LLaVA (Large Language-and-Vision Assistant) was updated to version 1.6 in February. I figured it was time to look at how to use it to describe an image in Node.js. LLaVA 1.6 is an advanced vision-language model created for multi-modal tasks, seamlessly integrating visual and textual data. Last month, we looked at how to use the official Ollama JavaScript Library. We are going to use the same library, today.
Basic CLI Example
Let’s start with a CLI app. For this example, I am using my remote Ollama server but if you don’t have one of those, you will want to install Ollama locally and replace const ollama = new Ollama({ host: 'http://100.74.30.25:11434' }); with const ollama = new Ollama({ host: 'http://localhost:11434' });.
To run it, first run npm i ollama and make sure that you have "type": "module" in your package.json. You can run it from the terminal by running node app.js <image filename>. Let’s take a look at the result.
Its ability to describe an image is pretty awesome.
Basic Web Service
So, what if we wanted to run it as a web service? Running Ollama locally is cool and all but it’s cooler if we can integrate it into an app. If you npm install express to install Express, you can run this as a web service.
The web service takes posts to http://localhost:4040/describe-image with a binary body that contains the image that you are trying to get a description of. It then returns a JSON object containing the description.
So, how can we get a proper answer? Ten years ago, when I wrote “The Milwaukee Soup App”, I used the Kimono (which is long dead) to scrape the soup of the day. You could also write a fiddly script to scrape the value manually. It turns out that there is another option, though. You could use Scrapegraph-ai. ScrapeGraphAI is a web scraping Python library that uses LLM and direct graph logic to create scraping pipelines for websites, documents, and XML files. Just say which information you want to extract and the library will do it for you.
Let’s take a look at an example. The project has an official demo where you need to provide an OpenAI API key, select a model, provide a link to scrape, and write a prompt.
As you can see, it reliably gives you the flavor of the day (in a nice JSON object). It will go even further, though because if you point it at the monthly calendar, you can ask it for the flavor of the day and soup of the day for the remainder of the month and it can do that as well.
I am running Python 3.12 on my Mac but when you run pip install scrapegraphai to install the dependencies, it throws an error. The project lists the prerequisite of Python 3.8+, so I downloaded 3.9 and installed the library into a new virtual environment.
Let’s see what the code looks like.
You will notice that just like in yesterday’s How to build a RAG system post, we are using both a main model and an embedding model.
At this point, if you want to harvest flavors of the day for each location, you can do so pretty simply. You just need to loop through each of Culver’s location websites.
Have a question, comment, etc? Please feel free to drop a comment, below.
Back in January, we started looking at AI and how to run a large language model (LLM) locally (instead of just using something like ChatGPT or Gemini). A tool like Ollama is great for building a system that uses AI without dependence on OpenAI. Today, we will look at creating a Retrieval-augmented generation (RAG) application, using Python, LangChain, Chroma DB, and Ollama. Retrieval-augmented generation is the process of optimizing the output of a large language model, so it references an authoritative knowledge base outside of its training data sources before generating a response. If you have a source of truth that isn’t in the training data, it is a good way to get the model to know about it. Let’s get started!
Your RAG will need a model (like llama3 or mistral), an embedding model (like mxbai-embed-large), and a vector database. The vector database contains relevant documentation to help the model answer specific questions better. For this demo, our vector database is going to be Chroma DB. You will need to “chunk” the text you are feeding into the database. Let’s start there.
Chunking
There are many ways of choosing the right chunk size and overlap but for this demo, I am just going to use a chunk size of 7500 characters and an overlap of 100 characters. I am also going to use LangChain‘s CharacterTextSplitter to do the chunking. It means that the last 100 characters in the value will be duplicated in the next database record.
The Vector Database
A vector database is a type of database designed to store, manage, and manipulate vector embeddings. Vector embeddings are representations of data (such as text, images, or sounds) in a high-dimensional space, where each data item is represented as a dense vector of real numbers. When you query a vector database, your query is transformed into a vector of real numbers. The database then uses this vector to perform similarity searches.
