How to Deploy an AI Agent

1from claude_agent_sdk import query, ClaudeAgentOptions, ResultMessage
2
3SLOTS = ["Tue 10am", "Tue 2pm", "Wed 9am"]
4
5async def run_scheduling_agent(patient_request: str) -> str:
6    options = ClaudeAgentOptions(
7        system_prompt=(
8            f"You are a dental clinic scheduling assistant. Open slots: {SLOTS}. "
9            "Pick the closest matching slot to the patient's preference and confirm it."
10        ),
11        allowed_tools=[],
12    )
13    final_text = ""
14    async for message in query(prompt=patient_request, options=options):
15        if isinstance(message, ResultMessage) and message.subtype == "success":
16            final_text = message.result
17    return final_text

1from fastapi import FastAPI, HTTPException
2from pydantic import BaseModel
3from typing import Any, Dict
4
5app = FastAPI(title="Scheduling Agent")
6
7class InvocationRequest(BaseModel):
8    input: Dict[str, Any]
9
10class InvocationResponse(BaseModel):
11    output: Dict[str, Any]
12
13@app.post("/invocations", response_model=InvocationResponse)
14async def invoke_agent(request: InvocationRequest):
15    prompt = request.input.get("prompt", "")
16    if not prompt.strip():
17        raise HTTPException(status_code=400, detail="'prompt' must be a non-empty string")
18    try:
19        reply = await run_scheduling_agent(prompt)
20    except Exception as e:
21        raise HTTPException(status_code=500, detail=f"Agent processing failed: {e}")
22    return InvocationResponse(output={"message": reply})
23
24@app.get("/ping")
25async def ping():
26    return {"status": "healthy"}

1uvicorn agent:app --host 0.0.0.0 --port 8080
2
3# In another terminal:
4curl http://localhost:8080/ping
5curl -X POST http://localhost:8080/invocations \
6  -H "Content-Type: application/json" \
7  -d '{"input": {"prompt": "I need a cleaning next Tuesday morning"}}'

1# Dockerfile
2FROM --platform=linux/arm64 python:3.11-slim
3
4WORKDIR /app
5COPY requirements.txt .
6RUN pip install --no-cache-dir -r requirements.txt
7
8COPY agent.py .
9
10EXPOSE 8080
11CMD ["uvicorn", "agent:app", "--host", "0.0.0.0", "--port", "8080"]

1docker buildx build --platform linux/arm64 \
2  -t <account-id>.dkr.ecr.us-west-2.amazonaws.com/scheduling-agent:latest --push .
3
4import boto3
5
6client = boto3.client("bedrock-agentcore-control", region_name="us-west-2")
7response = client.create_agent_runtime(
8    agentRuntimeName="dental_scheduling_agent",
9    agentRuntimeArtifact={
10        "containerConfiguration": {
11            "containerUri": "<account-id>.dkr.ecr.us-west-2.amazonaws.com/scheduling-agent:latest"
12        }
13    },
14    networkConfiguration={"networkMode": "PUBLIC"},
15    roleArn="arn:aws:iam::<account-id>:role/AgentRuntimeRole",
16    lifecycleConfiguration={"idleRuntimeSessionTimeout": 300, "maxLifetime": 1800},
17)
18print(response["agentRuntimeArn"])

1import json, uuid
2
3agent_core_client = boto3.client("bedrock-agentcore", region_name="us-west-2")
4session_id = str(uuid.uuid4()) + str(uuid.uuid4())  # ensure 33+ characters
5
6resp = agent_core_client.invoke_agent_runtime(
7    agentRuntimeArn=response["agentRuntimeArn"],
8    runtimeSessionId=session_id,
9    payload=json.dumps({"input": {"prompt": "I need a cleaning next Tuesday morning"}}),
10    qualifier="DEFAULT",
11)
12print(json.loads(resp["response"].read()))
13
14# Once the booking conversation is done:
15agent_core_client.stop_runtime_session(
16    agentRuntimeArn=response["agentRuntimeArn"], runtimeSessionId=session_id, qualifier="DEFAULT",
17)