Unofficial Azure Club

Automated Prompt Optimization in DSPy: Mechanisms, Algorithms, and Observability

1. Introduction: The Shift from Manual Prompting to Programmatic Optimization The advent of powerful Large Language Models (LLMs) has revolutionized natural language processing, yet harnessing their full potential often hinges on the art and science of prompt engineering.1 Traditionally, crafting effective prompts has been a manual, iterative, and often

ReAct Agents: Building From Scratch - Native Function Calling vs. Custom TAO Parsing

Introduction In this article, I'll walk you through two different ways to implement ReAct agents without relying on any existing frameworks: 1. Native Function Calling: Tapping into the built-in capabilities of modern LLMs 2. Custom TAO Parsing: A more flexible text-based approach for parsing Thoughts, Actions, and Observations I've deliberately

Building an English Dictionary with LLM

In my previous blog, I mentioned about online dictionary feature from my English learning app. Today, I'll share the implementation details and how I overcame several challenges using Large Language Models (LLMs). The Challenge The requirement was straightforward: when a user selects a word in a passage and clicks "explain,

Vibe Coding Featured

My Vibe Coding Journey - Developing an AI-Powered English Learning Application

I want to share my story about vibe coding, I started this idea to assist my son's english learning, I also wanted to test to see if I can rely on generative ai to complete a full project including client and server. My goal was twofold: to support his learning

Analysis of the latest advancements in AI(Created by AI itself)

Unleashed: The Dawn of a Smarter Tomorrow Imagine a world where creativity knows no bounds, where every industry thrives on innovation, and mundane tasks are a thing of the past. This isn't just a dream—it’s the reality shaped by AI’s most significant advancements. Let’s dive into

Featured

Cost-Effective Private Large Language Model Inference on Azure Kubernetes Service

Introduction This article provides a method for private and cost-optimized deployment of large language models on the Azure cloud. It assumes the reader has foundational knowledge of large language models and Kubernetes. It utilizes AKS spot instances, quantization techniques, and batching inference to reduce compute costs compared to traditional deployment

WASM Featured

Building a WASM Serverless Solution with KEDA HTTP Add-on and Slight Containerd Shim

Introduction In this article, we will explore how to use KEDA, KEDA HTTP add-on, with Slight Containerd Shim to build a WASM serverless solution. The solution provides scale to/from zero support, reduced time for cold start, and cloud integration capabilities such as Azure blob, App Configuration, ServiceBus etc. The

Deploy Kafka without ZooKeeper on Kubernetes

Most of the work is based on https://github.com/bitnami/charts/issues/13624, this article here is only for documentation purposes only, we use community kafka helm chart from bitnami/kafka First, we need to have a values.yaml files with content in below, to test purpose, we set

Linux

How to Setup an OpenVPN Server on Azure (Updated)

This is an updated version for my previous article How to Setup an OpenVPN Server on Azure to support configure OpenVPN on Ubuntu 22.04 since a lot of changes happened on Easy-RSA. Prerequisites * Ubuntu 22.04 VM deployed in Azure at least with one NIC which has public IP

Kusto Detective Agency Case 5

El Puente - Big heist Hello. It's going to happen soon: a big heist. You can stop it if you are quick enough. Find the exact place and time it’s going to happen. Do it right, and you will be rewarded, do it wrong, and you will miss your

Kusto Detective Agency Case 4

El Puente - Ready to play? Hello. I have been watching you, and I am pretty impressed with your abilities of hacking and cracking little crimes. Want to play big? Here is a prime puzzle for you. Find what it means and prove yourself worthy. 20INznpGzmkmK2NlZ0JILtO4OoYhOoYUB0OrOoTl5mJ3KgXrB0[8LTSSXUYhzUY8vmkyKUYevUYrDgYNK07yaf7soC3kKgMlOtHkLt[kZEclBtkyOoYwvtJGK2YevUY[v65iLtkeLEOhvtNlBtpizoY[v65yLdOkLEOhvtNlDn5lB07lOtJIDmllzmJ4vf7soCpiLdYIK0[

Kusto Detective Agency Case 3

Bank robbery We have a situation, rookie. As you may have heard from the news, there was a bank robbery earlier today. In short: the good old downtown bank located at 157th Ave / 148th Street has been robbed. The police were too late to arrive and missed the gang, and

Kusto Detective Agency Case 2

Election fraud? The mayor of Digitown, Mrs. Gaia Budskott, has found herself in quite a pickle. The election for the city’s mascot was run online for the first time, and it was a huge success! Or was it?? Over 5 million people voted. Four candidates made it to the

Kusto Detective Agency Case 1

The rarest book is missing! This was supposed to be a great day for Digitown’s National Library Museum and all of Digitown. The museum has just finished scanning more than 325,000 rare books, so that history lovers around the world can experience the ancient culture and knowledge of

Run WASM applications from Kubernetes

* Interested in running WASM workload from Kubernetes? * Interested in turning your Kubernetes cluster to support multiple container shims? * Even more, wasm with dapr support? The steps documented here help run WASM workload from Kubernetes cluster based on great work from * Containerd Wasm Shims * WasmEdge Runtime - Kubernetes + Containerd * Spin Sample

Architectural Characteristics

Azure

Azure Active-Active High-Availability Solutions

To show various options to deploy an application in active-active high availability mode, including - Azure Traffic Manager - Azure Front Door - Azure Cross Region Load Balancer

Business Architecture Reference Model

Create SSH VPN over PPP

sudo vi /lib/systemd/system/pppd.service [Unit] Description=PPP over Serial link After=network.target [Service] ExecStart=/usr/sbin/pppd nodetach noauth silent nodeflate pty "/usr/bin/ssh root@ /usr/sbin/pppd nodetach notty noauth" ipparam vpn 172.18.8.1:172.18.8.2 Restart=on-failure [Install] WantedBy=

Enable WAF with Modsecurity from Ingress Nginx

Export modsecurity.conf from ingress nginx controller POD_NAME=$(kubectl get pods --namespace ingress-nginx -l "app.kubernetes.io/component=controller,app.kubernetes.io/name=ingress-nginx" -o jsonpath="{.items[0].metadata.name}") kubectl exec -it $POD_NAME -n ingress-nginx -- cat /etc/nginx/modsecurity/modsecurity.conf > modsecurity.conf Change configuration to

Azure

Azure Custom Roles - Cook Guide

This article tries to provide a workflow so you can easily customize roles to suit your needs Azure role-based access control (Azure RBAC) has several Azure built-in roles that you can assign to users, groups, service principals, and managed identities. RBAC ensures the proper segregation of administration between the different

Cloud

Cloud Migration Checklist

Cloud migration checklist

Kubernetes Featured

Deploy Traefik2 Ingress with Cert-Manager and Azure AD Authentication

This article discuss on how to integrate your application with traefik2 for azure ad authentcation, automatic certificate generate as well as multiple traefik instances support for community version.

Azure Hub-Spoke, Hub-Hub Setup

Thsoe setups are servered as architecture reference for network design in Azure. A simple hub spoke setup with NVA Another hub spoke setup with virtual network gateway Hub-hub vnet setup with NVA Hub-hub plus on-premise network setup with VPN gateway

Load Balancer Featured

Azure load balancer SNAT behavior explained - Annotations to tcp port numbers reused, ACK with wrong sequence number plus RST from 3-way handshake and SNAT port exhaustion

This article will address azure external load balancer and focus on SNAT, explains a few of behaviors seen from network trace, provides a few of suggestions for application when it is behind of load balancer and requires SNAT.