Autogen Agent Definitions

Operator builds an AssistantAgent and passes the task instructions only via the first user message in team.run(task=...), leaving system_message at its default.

Put persistent role/persona/policy instructions in AssistantAgent(system_message=...) so they appear as the system turn on every model_client call. Use the task argument only for the per-run user instruction. Without system_message, the agent loses role anchoring on multi-turn loops because the mod…

Two AssistantAgents share a single OpenAIChatCompletionClient instance and run inside a RoundRobinGroupChat. One agent calls await model_client.close() at end of its first turn.

Either construct a model_client per agent OR treat the shared client as team-scoped and close it ONCE at team teardown — not from within an agent. Closing mid-run kills the other agent's next call. Document the lifecycle (per-agent vs per-team) in the integration.

AssistantAgent is configured with reflect_on_tool_use=True. The model emits a tool call, the tool returns a 4KB JSON blob, and the operator expects the next message to be a natural-language summary.

With reflect_on_tool_use=True, after the tool returns AutoGen re-invokes the model with the tool result appended to the context so the model produces a natural-language reply. Verify the second model call actually happens (extra latency + tokens) and do not assume the raw tool output IS the assista…