dojo-sdk-client 0.1.9

Client sdk to intearct with Dojo

Dojo SDK Client

Python client SDK for interacting with the Dojo API

Installation

uv add dojo-sdk-client

if you want to use our example agents:

uv add "dojo-sdk-client[agents]"

if you want to use our multi-turn environment with verifiers

uv add "dojo-sdk-client[multi_turn_env]"

if you want to use our browserbase engine:

uv add "dojo-sdk-client[browserbase]"

Quick Start

Basic Task Execution

from dojo_sdk_client import BaseDojoClient
from dojo_sdk_client.types import TaskStatus

client = BaseDojoClient(api_key="your-api-key")

# Create and start task
exec_id = await client.create_task(
    task_id="tic-tac-toe",
    state={"board": [[0]*3 for _ in range(3)]}
)
await client.start_task(exec_id)

status = await client.get_task_status(exec_id)
while status.status == TaskStatus.QUEUED:
    await asyncio.sleep(1)
    status = await client.get_task_status(exec_id)

# Submit action
action = {"type": "click", "x": 100, "y": 100}
await client.submit_action(exec_id, action, "Making move")

Agent Evaluation

from dojo_sdk_client import DojoEvalClient

eval_client = DojoEvalClient(agent=your_agent, api_key="your-api-key")

results = await eval_client.evaluate(
    tasks=["tic-tac-toe", "2048"],
    num_runners=2
)

Changing Engines

from dojo_sdk_client.engines.browserbase_engine import BrowserBaseEngine

engine = BrowserBaseEngine(
    api_key="your-browserbase-api-key",
    project_id="your-browserbase-project-id",
    dojo_api_key="your-dojo-api-key"
)

eval_client = DojoEvalClient(agent=your_agent, api_key="your-api-key", engine=engine)

BrowserBase engine has a concurrency limit that can be controlled with the BROWSERBASE_CONCURRENT_LIMIT environment variable.

Clients

BaseDojoClient - Low-level HTTP client for direct API interactions

DojoEvalClient - High-level client for running automated evaluations

MultiTurnEnv - Multi-turn environment for running evaluations with Verifiers

Engines

DojoEngine - Default engine for running tasks using Dojo

BrowserBaseEngine - Engine for running tasks using BrowserBase

Documentation

Visit docs.trydojo.ai for complete documentation.