Agent Mesh
Go a2aa2a-protocolagent-meshagentic-aiinferencemcpmeshmesh-networknetwork-securitynetworking-aip2p-network
examples: add chat-a2a, a Gemini-backed A2A agent with a chat REPL for the kind mesh
Tomas Tormo20e5799acba0
diff --git a/development/examples/chat-a2a/Dockerfile b/development/examples/chat-a2a/Dockerfilenew file mode 100644index 00000000..2e474c2b--- /dev/null+++ b/development/examples/chat-a2a/Dockerfile@@ -0,0 +1,10 @@+FROM python:3.12-slim++WORKDIR /srv+COPY requirements.txt /srv/+RUN pip install --no-cache-dir -r requirements.txt+COPY agent.py /srv/++ENV GEMINI_API_KEY=<API_KEY>++CMD ["python3", "/srv/agent.py"]diff --git a/development/examples/chat-a2a/README.md b/development/examples/chat-a2a/README.mdnew file mode 100644index 00000000..f98cdc77--- /dev/null+++ b/development/examples/chat-a2a/README.md@@ -0,0 +1,70 @@+# chat-a2a++A Gemini-backed A2A agent plus a tiny chat REPL, exercising the mesh's a2a+support end to end: agent-card fetch with caller-side rewrite, `message/send`+routing over libp2p, and `contextId` continuity across turns.++## 1. Set your Gemini key++Edit `Dockerfile` and replace `<API_KEY>` in `ENV GEMINI_API_KEY=<API_KEY>`+(same pattern as `gemini-buddy-mcp`). Optionally override the model with+`GEMINI_MODEL` (default `models/gemini-3.6-flash`).++## 2. Host the agent on a mesh node++Assign it in `development/kind/mesh-config.yaml`:++```yaml+node-a:+node-b: chat-a2a+```++Then bring the mesh up:++```sh+make build+make kind-up+```++## 3. Enroll a local caller node++```sh+./development/kind/run-local-node.sh+```++Sidecar API lands on `127.0.0.1:9099` with token `devtoken`.++## 4. Find the provider peer++```sh+./bin/mcp-client -url http://127.0.0.1:9099/mcp -token devtoken \+ -tool discover_remote_services -args '{"type":"a2a","name":"chat"}'+```++Note the peer ID and export it: `export PEER=<peer-id>`. Discovery is+gossip-fed; retry for a few seconds after startup if the list comes back empty.++## 5. See the card rewrite++```sh+curl -s -H 'X-Sam-Authentication: Bearer devtoken' \+ "http://127.0.0.1:9099/sam/$PEER/a2a/chat/.well-known/agent-card.json" | jq+```++The interface URLs point back at this mesh path (not the agent's own address)+and `capabilities.streaming` is forced to `false` — that rewrite is what lets+a stock A2A client work against the mesh unmodified.++## 6. Chat++Requires [`uv`](https://docs.astral.sh/uv/).++```sh+uv run --with-requirements requirements.txt chat.py "http://127.0.0.1:9099/sam/$PEER/a2a/chat"+```++Tell the agent your name, then ask for it back a couple of turns later: the+client carries the `contextId` the server minted on the first reply, and the+agent keeps one Gemini chat session per context, so the answer proves the+conversation survived the mesh hop. Each turn is still its own short-lived+A2A task — `taskId` changes every turn, `contextId` is what persists.diff --git a/development/examples/chat-a2a/agent.py b/development/examples/chat-a2a/agent.pynew file mode 100644index 00000000..2f4fd4c7--- /dev/null+++ b/development/examples/chat-a2a/agent.py@@ -0,0 +1,108 @@+"""Gemini-backed A2A chat agent hosted by a node in the local dev mesh."""+import os+import time+import uuid++import uvicorn+from a2a.server.agent_execution.agent_executor import AgentExecutor+from a2a.server.agent_execution.context import RequestContext+from a2a.server.events.event_queue import EventQueue+from a2a.server.request_handlers import DefaultRequestHandler+from a2a.server.routes import create_agent_card_routes, create_jsonrpc_routes+from a2a.server.tasks.inmemory_task_store import InMemoryTaskStore+from a2a.types import (+ AgentCapabilities,+ AgentCard,+ AgentInterface,+ AgentSkill,+ Message,+ Part,+ Role,+)+from google import genai+from google.genai import types+from starlette.applications import Starlette++PORT = 7777+MODEL = os.environ.get("GEMINI_MODEL", "models/gemini-3.6-flash")++class ChatExecutor(AgentExecutor):+ """One Gemini chat session per A2A contextId; the session carries the history."""