AI doesn't always have to rely on the cloud.You can use local LLMs for a lot and give them access to information from the internet without having to expose them outside your home network.I built an AI-generated morning news briefing using a local LLM that never touches the cloud.
I wanted a morning briefing that was tailored to me There's too much news to take everything in The idea was simple.My Alexa smart speakers used to give me a morning news briefing through a feature called Flash Briefing.It would give me local weather updates, headlines from major news outlets, and other things such as sports and stocks.
The problem was that a lot of the content was news I wasn't interested in.I wanted to create my own Flash Briefing that only gave me content from the sources that I wanted.I could then have a smart speaker read out a custom morning briefing that covered the exact topics that I was interested in and nothing else.
Beelink S13 PRO CPU Celeron FCBGA1264 3.6GHz Graphics Integrated Intel Graphics 24EUs 1000MHz The Beelink Mini S13 Pro desktop PC is a ultra-compact computer powered by the Intel N150 processor.Shipping with 16GB of DDR4 RAM and a 500GB SSD, this micro desktop is perfect for a variety of workloads.From running simple server programs to replacing your old PC, the Beelink S13 Pro is up to the task. Memory 16 GB DDR4 Storage 500GB Operating System Windows 11 Home Dimension 4.52 x 4 x 1.54 inches USB Ports 4 $299 at Amazon Expand Collapse RSS feeds are still a great source I can pull stories from the feeds that I want I needed a simple way to pull the latest news content from specific providers or on specific topics.
The solution was to turn to a source that's been around for decades.RSS feeds may not be the most sophisticated way to access information, but they're simple to use, and most major news sites, blogs, and publications still offer RSS feeds.I managed to find feeds for all the sources that I wanted to use, including BBC News, The Guardian, and a site dedicated to the greatest soccer team in the history of the planet, Liverpool FC.
I decided to create the automation that powers my morning briefing in the self-hosted n8n automation software.There's an RSS Read node in n8n that can pull the latest items from an RSS feed if you provide the relevant URL.I used three of these nodes to pull all the latest stories from the RSS feeds.
Using all the stories returned by the RSS feeds would have given me more than 70 different stories, which was far too many for my morning briefing.I used Limit nodes to limit the outputs to the first three stories from each source.This gave me a list of nine headlines which seemed about the right amount for my briefing.
Related How I Use Home Assistant to Describe Who's At the Door Using AI Get AI-generated descriptions of anyone your video doorbell detects.Posts By Adam Davidson A list of headlines is not a morning briefing A local LLM turns a list of headlines into a proper text Having a smart speaker read out a list of nine different news headlines one after another is not my idea of a good morning briefing.I wanted it to feel like a real person was delivering the news to me in a natural way.
That's why I knew I needed to bring an LLM into the loop.LLMs are great at taking a bunch of text and turning it into natural-language prose.I could use an LLM to convert my list of headlines into a coherent morning briefing that would be more pleasant to listen to.
Since I wanted my news briefing to work completely locally, I decided to use a local LLM.While I don't have powerful hardware, I can still run some reasonably useful models.The output takes some time to generate, but the plan was to have the LLM do its work long before I woke up, so that the morning briefing was ready and waiting for me each morning.
I run Ollama in a container on my mini PC, and using n8n, I'm able to send a prompt to Ollama with the news headlines from the RSS feeds along with instructions to combine them into a coherent spoken news briefing.The generation can take several minutes, but the automation is set to run early in the morning, so it's always finished by the time I wake up.My smart speaker reads the briefing each morning I'm up to speed before I'm even out of bed The final piece of the puzzle was getting my smart speaker to read my morning briefing to me each morning.
I use Home Assistant to control my smart home, and my smart speaker was already connected to Home Assistant, so it was a no brainer to pass the text of the morning briefing to Home Assistant, which could then use text-to-speech (TTS) to get the smart speaker in my bedroom to read the morning briefing aloud.The result is exactly what I wanted.Each morning, the morning briefing plays through my smart speaker, telling me about the latest news from my chosen sources.
Just this morning, for example, I discovered that Liverpool had signed a new player, news that had broken after I'd gone to bed.AI doesn't always need to rely on the cloud Even with relatively modest hardware, there's still a lot you can do with a local LLM.Creating my own custom morning briefing works really well and replicates one of the useful features of Alexa without any of my information getting sent to the cloud.
Read More
