feat(examples): expand CrewAI examples with multi-agent crew and Flow integration

[joaomdmoura]

Add 4 new CrewAI example directories demonstrating the full range of Agent Control capabilities:

• steering_financial_agent: All 3 action types (deny/steer/warn) in a wire-transfer scenario with 5 tested scenarios
• evaluator_showcase: All 4 built-in evaluators (SQL, LIST, REGEX, JSON) in a data-analyst scenario with 12 tested scenarios
• secure_research_crew: 3-agent sequential crew (Researcher, Analyst, Writer) with per-role policies and 5 tested scenarios
• content_publishing_flow: Full CrewAI Flow with @start/@listen/@router, conditional routing (low_risk/high_risk/escalation), embedded crews, client-side steering, and human-in-the-loop with 6 tested scenarios

Update the top-level CrewAI README with comprehensive feature coverage matrix, architecture diagrams, and scenario tables for all 5 examples.

[namrataghadi-galileo]

content_publishing_flow = "content_publishing_flow.main:run"

[namrataghadi-galileo]

def run():
"""Entry point for [project.scripts]."""
try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

evaluator_showcase = "evaluator_showcase.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

steering_financial_agent = "steering_financial_agent.main:run"

[namrataghadi-galileo]

run_crew = "steering_financial_agent.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

@joaomdmoura This looks great. Just some minor improvements to handle graceful shutdown. And Tasks were missing agent in the secure_research_crew example.

[abhinav-galileo]

Thank you for this substantial contribution and for helping expand Agent Control’s CrewAI coverage. The work proposed here has since been incorporated into main through #142, with additional compatibility fixes in #179.

Since these example families are now available on main, I’m closing this PR as superseded. We really appreciate the work and thought that went into demonstrating these CrewAI patterns.

[joaomdmoura]

Add 4 new CrewAI example directories demonstrating the full range of Agent Control capabilities:

• steering_financial_agent: All 3 action types (deny/steer/warn) in a wire-transfer scenario with 5 tested scenarios
• evaluator_showcase: All 4 built-in evaluators (SQL, LIST, REGEX, JSON) in a data-analyst scenario with 12 tested scenarios
• secure_research_crew: 3-agent sequential crew (Researcher, Analyst, Writer) with per-role policies and 5 tested scenarios
• content_publishing_flow: Full CrewAI Flow with @start/@listen/@router, conditional routing (low_risk/high_risk/escalation), embedded crews, client-side steering, and human-in-the-loop with 6 tested scenarios

Update the top-level CrewAI README with comprehensive feature coverage matrix, architecture diagrams, and scenario tables for all 5 examples.

[namrataghadi-galileo]

content_publishing_flow = "content_publishing_flow.main:run"

[namrataghadi-galileo]

def run():
"""Entry point for [project.scripts]."""
try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

evaluator_showcase = "evaluator_showcase.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

steering_financial_agent = "steering_financial_agent.main:run"

[namrataghadi-galileo]

run_crew = "steering_financial_agent.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

@joaomdmoura This looks great. Just some minor improvements to handle graceful shutdown. And Tasks were missing agent in the secure_research_crew example.

[abhinav-galileo]

Thank you for this substantial contribution and for helping expand Agent Control’s CrewAI coverage. The work proposed here has since been incorporated into main through #142, with additional compatibility fixes in #179.

Since these example families are now available on main, I’m closing this PR as superseded. We really appreciate the work and thought that went into demonstrating these CrewAI patterns.

[joaomdmoura]

Add 4 new CrewAI example directories demonstrating the full range of Agent Control capabilities:

• steering_financial_agent: All 3 action types (deny/steer/warn) in a wire-transfer scenario with 5 tested scenarios
• evaluator_showcase: All 4 built-in evaluators (SQL, LIST, REGEX, JSON) in a data-analyst scenario with 12 tested scenarios
• secure_research_crew: 3-agent sequential crew (Researcher, Analyst, Writer) with per-role policies and 5 tested scenarios
• content_publishing_flow: Full CrewAI Flow with @start/@listen/@router, conditional routing (low_risk/high_risk/escalation), embedded crews, client-side steering, and human-in-the-loop with 6 tested scenarios

Update the top-level CrewAI README with comprehensive feature coverage matrix, architecture diagrams, and scenario tables for all 5 examples.

[namrataghadi-galileo]

content_publishing_flow = "content_publishing_flow.main:run"

[namrataghadi-galileo]

def run():
"""Entry point for [project.scripts]."""
try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

evaluator_showcase = "evaluator_showcase.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

steering_financial_agent = "steering_financial_agent.main:run"

[namrataghadi-galileo]

run_crew = "steering_financial_agent.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

@joaomdmoura This looks great. Just some minor improvements to handle graceful shutdown. And Tasks were missing agent in the secure_research_crew example.

