Sre Publish

AWS SRE health check with FinOps — queries CloudWatch, SQS DLQ, and Cost Explorer, generates a Bedrock-powered incident diagnosis (Contexto, Soluções, CTAs), and sends a structured report to Telegram.

AWS SRE Report

Activate this skill when the user asks for an SRE health check, infrastructure status, pipeline report, or AWS cost overview on a stack that uses:

• Amazon SQS (main queue + Dead Letter Queue)
• AWS Lambda functions
• Amazon CloudWatch metrics
• AWS Cost Explorer
• Amazon Bedrock for AI diagnosis
• Telegram for notifications

Credentials required

Variable	Description
AWS_REGION	AWS region of your workload (e.g. us-east-1)
TELEGRAM_BOT_TOKEN	Telegram Bot API token
TELEGRAM_CHAT_ID	Telegram chat ID to receive the report
DLQ_URL	Full SQS Dead Letter Queue URL
AUDIO_QUEUE_URL	Full SQS main queue URL
LAMBDA_FUNCTIONS	Comma-separated Lambda function names to monitor
MONTHLY_BUDGET_USD	Monthly AWS budget in USD (default: 50)
SLO_TARGET	SLO target percentage (default: 99.5)

Required AWS IAM permissions

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": ["cloudwatch:GetMetricStatistics", "cloudwatch:ListMetrics"],
      "Resource": "*"
    },
    {
      "Effect": "Allow",
      "Action": ["sqs:GetQueueAttributes", "sqs:ReceiveMessage"],
      "Resource": "<YOUR_DLQ_ARN>"
    },
    {
      "Effect": "Allow",
      "Action": ["ce:GetCostAndUsage"],
      "Resource": "*"
    },
    {
      "Effect": "Allow",
      "Action": ["bedrock:InvokeModel"],
      "Resource": [
        "arn:aws:bedrock:*::foundation-model/anthropic.*",
        "arn:aws:bedrock:*:*:inference-profile/*"
      ]
    }
  ]
}

Step-by-step execution

1 — Import and configure the reporter

import { SreReporter } from "openclaw-aws-sre-report";

const reporter = new SreReporter({
  region:           process.env.AWS_REGION!,
  telegramBotToken: process.env.TELEGRAM_BOT_TOKEN!,
  telegramChatId:   process.env.TELEGRAM_CHAT_ID!,
  dlqUrl:           process.env.DLQ_URL!,
  audioQueueUrl:    process.env.AUDIO_QUEUE_URL!,
  lambdaFunctions:  process.env.LAMBDA_FUNCTIONS!.split(","),
  monthlyBudgetUsd: parseFloat(process.env.MONTHLY_BUDGET_USD ?? "50"),
  sloTarget:        parseFloat(process.env.SLO_TARGET ?? "99.5"),
});

2 — Run the report

const report = await reporter.run();

console.log(`SLO: ${report.sloValue.toFixed(1)}%`);
console.log(`DLQ: ${report.dlqCount} messages`);
console.log(`Lambda errors (24h): ${report.lambdaErrors}`);
console.log(`AWS cost yesterday: $${report.costs?.yesterday.toFixed(2) ?? "N/A"}`);

if (report.incident) {
  console.log("\n🔴 Incident detected:");
  console.log("Contexto:", report.incident.contexto);
  console.log("CTAs:", report.incident.ctas);
}

3 — What the Telegram report looks like

Healthy pipeline:

📊 AWS SRE Report

🔋 Pipeline (24h)
  Processed: 47   DLQ: 0 ✅
  Lambda errors: 0 ✅

📈 SLO
  Availability: 100.0% ✅ (target 99.5%)
  Budget burned: 0.0%

💰 FinOps (2026-04-12)
  AWS: $2.14
  ├ Fargate/ECS: $1.68 (79%)
  └ Bedrock: $0.28 (13%)
  Cost/item: $0.0455
  MTD: $21.40 / $50 (43%) ✅

When an incident is detected:

🔴 Incident Detected

Context
3 DLQ messages indicate a validation error in the pipeline
starting yesterday at 14:00 UTC. The DLQ sample shows a
ZodError — likely a malformed payload from an edge case.

Findings
  ✓ DLQ sample read: ZodError on field "target" (null not allowed)

Next Actions
  1. Tail ECS logs: aws logs tail /ecs/my-worker --follow --region us-east-1
  2. Inspect DLQ: aws sqs receive-message --queue-url <DLQ_URL> --region us-east-1
  3. Redrive after fix: aws sqs start-message-move-task --source-queue-url <DLQ_URL> --destination-queue-url <QUEUE_URL>

MUST DO

• Always send the report to Telegram even if Cost Explorer or Bedrock calls fail — use "N/A" for unavailable data
• Never delete DLQ messages — ReceiveMessage is used for peek only
• Escape all MarkdownV2 special characters in dynamic values before building the Telegram message
• Use us-east-1 as the Cost Explorer endpoint regardless of your workload region (AWS requirement)

MUST NOT DO

• Abort if Cost Explorer returns no data — the 24 h lag is normal, skip the FinOps block gracefully
• Purge or redrive DLQ messages automatically — always present these as operator CTAs
• Hard-code any AWS account IDs, ARNs, or resource names

---

Credits

Built by @vieiradiego — open-sourced as give back to the community. MIT License.