You can think of it as being like a two-dimensional chart with points on it. One of those points is your query. The rest are your database records. What are the points that are closest to the query point?
Our main model for this demo is going to be phi3. It is a 3.8B parameters model that was trained by Microsoft.
LangChain
You will notice that today’s demo is heavily using LangChain. LangChain is an open-source framework designed for developing applications that use LLMs. It provides tools and structures that enhance the customization, accuracy, and relevance of the outputs produced by these models. Developers can leverage LangChain to create new prompt chains or modify existing ones. LangChain pretty much has APIs for everything that we need to do in this app.
The Actual App
Before we start, you are going to want to pip install tiktoken langchain langchain-community langchain-core. You are also going to want to ollama pull phi3 and ollama pull nomic-embed-text. This is going to be a CLI app. You can run it from the terminal like python3 app.py "<Question Here>".
You also need a sources.txt file containing the URLs of things that you want to have in your vector database.
The May 7th event is too recent to be in the model’s training data. This makes sure that the model knows about it. You could also feed the model company policy documents, the rules to a board game, or your diary and it will magically know that information. Since you are running the model in Ollama, there is no risk of that information getting out, too. It is pretty awesome.
Have any questions, comments, etc? Feel free to drop a comment, below.
When we looked at how to dockerize a node app, I said that you create a docker image and then run it as a container. So, how do you list the docker images on a system? You run docker images.
Like a VM or a system running on bare metal, you can get a shell inside of the docker container. The first step is knowing the container ID for the container you want a shell for. If you look at the output from the docker ps command, you can find it.
Once you know that the image is there, know if it is running or not, and have a shell inside the container, you should be able to find what is wrong with your container.
Have a questions, comment, etc? Feel free to drop a comment, below.
Back in 2022, I created “Good Morning, Milwaukee!“. It is a bot that posts every day at 6 am with the weather, the times for sunrise and sunset, and a photo from around the city. When I first wrote it, I wrote it in Node and put it up on Pipedream. Lately, there have been some issues with the weather API that it was using, so I decided to replace it with the OpenWeather API but I figured that while I was at it, I would rewrite it in Python, dockerize it, and run it on my new home lab server.
Let’s start with what the actual Python script looks like.
If you want to reuse this code to create your own bot, there are variables at the top for api_key, zip_code, and mastodon_access_token. The actual posting is done using Mastodon.py.
So, what would the Dockerfile look like?
You’ll notice that it also needs a requirements.txt and a crontab file. Lets see what those look like.
Just make sure that you have a newline at the end of your crontab file. At this point, you can run docker build -t gmmke-app . to build the docker image and then run docker run -d gmmke-app run the container.
In yesterday’s post, we asked the basic question of what is machine learning. I hoped to illustrate the similarities and differences between artificial intelligence and machine learning. Lately, on this site, we have been spending a bit of time using Python and I wanted to take a moment today to look at a great library for machine learning in Python.
Scikit-learn is the go-to library for machine learning with an amazing ecosystem of plugins. It is open-source and supports supervised and unsupervised learning. It also provides various tools for model fitting, data preprocessing, model selection, model evaluation, and many other utilities. After you python3 -m venv EnvironmentName and source EnvironmentName/bin/activate, you can install it by running pip install scikit-learn. At that point, you can reference it in your code as sklearn.
The way that scikit-learn works is that you start with some data, you give it to a model, the model learns from it, and then you will be able to make predictions. The common notation is splitting up the data into a part called X (everything you are using to make a prediction) and another part called Y (the prediction you are interested in making). The X could be information about a house (square feet, number of bathrooms, etc) where Y is the house price, or X could be a patient’s health statistics where Y is whether or not they develop diabetes. The model then uses X to try to predict Y.
