I’m using the Agno framework, where a Team Coordinator routes each user request to the appropriate agent based on intent. Each agent follows its own instructions defined in separate README files. However, the average response time is around 4–5 seconds, which feels quite high.
Is there any way to reduce this latency?
Currently, I’m using the gpt-4.1 model for both the Team Coordinator and all agents.
Configs we are using currently:
Team(
model="gpt-4.1",
members=[payment_agent, account_agent],
instructions=TEAM_COORDINATOR_INSTRUCTIONS, # Router prompt
tools=router_tools,
session_id=session_id,
num_history_runs=50,
add_session_state_to_context=True, # Share context with agents
enable_agentic_state=True,
add_history_to_context=True,
db=self.storage,
determine_input_for_members=False, # Router provides minimal input; agents fetch their own context
respond_directly=True, # Agents respond directly to user (no Team mediation)
store_member_responses=True, # Enable to get member_responses with timing info
# Performance optimizations
share_member_interactions=True,
# Memory and session management
enable_session_summaries=False,
enable_user_memories=False,
# Tool optimization
tool_call_limit=10,
debug_mode=True,
markdown=True,
)
My queries:
- Is there a way to reduce the latency involved in generating responses?
- Also, how can I measure the time taken specifically by the Team Coordinator to route the user’s query to the appropriate agent — i.e., the coordinator’s routing time?