Python SDK
pip install and chat in 5 lines
Quickstart: Python SDK
Go from zero to working chat in 5 lines of Python.
### Install the SDK ```bash pip install archipelag ``` ### Send a chat request ```python from archipelag import Client client = Client(api_key="your-api-key-here") result = client.chat("Say hello in three languages") print(result.content) ``` ### Run it ```bash export ARCHIPELAG_API_KEY="your-api-key-here" python hello.py ``` ``` Hello! (English) Bonjour! (French) Hola! (Spanish) ```
Using the response
client.chat() returns a ChatResult with these fields:
result = client.chat("Say hello in three languages")
result.content # "Hello! (English)\nBonjour!..."
result.job_id # "550e8400-e29b-41d4-a716-446655440000"
result.finish_reason # "stop"
result.usage.prompt_tokens # 12
result.usage.completion_tokens # 18
result.usage.total_tokens # 30
result.usage.credits_used # Decimal("0.001")
Streaming
Use client.chat_stream() to receive tokens as they’re generated:
from archipelag import Client
client = Client(api_key="your-api-key-here")
for event in client.chat_stream("Say hello in three languages"):
if event.type == "token":
print(event.content, end="", flush=True)
elif event.type == "done":
print() # newline
print(f"Tokens used: {event.usage.total_tokens}")
Options
Both methods accept the same options:
result = client.chat(
"Translate 'hello world' to Japanese",
system_prompt="You are a helpful translator.",
max_tokens=256,
temperature=0.3,
workload="llm-chat", # default
)
Environment variable
The client reads ARCHIPELAG_API_KEY automatically if no api_key is passed:
# Uses ARCHIPELAG_API_KEY from environment
client = Client()
Next steps
- Node.js SDK quickstart — same experience in JavaScript
- OpenAI-compatible quickstart — use your existing OpenAI code
- User Guide — models, pricing, error handling