In the above code, we load the 20,640 records and 9 columns into the data variable and then we set the things that we are using to make a prediction to X and the prediction that we are interested in making to y. So, what are the feature (column) names for the data? If you print(data.feature_names), it will print them.
sklearn.model_selection
Once you have data, you can start working on creating a model. The model itself is nothing more than a Python object but the goal after you create it is to train it. You will want to split your data into a training set and a test set. Using <a href="https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html#sklearn.model_selection.train_test_split">train_test_split</a> in sklearn.model_selection, you can split it into 70% of the data for training the model and 30% of the data for testing the model (or whatever split you want).
In the above example, we are taking any X values except num_preg (the number of pregnancies) that have the value 0 and setting it to the mean. That makes it so that missing values don’t scew things when you go to train the model.
The neat thing about .fit() is that if you want to swap out the KNeighborsRegressor model with a new one, .fit() still works just the same. Let’s look at what it would look like using a linear regression model.
In the above code, we are predicting the value for y and then comparing it against the actual value of y. Using just the training data, it is predicting the values with a 75.23% level of accuracy.
So, what is next?
In a future post, I want to step through the whole process of picking a statement to test, adjusting the data, building and training a model, testing, adjusting the model, and making predictions. Let’s save that for another day, though.
Last week, we went over some basics of Artificial Intelligence (AI) using Ollama, Llama3, and some custom code. Artificial intelligence (AI) encompasses a broad range of technologies designed to enable machines to perform tasks that typically require human intelligence. These tasks include understanding spoken or written language, recognizing visual patterns, making decisions, and providing recommendations. Machine learning (ML) is a specialized subset of AI that focuses on developing systems that improve their performance over time without being explicitly programmed. Instead, ML algorithms analyze and learn from large datasets to identify patterns and make decisions based on these insights. This learning process allows ML models to make increasingly accurate predictions or decisions as they are exposed to more data.
A few months ago, I added Liner to the resource page of my website. It allows you to easily train an ML model so that you can do image, text, audio, or video classification, object detection, image segmentation, or pose classification. I created “Is this Joe or Not Joe?” using that tool. TensorFlow.js is running client-side with a model that is trained on a half dozen examples of photos that are Joe and a half dozen examples of photos that are not Joe. You can supply a photo and get a prediction if Joe is in the image or not. You can always retrain the existing model with more examples. That is an example of machine learning.
So, you can think of ML as a subset of AI and Deep Learning (DL) as a subset of ML.
Have any questions, comments, etc? Please feel free to drop a comment, below.
Yesterday, we looked at how to write a JavaScript app that uses Ollama. Recently, we started to look at Python on this site and I figured that we better follow it up with how to write a Python app that uses Ollama. Just like with JavaScript, Ollama offers a Python library, so we are going to be using that for our examples. Also just like we did with the JavaScript demo, I am going to be using the generate endpoint instead of the chat endpoint. That keeps things simpler but I am going to explore the chat endpoint also at some point.
Install the Ollama Library
The first step is to run pip3 install ollama from the terminal. First, you need to create a virtual environment to isolate your project’s libraries from the global Python libraries.
At this point, we can start writing code. When we used the web service earlier this week, we used the generate endpoint and provided model, prompt, and stream as parameters. We set the stream parameter to false so that it would return a single response object instead of a stream of objects. When using the python library, the stream parameter isn’t necessary because it returns a single response object by default. We still provide it with a model and a prompt, though.
If you run it from the terminal, the response will look familiar.
If you pip install flask to install flask, you can host a simple HTTP page at port 8080 and with the magic of json.loads() and a for loop, you can build your unordered list.
Every time you load the page, it makes a server-side API call to Ollama, gets a list of large cities in Wisconsin, and displays them on the website. The list is never the same (because of hallucinations) but that is another issue.
Have any questions, comments, etc? Please feel free to drop a comment, below.
That installs Ollama as a dependency in package.json.
