I have been building a voice AI receptionist, and there are a lot of terms and acronyms. Some of them are normal telephony acronyms that have been around forever, like DTMF and PSTN (and might need a refresher). Some are AI acronyms, like STT, TTS, LLM and RAG (that we are now sick of hearing every day). Then there are the “glue” terms that only start to matter once you actually try to connect Telephony, Voice and LLMs. This is my attempt to write down the useful ones, A.K.A The ABCs of Voice AI Agents.
It is not a full architecture doc. It is more of a map of the words that kept coming up while building the bot.
I have a little project I’m working on playing with, MentionVault.com. It’s a platform that allows you to look for guests on various podcasts and what was mentioned in each episode. So I was thinking, I can’t be that shoeless cobbler, how come I have an application and don’t have any Observability for it?! That’s how I decided to try a Splunk O11y deployment for my app.
Splunk MLTK 5.6.0+ allows you to configure LLM inference endpoints, but the list is somewhat limited. Below, I’ll explain how you can add new LLM models to Splunk MLTK.
I started playing with n8n.io, specifically with the “My first AI Agent in n8n” workflow that comes OOTB.
I didn’t have OpenAI subscription, but I do have an Azure subscription and Azure OpenAI deployment to play with, so I replaced the “standard” OpenAI node with the Azure OpenAI one.
But when I started the execution, the Azure OpenAI Chat Model node threw an exception, straight in my face: “The response was filtered due to the prompt triggering Azure OpenAI’s content management policy.”.
I tried to follow the “Create a Custom Skill for Azure AI Search” but it failed with this error “The request is invalid. Details: The property ‘includeTypelessEntities’ does not exist on type ‘Microsoft.Skills.Text.V3.EntityRecognitionSkill’. Make sure to only use property names that are defined by the type.”
I am learning about different concepts and architectures used in the LLM/AI space and one of them is Retrieval-Augmented Generation. As always I prefer learning concepts by tinkering with them and here is my first attempt at learning about RAG and Vector Databases.
While there is a QuickStart example on the Streamlit site that shows how to connect to OpenAI using LangChain I thought it would make sense to create Streamlit Langchain Quickstart App with Azure OpenAI.
Pinecone documentation is quite good, but when I wanted to create a free pod index in Pinecone using Python, I didn’t know what parameters I should supply.
I mainly use Mac for work, but occasionally need access to a Windows box. I am using UTM to achieve that. I have noticed that if you leave your Windows VM running and then your host Mac goes to sleep (overnight for example), there will be a time drift on the VM. So here is how to fix time drift on UTM Windows VM.
