(base_url="https://osarai.xyz", api_key=os.environ["OSARAI_API_KEY"])

Messages

Python
message = client.messages.create(
    model="qwen3.8-27b",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Explain recursion in one paragraph."}],
)
for block in message.content:
    if block.type == "text":
        print(block.text)

Reasoning models answer with a thinking block before the text block, so read blocks by type rather than taking content[0].

Streaming

Python
with client.messages.stream(
    model="qwen3.8-27b",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Write a limerick about GPUs."}],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

Thinking

Python
message = client.messages.create(
    model="qwen3.8-27b",
    max_tokens=4096,
    thinking={"type": "enabled", "budget_tokens": 2000},   # low effort
    messages=[{"role": "user", "content": "Is 221 prime?"}],
)

Use {"type": "disabled"} for the fastest answers. See Thinking for how budgets map to effort.

What differs from Anthropic's API

  • Anthropic's server tools (web search, code execution), prompt caching and the Files API are not available; requests that include them still work, without those parts.
  • count_tokens counts with your model's tokenizer, not Claude's.
  • Model names are your instance's ids, such as qwen3.8-27b.