++ def __init__(self):+ self.gemini = genai.Client()+ self.chats = {}++ async def execute(self, context: RequestContext, event_queue: EventQueue) -> None:+ chat = self.chats.get(context.context_id)+ if chat is None:+ # Gemini 3 Flash defaults to thinking_level=high, which dominates latency.+ chat = self.gemini.aio.chats.create(+ model=MODEL,+ config=types.GenerateContentConfig(+ thinking_config=types.ThinkingConfig(thinking_level="minimal")+ ),+ )+ self.chats[context.context_id] = chat+ started = time.monotonic()+ reply = await chat.send_message(context.get_user_input())+ print(+ f"[chat] context={context.context_id} gemini took "+ 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,+ )+ )++ async def cancel(self, context: RequestContext, event_queue: EventQueue) -> None:+ pass+++agent_card = AgentCard(+ name="chat",+ description="Gemini-backed conversational agent; remembers the conversation per contextId",+ version="0.1.0",+ capabilities=AgentCapabilities(streaming=False),+ default_input_modes=["text"],+ default_output_modes=["text"],+ skills=[+ AgentSkill(+ id="chat",+ name="chat",+ description="Multi-turn small talk",+ tags=["chat"],+ examples=["hi, my name is Ada", "what is my name?"],+ )+ ],+ supported_interfaces=[+ AgentInterface(+ protocol_binding="JSONRPC",+ protocol_version="1.0",+ url=f"http://127.0.0.1:{PORT}/",+ )+ ],+)++handler = DefaultRequestHandler(+ agent_executor=ChatExecutor(),+ task_store=InMemoryTaskStore(),+ agent_card=agent_card,+)+# Starlette over FastAPI: the SDK generates the routes, so FastAPI would add nothing.+# JSON-RPC at "/": the mesh card rewrite drops URL subpaths, so clients land on the root.+app = Starlette(+ routes=[+ *create_jsonrpc_routes(request_handler=handler, rpc_url="/"),+ *create_agent_card_routes(agent_card=agent_card),+ ]+)++if __name__ == "__main__":+ uvicorn.run(app, host="0.0.0.0", port=PORT)diff --git a/development/examples/chat-a2a/chat.py b/development/examples/chat-a2a/chat.pynew file mode 100644index 00000000..f23fa2dc--- /dev/null+++ b/development/examples/chat-a2a/chat.py@@ -0,0 +1,44 @@+"""Minimal A2A chat REPL: resolves the agent card through the mesh and talks to it."""+import asyncio+import os+import sys+import uuid++import httpx+from a2a.client import A2ACardResolver, ClientConfig, create_client+from a2a.helpers import get_message_text+from a2a.types import Message, Part, Role, SendMessageRequest+++async def main(url: str) -> None:+ token = os.environ.get("SAM_API_TOKEN", "devtoken")+ async with httpx.AsyncClient(timeout=120, headers={"X-Sam-Authentication": f"Bearer {token}"}) as http:+ card = await A2ACardResolver(http, url).get_agent_card()+ print(f"{card.name}: {card.description}")+ for iface in card.supported_interfaces:+ print(f" {iface.protocol_binding} -> {iface.url}")+ client = await create_client(card, client_config=ClientConfig(httpx_client=http))+ context_id = None+ while True:+ try:+ text = input("you> ")+ except EOFError:+ return+ if not text.strip():+ continue+ message = Message(+ role=Role.ROLE_USER,+ message_id=str(uuid.uuid4()),+ parts=[Part(text=text)],+ context_id=context_id,+ )+ async for event in client.send_message(SendMessageRequest(message=message)):+ if event.HasField("message"):+ context_id = event.message.context_id or context_id+ print(f"agent> {get_message_text(event.message)}")+++if __name__ == "__main__":+ if len(sys.argv) != 2:+ sys.exit("usage: chat.py http://127.0.0.1:9099/sam/<peer-id>/a2a/chat")+ asyncio.run(main(sys.argv[1]))diff --git a/development/examples/chat-a2a/requirements.txt b/development/examples/chat-a2a/requirements.txtnew file mode 100644index 00000000..ee9c850d--- /dev/null+++ b/development/examples/chat-a2a/requirements.txt@@ -0,0 +1,6 @@+a2a-sdk>=1.0+google-genai>=1.0+httpx>=0.27+sse-starlette>=2.0+starlette>=0.40+uvicorn>=0.30diff --git a/development/examples/chat-a2a/sam-node-config.yaml b/development/examples/chat-a2a/sam-node-config.yamlnew file mode 100644index 00000000..910a4901--- /dev/null+++ b/development/examples/chat-a2a/sam-node-config.yaml@@ -0,0 +1,8 @@+version: "v1alpha1"+attenuation:+ policies: []+services:+ - type: "a2a"+ name: "chat"+ description: "Gemini-backed conversational A2A agent"+ target_url: "http://127.0.0.1:7777"