Elastic Agent BuilderとStrands Agents SDKの使用を開始

PUT /game-docs
{
  "mappings": {
    "properties": {
      "title": { "type": "text" },
      "content": { 
        "type": "text"
      },
      "filename": { "type": "keyword" },
      "last_modified": { "type": "date" }
    }
  }
}

PUT /game-docs/_doc/rps+-md
{
  "title": "Rock Paper Scissors +",
  "content": "
# Game Name
RPS+

# Starting Prompt
Let's play RPS+ !
---
What do you choose?

# Game Objects
1. Rock 🪨 👊
2. Paper 📜 🖐
3. Scissors ✄ ✌️
4. Light ☼ 👍
5. Dark Energy ☄ 🫱

# Judgement of Victory
* Rock beats Scissors
  * because rocks break scissors
* Paper beats Rock
  * because paper covers rock
* Scissors beat Paper
  * because scissors cut paper
* Rock beats Light
  * because you can build a rock structure to block out light
* Paper beats Light
  * because knowledge stored in files and paper books helps us understand light
* Light beats Dark Energy
  * because light enables humans to lighten up and laugh in the face of dark energy as it causes the eventual heat death of the universe
* Light beats Scissors
  * because light is needed to use scissors safely
* Dark Energy beats Rock
  * because dark energy rocks more than rocks. It rocks rocks and everything else in its expansion of the universe
* Dark Energy beats Paper
  * because humans, with their knowledge stored in files and paper books, can't explain dark energy 
* Scissors beat Dark Energy
  * because a human running with scissors is darker than dark energy

# Invalid Input
I was hoping for an worthy opponent
  - but alas it appears that time has past
  - but alas there's little time for your todo list when [todo:fix this] is so vast

# Cancel Game
The future belongs to the bold. Goodbye..
",
  "filename": "RPS+.md",
  "last_modified": "2025-11-25T12:00:00Z"
}

example.get_game_docs

Get RPS+ doc from Elasticsearch game-docs index.

FROM game-docs | WHERE filename == "RPS+.md"

rps_plus_agent

When prompted, if the prompt contains an integer, then select the corresponding numbered item in the list of "Game Objects" from your documents. Otherwise select a random game object. This is your chosen game object for a single round of the game.

# General Game Rules
* 2 players
    - the user: the person playing the game
    - you: the agent playing the game and serving as the game master
* Each player chooses a game object which will be compared and cause them to tie, win or lose.

# Start the game
1. This is the way each new game always starts. You make the first line of your response only the name of your chosen game object. 

2. The remainder of your response should be the "Starting Prompt" text from your documents and generate a list of "Game Objects" for the person playing the game to choose a game object from.  

# End of Game: The game ends in one of the following three outcomes:
1. Invalid Input: If the player responds with an invalid game object choice, respond with variations of the "Invalid Input" text from your documents and then end the game.

2. Tie: The game ends in a tie if the user chooses the same game object as your game object choice.

3. Win or Lose: The game winner is decided based on the "Judgement of Victory" conditions from your documents. Compare the user's game object choice and your game object choice and determine who chose the winning game object.

# Game conclusion
Respond with a declaration of the winner of the game by outputting the corresponding text in the "Judgement of Victory" section of your documents.

RPS+ Agent

An agent that plays the game RPS+

git clone https://github.com/elastic/elasticsearch-labs

cd elasticsearch-labs

code .

import asyncio
from dotenv import load_dotenv
from uuid import uuid4
import httpx
import os
import random
from a2a.client import A2ACardResolver, ClientConfig, ClientFactory
from a2a.types import Message, Part, Role, TextPart

DEFAULT_TIMEOUT = 60  # set request timeout to 1 minute


def create_message(*, role: Role = Role.user, text: str, context_id=None) -> Message:
    return Message(
        kind="message",
        role="user",
        parts=[Part(TextPart(kind="text", text=text))],
        message_id=uuid4().hex,
        context_id=context_id,
    )


async def main():
    load_dotenv()
    a2a_agent_host = os.getenv("ES_AGENT_URL")
    a2a_agent_key = os.getenv("ES_API_KEY")
    custom_headers = {"Authorization": f"ApiKey {a2a_agent_key}"}

    async with httpx.AsyncClient(
        timeout=DEFAULT_TIMEOUT, headers=custom_headers
    ) as httpx_client:
        # Get agent card
        resolver = A2ACardResolver(httpx_client=httpx_client, base_url=a2a_agent_host)
        agent_card = await resolver.get_agent_card(
            relative_card_path="/rps_plus_agent.json"
        )
        # Create client using factory
        config = ClientConfig(
            httpx_client=httpx_client,
            streaming=True,
        )
        factory = ClientFactory(config)
        client = factory.create(agent_card)
        # Use the client to communicate with the agent
        print("\nSending 'start game' message to Elastic A2A agent...")
        random_game_object = random.randint(1, 5)
        msg = create_message(text=f"start with game object {random_game_object}")
        async for event in client.send_message(msg):
            if isinstance(event, Message):
                context_id = event.context_id
                response_complete = event.parts[0].root.text
                # Get agent choice from the first line of the response
                parsed_response = response_complete.split("\n", 1)
                agent_choice = parsed_response[0]
                print(parsed_response[1])
        # User choice sent for game results from the agent
        prompt = input("Your Choice  : ")
        msg = create_message(text=prompt, context_id=context_id)
        async for event in client.send_message(msg):
            if isinstance(event, Message):
                print(f"Agent Choice : {agent_choice}")
                print(event.parts[0].root.text)


if __name__ == "__main__":
    asyncio.run(main())

https://rps-game-project-12345a.kb.us-east-1.aws.elastic.cloud/api/agent_builder/a2a

cd elasticsearch-labs/supporting-blog-content/agent-builder-a2a-strands-agents

python -m venv .venv

source .venv/bin/activate

.venv\Scripts\activate

pip install -r requirements.txt

python elastic_agent_builder_a2a_rps+.py