New Let’s Do DevOps post: Teaching Strands Agents to assume cross-account IAM roles and read AWS resources.
Includes full setup for AWS CLI MCP + Lambda + boto3 profile chaining and lots of debugging tips.
#DevOps #AWS #AI #Strands #LetsDoDevOps
https://www.letsdodevops.com/p/teaching-agentic-strands-agents-to
In Part 5 of my “Building a Teams Bot with AI Capabilities” series, we dive into the fun stuff: how the bot actually reads conversations, grabs context, and posts AI-generated responses back into Teams.
This one gets into message threading, attachments (SharePoint vs OneDrive 🤯), and how to make your Worker Lambda do the AI magic.
If you’re curious how GenAI really integrates with Teams — this is your guide.
Full article’s free to read, link in image
The next article in the Vera Teams series is now out! PHEW
This one covers all the magic I'm doing to make a stateless app (lambda) stateful using dynamoDB, and how Entra works to get an auth_code from the IdP, and exchange it for a token. Lots of cool Entra things there, as well as stateless architecture.
It's REAL cool, check it out: https://www.letsdodevops.com/p/teamsai-4
🚨 Part 9 of the Slack Bot with AI series is now FREE!
✅ Add a Lambda Receiver pattern
✅ Avoid Slack timeouts
✅ Improve reliability and cut costs
This wraps up the series with a clean async handoff between Lambdas so your AI bot doesn’t break Slack’s 3s rule.
Easy to implement, free to read, and MIT licensed!
https://www.letsdodevops.com/p/building-a-slack-bot-receiver
🚨 Part 8 of the Slack Bot with AI series is now FREE!
✅ Add ReRanking to improve KB response accuracy
✅ Score results by actual relevance
✅ Boost answer quality with minimal cost or latency
It’s fast, cheap, and dramatically improves results from your Confluence-powered RAG bot using AWS Bedrock.
Read it now:
https://www.letsdodevops.com/p/building-a-slack-bot-reranking
🚨 Part 7 of the Slack Bot with AI series is now FREE!
✅ Stream token-by-token responses with AWS Bedrock
✅ Show live status updates in Slack
✅ Make your bot feel faster and more responsive
This upgrade uses converse_stream() to show real-time progress instead of waiting for the full reply.
Code is open source and MIT licensed:
https://www.letsdodevops.com/p/building-a-slack-bot-streaming
Just dropped a new Building Better GenAI post:
Pre-Contextualizing Knowledge Base Results
We’re improving your RAG queries by adding a step before querying the knowledge base — using a model to standardize and extract structured data from user input. Think: turn IPs into CIDRs before hitting the KB.
It’s already improving answers in tools like NetBot (my “is it the network?” assistant). Works great when your data is structured but your users aren’t.
https://www.letsdodevops.com/p/building-better-genai-pre-contextualizing
🚨 Part 6 of the Slack Bot with AI series is now FREE!
Learn how to use AWS Bedrock’s .converse() API to let your bot read and analyze DOCX, XLSX, PDFs & more.
✅ Summarize contracts
✅ Analyze sales data
✅ Compare resumes to job reqs
✅ Give project feedback
No AWS or Python expertise required. Just clear steps and pictures.
Read it now:
https://www.letsdodevops.com/p/building-a-slack-bot-with-ai-capabilities-bdf
🚀 Part 9 of Building a Slack Bot with AI is live!
We wrap the series by solving a big problem: Slack needs fast responses, but our AI Lambda takes time. The fix? A Receiver Lambda that immediately replies to Slack, then hands off the work.
Clean, async, and cost-effective.
Catch the final post here:
https://www.letsdodevops.com/p/building-a-slack-bot-receiver
AWS Anti-Patterns: A Security Superpower?
I had the pleasure of contributing to the Cloud Security Newsletter with an interview exploring why AWS anti-patterns around Network Firewall - yes, the things we’re told not to do, and how they might actually be the key to strengthening your cloud security posture. 🚀
Curious how? Check it out here:
Link to the newsletter: https://www.cloudsecuritynewsletter.com/p/why-aws-anti-patterns-might-be-your-next-cloud-security-superpower
🚀 Exciting preview of an upcoming Let’s Do DevOps post!
What if AI could truly follow along in a conversation? I’ve been teaching DevOpsBot to understand threading and connect thoughts more effectively. I'm working diligently to get this content ready for next week! Please let me know what features you would want in a conversational AI bot!
Next up? Training it for domain-specific knowledge with jsonl.
If you're building Auth0 Applications for teams to use in their CI processes, you're likely granting most permissions (Scopes) against the Auth0 Management API, and removing some of them that you deem too dangerous.
The problem is, those Scopes keep changing as Auth0 adds new cool stuff, and keeping all of them up to date with new Scopes is a bit annoying, right?
So let's automate it. Here we use a data lookup to find all the Scopes that are possible to grant against the Auth0 Management API. We store all of them in a variable, and then use the !can(regex()) pattern to remove some of them (all the deletes).
That way, when folks need a CI, you can just reference this locals value, and you're done, forever. If new Scopes are created or removed, just run the terraform/tofu workspace again, and it'll find them, filter them, and assign them. And you can kick your feet up.
Here's the gist with code you can copy: https://gist.github.com/KyMidd/ddf51b2aa241db45532a54ebaf9ceb39
