|
58 | 58 | - "tracing": Telemetry, observability, structured logs, or spans. |
59 | 59 | - "core": Core ADK runtime (Agent definitions, Runner, planners, |
60 | 60 | thinking config, CLI commands, GlobalInstructionPlugin, CPU usage, or |
61 | | - general orchestration). Default to "core" when the topic is about ADK |
62 | | - behavior and no other label is a better fit. |
| 61 | + general orchestration including agent transfer for multi-agents system). |
| 62 | + Default to "core" when the topic is about ADK behavior and no other |
| 63 | + label is a better fit. |
63 | 64 | - "agent engine": Vertex AI Agent Engine deployment or sandbox topics |
64 | 65 | only (e.g., `.agent_engine_config.json`, `ae_ignore`, Agent Engine |
65 | 66 | sandbox, `agent_engine_id`). If the issue does not explicitly mention |
66 | 67 | Agent Engine concepts, do not use this label—choose "core" instead. |
67 | | - - "a2a": Agent-to-agent workflows, coordination logic, or A2A protocol. |
| 68 | + - "a2a": A2A protocol, running agent as a2a agent with "--a2a" option for |
| 69 | + remote agent to talk with. Talking to remote agent via RemoteA2aAgent. |
| 70 | + NOT including those local multi-agent systems. |
68 | 71 | - "bq": BigQuery integration or general issues related to BigQuery. |
69 | 72 | - "workflow": Workflow agents and workflow execution. |
70 | 73 | - "auth": Authentication or authorization issues. |
|
0 commit comments