Agent Mesh
Go a2aa2a-protocolagent-meshagentic-aiinferencemcpmeshmesh-networknetwork-securitynetworking-aip2p-network
examples: surface clarifying questions as input-required tasks
Tomas Tormo57c93544a23a
diff --git a/development/examples/chat-a2a/agent.py b/development/examples/chat-a2a/agent.pyindex 4d15c5c7..51214ef6 100644--- a/development/examples/chat-a2a/agent.py+++ b/development/examples/chat-a2a/agent.py@@ -2,7 +2,6 @@import jsonimport osimport time-import uuidfrom google.protobuf.json_format import MessageToDict@@ -13,14 +12,13 @@ from a2a.server.events.event_queue import EventQueuefrom a2a.server.request_handlers import DefaultRequestHandlerfrom a2a.server.routes import create_agent_card_routes, create_jsonrpc_routesfrom a2a.server.tasks.inmemory_task_store import InMemoryTaskStore+from a2a.server.tasks.task_updater import TaskUpdaterfrom a2a.types import (AgentCapabilities,AgentCard,AgentInterface,AgentSkill,- Message,Part,- Role,)from google import genaifrom google.genai import types@@ -29,6 +27,10 @@ from starlette.applications import StarlettePORT = 7777MODEL = os.environ.get("GEMINI_MODEL", "models/gemini-3.5-flash-lite")+# Sentinel Gemini emits so a clarifying question becomes the A2A+# input-required task state instead of hiding inside a completed reply.+NEED_INPUT = "[NEED_INPUT]"+class ChatExecutor(AgentExecutor):"""One Gemini chat session per A2A contextId; the session carries the history."""@@ -43,7 +45,12 @@ class ChatExecutor(AgentExecutor):chat = self.gemini.aio.chats.create(model=MODEL,config=types.GenerateContentConfig(- thinking_config=types.ThinkingConfig(thinking_level="minimal")+ thinking_config=types.ThinkingConfig(thinking_level="minimal"),+ system_instruction=(+ "If you cannot complete the request without the user "+ f"answering a clarifying question first, start your reply "+ f"with the exact token {NEED_INPUT} followed by the question."+ ),),)self.chats[context.context_id] = chat@@ -73,15 +80,13 @@ class ChatExecutor(AgentExecutor):f"{time.monotonic() - started:.1f}s usage={reply.usage_metadata}",flush=True,)- await event_queue.enqueue_event(- Message(- role=Role.ROLE_AGENT,- message_id=str(uuid.uuid4()),- parts=[Part(text=reply.text or "")],- context_id=context.context_id,- task_id=context.task_id,- )- )+ updater = TaskUpdater(event_queue, context.task_id, context.context_id)+ text = reply.text or ""+ if text.startswith(NEED_INPUT):+ question = text.removeprefix(NEED_INPUT).strip()+ await updater.requires_input(updater.new_agent_message([Part(text=question)]))+ else:+ await updater.complete(updater.new_agent_message([Part(text=text)]))async def cancel(self, context: RequestContext, event_queue: EventQueue) -> None:pass