Not able to run crewai on databricks when crew is triggered

General
1

Getting following error
ERROR: LiteLLM call failed: litellm.BadRequestError: databricksException - {“error_code”:“BAD_REQUEST”,“message”:“Bad request: Chat message input must end with a ‘user’ or ‘tool’ role\n”}

on the following code when crew is triggered

import os
import json
from datetime import datetime

from crewai import Agent, Task, Crew, Process
from crewai_tools import SerperDevTool, YoutubeVideoSearchTool
from crewai import Agent, LLM

To use DBRX model, set the following environment variables

os.environ[‘OPENAI_API_KEY’] = ‘NA’
print(os.environ[‘SERPER_API_KEY’])
print(os.environ[‘DATABRICKS_HOST’])
print(os.environ[‘DATABRICKS_TOKEN’])

import agentops
agentops.init()

Override default LLM with Databricks DBRX model

llm = LLM(
model=“databricks/databricks-meta-llama-3-1-405b-instruct”,

)

search_tool = SerperDevTool()
youtube_tool = YoutubeVideoSearchTool()

today = datetime.today().strftime(‘%Y-%m-%d %H-%M’)

researcher = Agent(
role=‘Senior Research Analyst’,
goal=‘Uncover cutting-edge developments in AI, data, software engineering’,
backstory=“”“You work at a leading tech think tank.
Your expertise lies in identifying emerging credible tech trends from blogs, articles, and videos.
You have a knack for dissecting complex data and presenting actionable insights.”“”,
verbose=True,
allow_delegation=False,
tools=[search_tool, youtube_tool],
llm=llm,
)

writer = Agent(
role=‘Tech Content Strategist’,
goal=‘Create engaging content on tech advancements in AI, data, and software engineering’,
backstory=“”“You are a renowned Content Strategist, known for your insightful
and engaging tech articles. You have a keen eye for captivating narratives and real-world applications.”“”,
verbose=True,
allow_delegation=True,
llm=llm,
)

Create tasks for your agents

task1 = Task(
description=“”“Conduct a comprehensive analysis of the latest advancements in tech,
focusing on fields like AI, data engineering, software engineering, AWS and Azure cloud, and Databricks
in the past month from 2024. Identify key trends, breakthrough technologies, and potential impact on industries.”“”,
expected_output=“Full analysis report in bullet points”,
agent=researcher,
)

task2 = Task(
description=“”“Using the insights provided, provide 10 or more suggestions for engaging blog posts, videos, and projects,
that highlight the most significant advancements in fields like AI, data engineering, software engineering, AWS and Azure cloud, and Databricks.
Your suggestions should focus on real-world use cases and cater to a tech-savvy audience.
The projects should be specific ideas and small enough to be completed within 1-4 weeks.
Make it sound cool, avoid complex words so it doesn’t sound like AI generated.”“”,
expected_output=“Numbered list of engaging blog post ideas, project suggestions, and video ideas”,
agent=writer,
)

def main():
# Assemble your crew with a sequential process
crew = Crew(
agents=[researcher, writer],
tasks=[task1, task2],
verbose=True,
output_log_file=f"logs/crew_{today}.log",
)

# Start the crew to work
result = crew.kickoff()

print('#' * 30)
print(result)

# Append crew.usage_metrics to the log file
with open(f"logs/crew_{today}.log", "a") as text_file:
    text_file.write('\n\n')
    text_file.write(json.dumps(crew.usage_metrics, indent=2))

# Save the result to a file with today's date
with open(f"outputs/result_{today}.md", "w") as text_file:
    text_file.write(result)

if name == ‘main’:
main()