Using DataRobot Prompt Templates with CrewAI

CrewAI does not have a dedicated prompt template class. Instead, prompts are plain strings passed directly to Agent attributes like backstory, goal or Task attributes description or expected_output

Where	Field	Best for
Agent	backstory	Personality, role context, system prompt
Agent	goal	Short objective/directive
Task	description	Step-by-step task instructions
Task	expected_output	Output format constraints

Example with task description

# 1. Fetch and convert
import datarobot as dr

prompt_template = dr.genai.PromptTemplate.get("prompt_template_id")
prompt_template_version = prompt_template.get_latest_version()
prompt_text = prompt_template_version.to_fstring()

# 2. Build agents with the DR-managed prompt
llm = get_llm(model_name="provider/llm")

agent_writer = Agent(
    role="Content Writer",
    goal="Write a 2-3 sentence about: {topic}.",
    backstory=make_system_prompt(
        "You are a concise writer, write a short response."
    ),
    llm=llm,
)

agents = [agent_writer]

# 3. Define task
task_writer = Task(
    description=prompt_text
    expected_output="A concise 2-3 sentence response.",
    agent=agent_writer,
)

tasks = [task_writer]

# 4. Assemble and register
crew = Crew(agents=agents, tasks=tasks, stream=True)

kickoff_inputs = lambda user_prompt_content: {
    "topic": user_prompt_content,
    "chat_history": "",
}