Adding a new assistant in Ragna

Following up from my previous post on Ragna, I wanted to share following the announcement of Meta’s Llama2 13b model availability within Amazon Bedrock, how you can incorporate that.

I have also put together a GitHub repo that shares the code, something that I got quite a few questions from the original post.

Adding Meta’s Llama2

As with adding Amazon Bedrock’s Anthropic’s Claude support, it was pretty straight forward to modify the original code to add support for Llama2.

Unboxing Ragna: Getting hands on and making it to work with Amazon Bedrock

I am always on the look out for interesting new projects to check out, and this week I came across Ragna, an open source Retrieval Augmented Generation RAG orchestration framework. It is a new project with a committed and active community, so I wanted to find out more about this project.

What piqued my interest was reading this blog post, Unveiling Ragna: An Open Source RAG-based AI Orchestration Framework Designed to Scale From Research to Production which takes a look at the background, or as I like to think of it, the “scratch that needed to be itched”. I couldn’t help notice some parallels from an architectural perspective to one of my favourite open source projects, Apache Airflow.

Getting started with Griptape

Getting gnarly with AI

When I was much much younger, I was known to (very occasionally) drop an ollie, kick-turn, or very occasionally dare a drop in on my trusty skateboard. Thanks to the grip tape, my shoes would stick to the board, and gave me the confidence I could try these tricks. I was always grateful to that grip tape and who knew all these years later, I would become re-aquatinted with it (albeit in a different form!)

How are you using these new AI coding assistants like Amazon CodeWhisperer? I want to share a quick story about how I am finding some success with using these tools, which I hope will encourage you to experiment for yourselves.

As I work on new projects, demos, blog posts, I often need to do supporting activities such as creating or cleaning up AWS environments, which I sometimes do by hand but mostly do via the AWS CLI or via Python scripts and boto3. I am working on a some demos to show how you can use tools like Griptape, which is a Python framework for AI workflows and pipelines which you can think of as an enterprise grade alternative to LangChain. I needed to delete a bunch of resources in my Amazon Sagemaker environment that I had forgotten to clean up. I had a few hundred SageMaker Models that were really distracting me as I love to have clean environments when I start writing new content.

I have written in the past about setting up developer environments and tools when working with Apache Airflow. Today I came across a new tool from Kaxil Naik, directory of engineering at Astronomer and all round Apache Airflow good guy. Kaxil has put together airflowctl, a command-line tool for managing Apache Airflow™ projects, and making it super easy to get up and running. What does it do? Well, it helps you install and use different versions of Apache Airflow, work with Variables and Connections, provide live logs, and more.

Update, August 1st

I have been running this project for five weeks now, so have a better understanding of costs. My bill in July was $0.85, and I suspect this will be reasonably stable over the coming months. I think this represents great value! I will report back at the end of the year and update this

Like many folk, I run a personal blog. This blog runs Hugo and turns my markdown pages into static content, which I then deploy on Netlify. One of the key things about running a blog is understanding how readers are interacting with your content, and this is where web analytics solutions can help you. A common solution to this has been to use something like Google Analytics as this typically is very easy to integrate into your website.

This is the follow up post to Integrating Keycloak as my Identity Provider for IAM Identity Centre: Part one, deploying Keycloak on AWS, where I looked at how to deploy Keycloak on AWS in order to have an Identity Provider to use when configuring AWS Identity Centre. In this post, I am going to use that setup, and show you how I configured it to integrate with AWS Identity Centre to provide access to my AWS resources.

CIFS (Common Internet File System) and SMB (Server Message Block) are both Windows file-sharing protocols used in storage systems. As part of a new demo/blog post that looks at how to use data stored on SMB/CIF file shares with Apache Airflow, I have been exploring the various options of creating SMB/CIF compatible resources. (There are LOTS of ways you could do this, so there is plenty for me to play around with!)

Integrating Keycloak as my Identity Provider for IAM Identity Centre: Part one, deploying Keycloak on AWS

“It was the best of times, it was the worst of times…” A Tale of Two Cities

It started out innocently enough. As part of working on a new blog post, I needed a way to use an open source tool called saml2aws that generates AWS short lived credentials that you can use to access your AWS resources. This is a pretty common pattern, and a good practice to do and for many organisations this means they integrate with their “user directory” - sometimes this might be something like Active Directory, or an LDAP server. AWS IAM Identity Centre (which is the new name for AWS Single Sign-On) allows you to integrate those identity providers into your AWS Account, allowing your users to authenticate against that user directory and then get access to AWS resources.

Managed Workflows for Apache Airflow (MWAA) recently launched a new feature that a lot of folk had been asking for, which was the ability to add additional libraries, binaries, or environment variables when launching Airflow workers. If you missed the announcement, Amazon MWAA now supports Shell Launch Scripts, this new capability allows you to easily do this by creating a script and then configuring your MWAA environments to use that script during the start-up phase. The MWAA documentation has been updated to show you how you can use this, and you can read about that in the page, Using a startup script with Amazon MWAA. This is essential reference material for understanding more about what you can do with this new feature.