[abhinav-galileo]

Thank you for this substantial contribution and for helping expand Agent Control’s CrewAI coverage. The work proposed here has since been incorporated into main through #142, with additional compatibility fixes in #179.

Since these example families are now available on main, I’m closing this PR as superseded. We really appreciate the work and thought that went into demonstrating these CrewAI patterns.

[joaomdmoura]

Add 4 new CrewAI example directories demonstrating the full range of Agent Control capabilities:

• steering_financial_agent: All 3 action types (deny/steer/warn) in a wire-transfer scenario with 5 tested scenarios
• evaluator_showcase: All 4 built-in evaluators (SQL, LIST, REGEX, JSON) in a data-analyst scenario with 12 tested scenarios
• secure_research_crew: 3-agent sequential crew (Researcher, Analyst, Writer) with per-role policies and 5 tested scenarios
• content_publishing_flow: Full CrewAI Flow with @start/@listen/@router, conditional routing (low_risk/high_risk/escalation), embedded crews, client-side steering, and human-in-the-loop with 6 tested scenarios

Update the top-level CrewAI README with comprehensive feature coverage matrix, architecture diagrams, and scenario tables for all 5 examples.

[namrataghadi-galileo]

content_publishing_flow = "content_publishing_flow.main:run"

[namrataghadi-galileo]

def run():
"""Entry point for [project.scripts]."""
try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

evaluator_showcase = "evaluator_showcase.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

steering_financial_agent = "steering_financial_agent.main:run"

[namrataghadi-galileo]

run_crew = "steering_financial_agent.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

@joaomdmoura This looks great. Just some minor improvements to handle graceful shutdown. And Tasks were missing agent in the secure_research_crew example.

[abhinav-galileo]

Thank you for this substantial contribution and for helping expand Agent Control’s CrewAI coverage. The work proposed here has since been incorporated into main through #142, with additional compatibility fixes in #179.

Since these example families are now available on main, I’m closing this PR as superseded. We really appreciate the work and thought that went into demonstrating these CrewAI patterns.

[joaomdmoura]

Add 4 new CrewAI example directories demonstrating the full range of Agent Control capabilities:

• steering_financial_agent: All 3 action types (deny/steer/warn) in a wire-transfer scenario with 5 tested scenarios
• evaluator_showcase: All 4 built-in evaluators (SQL, LIST, REGEX, JSON) in a data-analyst scenario with 12 tested scenarios
• secure_research_crew: 3-agent sequential crew (Researcher, Analyst, Writer) with per-role policies and 5 tested scenarios
• content_publishing_flow: Full CrewAI Flow with @start/@listen/@router, conditional routing (low_risk/high_risk/escalation), embedded crews, client-side steering, and human-in-the-loop with 6 tested scenarios

Update the top-level CrewAI README with comprehensive feature coverage matrix, architecture diagrams, and scenario tables for all 5 examples.

[namrataghadi-galileo]

content_publishing_flow = "content_publishing_flow.main:run"

[namrataghadi-galileo]

def run():
"""Entry point for [project.scripts]."""
try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

evaluator_showcase = "evaluator_showcase.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

steering_financial_agent = "steering_financial_agent.main:run"

[namrataghadi-galileo]

run_crew = "steering_financial_agent.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

@joaomdmoura This looks great. Just some minor improvements to handle graceful shutdown. And Tasks were missing agent in the secure_research_crew example.

[abhinav-galileo]

Thank you for this substantial contribution and for helping expand Agent Control’s CrewAI coverage. The work proposed here has since been incorporated into main through #142, with additional compatibility fixes in #179.

Since these example families are now available on main, I’m closing this PR as superseded. We really appreciate the work and thought that went into demonstrating these CrewAI patterns.

[joaomdmoura]

Add 4 new CrewAI example directories demonstrating the full range of Agent Control capabilities:

• steering_financial_agent: All 3 action types (deny/steer/warn) in a wire-transfer scenario with 5 tested scenarios
• evaluator_showcase: All 4 built-in evaluators (SQL, LIST, REGEX, JSON) in a data-analyst scenario with 12 tested scenarios
• secure_research_crew: 3-agent sequential crew (Researcher, Analyst, Writer) with per-role policies and 5 tested scenarios
• content_publishing_flow: Full CrewAI Flow with @start/@listen/@router, conditional routing (low_risk/high_risk/escalation), embedded crews, client-side steering, and human-in-the-loop with 6 tested scenarios

Update the top-level CrewAI README with comprehensive feature coverage matrix, architecture diagrams, and scenario tables for all 5 examples.