Basic CLI example
At this point, we can start writing code. When we used the web service earlier this week, we used the generate endpoint and provided model, prompt, and stream as parameters. We set the stream parameter to false so that it would return a single response object instead of a stream of objects. When using the javascript library, the stream parameter isn’t necessary because it returns a single response object by default. We still provide it with a model and a prompt, though.
If you run it from the terminal, the response will look familiar.
If you npm install express to install express, you can host a simple HTTP page at port 8080 and with the magic of JSON.parse() and a for loop, you can build your unordered list.
Every time you load the page, it makes a server-side API call to Ollama, gets a list of large cities in Wisconsin, and displays them on the website. The list is never the same (because of hallucinations) but that is another issue.
Have any questions, comments, etc? Please feel free to drop a comment, below.
My favorite client for MacOS is MindMac. You can buy it for under $30, it works with multiple models, servers, and server types, and it is easy to use.
If you want to look further into it, you can check it out at mindmac.app.
Android
My favorite client for Android is Amallo. It is $23 and like MindMac, it works with multiple models, servers, and server types. My only complaint would be that uploading a base64-encoded image to the model doesn’t seem to work well.
A few months ago, I started trying to figure out how I could use AI without depending on Google, OpenAI, or Microsoft’s continued existence. The risk that Google would kill a product like Gemini is almost 100%. At work, I was asked to figure out embedding and the tools I used in the post are invaluable for the on-the-go use of both a stock model and something that you tinkered with.
If you like that, you are going to love the next dozen posts that I have planned. 🙂
Yesterday, we played with Llama 3 using the Ollama CLI client (or REPL). Today I figured that we would play with it using the Ollama API. The Ollama API is documented on their Github repo. Ollama has a client that runs when you run ollama run llama3 and a service that can be accessed from something like MindMac, Amallo, or Enchanted. The service is what starts when you run ollama serve.
In our first Llama 3 post, we asked the model for “a comma-delimited list of cities in Wisconsin with a population over 100,000 people”. Using Postman and the completion API endpoint, you can ask the same thing.
You will notice the stream parameter is set to false in the body. If the value is false, the response will be returned as a single response object, rather than a stream of objects. If you are using the API with a web application, you will want to ask the model for the answer as JSON and you will probably want to provide an example of how you want the answer formatted.
Last week, Meta announced Llama 3. Thanks to Ollama, you can run it pretty easily. There are 8b and 70b variants available. There are also pre-trained or instruction-tuned variants available. I am not seeing it on the Hugging Face Leader Board yet but the bit that I have played around with it has been promising.
We have been looking at a lot of new things on the blog, lately (React, web components, etc). I figured that it was time to add was time to add Python to the list. Today’s post is going to go over the basics but you can expect a series of posts on the topic over the next few weeks. Python is a lot like Node. It is a general-purpose programming language that is not specialized for any specific problems. You can use it to build websites and software, automate tasks, and analyze data.
How to run a Python script
If you want to write Python, you will probably want to also run what you write. If you are developing on a Windows computer and don’t yet have Python installed, you can run the command python3 from PowerShell to install it from the Microsoft Store. If you don’t know if you have it installed, you can run python3 --version from the same place. If you are using MacOS, you can download the Python installer from the Python website and run it to install Python. If you are a MacOS user, Python3 is likely already there from previously using Homebrew.
With Python installed, you be able to open the Python shell and run simple commands.
Next, let’s install Jupyter. Jupyter Notebook is a popular way to write and run Python code, especially for data analysis, data science, and machine learning. Jupyter Notebooks are easy to use because they let you execute code and review the output quickly. From the MacOS terminal, you can install Jupyter by running brew install jupyterlab (assuming that you have Homebrew installed).
Each cell in the notebook is going to be numbered, you can get as complex with it as you want, you can add markdown into cells, and you can use “Save and Export Notebook As…” to share your work. It is kind of cool.