[namrataghadi-galileo]

content_publishing_flow = "content_publishing_flow.main:run"

[namrataghadi-galileo]

def run():
"""Entry point for [project.scripts]."""
try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

evaluator_showcase = "evaluator_showcase.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

steering_financial_agent = "steering_financial_agent.main:run"

[namrataghadi-galileo]

run_crew = "steering_financial_agent.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

@joaomdmoura This looks great. Just some minor improvements to handle graceful shutdown. And Tasks were missing agent in the secure_research_crew example.

[abhinav-galileo]

Thank you for this substantial contribution and for helping expand Agent Control’s CrewAI coverage. The work proposed here has since been incorporated into main through #142, with additional compatibility fixes in #179.

Since these example families are now available on main, I’m closing this PR as superseded. We really appreciate the work and thought that went into demonstrating these CrewAI patterns.

[joaomdmoura]

Add 4 new CrewAI example directories demonstrating the full range of Agent Control capabilities:

• steering_financial_agent: All 3 action types (deny/steer/warn) in a wire-transfer scenario with 5 tested scenarios
• evaluator_showcase: All 4 built-in evaluators (SQL, LIST, REGEX, JSON) in a data-analyst scenario with 12 tested scenarios
• secure_research_crew: 3-agent sequential crew (Researcher, Analyst, Writer) with per-role policies and 5 tested scenarios
• content_publishing_flow: Full CrewAI Flow with @start/@listen/@router, conditional routing (low_risk/high_risk/escalation), embedded crews, client-side steering, and human-in-the-loop with 6 tested scenarios

Update the top-level CrewAI README with comprehensive feature coverage matrix, architecture diagrams, and scenario tables for all 5 examples.

[namrataghadi-galileo]

content_publishing_flow = "content_publishing_flow.main:run"

[namrataghadi-galileo]

def run():
"""Entry point for [project.scripts]."""
try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

evaluator_showcase = "evaluator_showcase.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

steering_financial_agent = "steering_financial_agent.main:run"

[namrataghadi-galileo]

run_crew = "steering_financial_agent.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

@joaomdmoura This looks great. Just some minor improvements to handle graceful shutdown. And Tasks were missing agent in the secure_research_crew example.

[abhinav-galileo]

Thank you for this substantial contribution and for helping expand Agent Control’s CrewAI coverage. The work proposed here has since been incorporated into main through #142, with additional compatibility fixes in #179.

Since these example families are now available on main, I’m closing this PR as superseded. We really appreciate the work and thought that went into demonstrating these CrewAI patterns.

[joaomdmoura]

Add 4 new CrewAI example directories demonstrating the full range of Agent Control capabilities:

• steering_financial_agent: All 3 action types (deny/steer/warn) in a wire-transfer scenario with 5 tested scenarios
• evaluator_showcase: All 4 built-in evaluators (SQL, LIST, REGEX, JSON) in a data-analyst scenario with 12 tested scenarios
• secure_research_crew: 3-agent sequential crew (Researcher, Analyst, Writer) with per-role policies and 5 tested scenarios
• content_publishing_flow: Full CrewAI Flow with @start/@listen/@router, conditional routing (low_risk/high_risk/escalation), embedded crews, client-side steering, and human-in-the-loop with 6 tested scenarios

Update the top-level CrewAI README with comprehensive feature coverage matrix, architecture diagrams, and scenario tables for all 5 examples.

[namrataghadi-galileo]

content_publishing_flow = "content_publishing_flow.main:run"

[namrataghadi-galileo]

def run():
"""Entry point for [project.scripts]."""
try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

evaluator_showcase = "evaluator_showcase.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

steering_financial_agent = "steering_financial_agent.main:run"

[namrataghadi-galileo]

run_crew = "steering_financial_agent.main:run"

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

try:
main()
finally:
agent_control.shutdown()

[namrataghadi-galileo]

@joaomdmoura This looks great. Just some minor improvements to handle graceful shutdown. And Tasks were missing agent in the secure_research_crew example.

[abhinav-galileo]

Thank you for this substantial contribution and for helping expand Agent Control’s CrewAI coverage. The work proposed here has since been incorporated into main through #142, with additional compatibility fixes in #179.

Since these example families are now available on main, I’m closing this PR as superseded. We really appreciate the work and thought that went into demonstrating these CrewAI patterns.