Python Identifiers and Literals
When you name a variable, function, or class in Python, it can’t start with a number, be a reserved word, or contain special characters. It is also case-sensitive. The function type(x) (where x is an identifier) returns the data type for the object passed into it. Let’s take a look at a few data types.
As you can see, you have a lot of options. Just remember that True and False values with booleans are case-sensitive.
Conditions and If Statements
Conditions are pretty easy with Python.
Equals: a == b
Not Equals: a != b
Less than: a < b
Less than or equal to: a <= b
Greater than: a > b
Greater than or equal to: a >= b
If statements are just if x > y: on the first line and what you want to conditionally run below it. Just remember to put an indentation after if x > y:.
We have talked about docker a few times in the past. Most recently, we talked about it in the context of running Ollama. For today’s post, I wanted to talk about how to turn your code into a docker container that you can run somewhere.
What is Docker
Docker provides the ability to package and run an application in a loosely isolated environment called a container. Docker containers can be deployed to just about any machine without any compatibility issues so your software stays system agnostic, making software simpler to use, less work to develop, and easier to maintain and deploy.
Once upon a time, a web application would be run on a physical piece of hardware that is running an operating system like Linux or Windows and then virtualization became a thing. Virtual machines access the hardware of a physical machine through a hypervisor. The host machine has an operating system (Ubuntu, Windows, MacOS, etc) and a hypervisor. Each virtual machine has an operating system of its own, binaries and libraries, and the actual web app. When using containers, the host machine has an operating system and a container engine but the containers only have binaries and libraries and the actual web app (no guest OS is necessary).
A dockerfile is needed to create an image and a container is the result of running an image. Today I am going to show how to go from a basic web app to a running docker container.
A Basic Node Example
If we are going to be dockerizing a web app, we need a web app to dockerize. In yesterday’s demo on how to pass an array as a property into a web component, we looked at three ways to turn an array into an unordered list. I figured that we could do the same with today’s demo.
In the above Node app, we are setting const items as being an array, using <a href="https://www.w3schools.com/nodejs/met_http_createserver.asp">createServer()</a> to create a new HTTP server, and then we are setting it to listen on port 8080. If you save the file locally as app.js, assuming that you have Node installed on your machine, you can run node app.js from the terminal to start the server.
You will notice that it also includes the line EXPOSE 8080, to expose port 8080 but as you will see below, it is more for documentation purposes than anything else.
Creating a Dockerignore
If you are familiar with git, you likely know what a .gitignore file is. A .dockerignore file does something similar. A .dockerignore is a configuration file that describes files and directories that you want to exclude when building a Docker image. Usually, you put the Dockerfile in the root directory of your project, but there may be many files in the root directory that are not related to the Docker image or that you do not want to include. .dockerignore is used to specify unwanted files and not include them in the Docker image.
Building a Docker Image
Now that you have what you are dockerizing, a Dockerfile, and a .dockerignore, you can simply build by running docker build . in the terminal.
If you want to aid in maintainability a little, you can add -t [image name] to the build command. When you run docker build -t node-app . in the terminal, it looks like this …
As I said above, an image becomes a container when you execute it. You can execute it by running docker run -d -p 8080:8080 6cced9894e8c where -d runs it as a daemon (a background process) and -p [port number]:[port number] tells the container what port to give it on the host machine. The 6cced9894e8c hash is the “Image ID” value from when I ran docker images above. If you tagged the image in the above step, you can use that value instead of the hash, though.
If you run docker ps after starting the container, you can verify that it is running. Go to http://localhost:8080/ and witness the splendor (now running in a docker container).
You will notice that the ArrayList class has an items property that is an array type. Lit won’t let you do something like <array-list items = ['Item 1', 'Item 2', 'Item 3']></array-list> but it is fine with you passing it in using javascript. That means that myList.items = ['Item 1', 'Item 2', 'Item 3']; does the job, fine.