FAISS implements similarity search. Sentence Transformers encodes sentences into vectors FAISS can use. Llama can be configured to perform semantic searches in a FAISS vector store. By combining all of the above together we can build a local GenAI powered assistant capable of performing semantic queries to extract information from a local corpus of documents, let’s see how.
I’ve made a digital photo frame that displays random photo from my Google Photo library enriched with an Haiku generated by a LLaVa instance using the same photo as reference.
This experiment combines hsl.js and tensorflow.js to perform real time object detection from a browser. When the mouse hovers the canvas the entire stream is shown, with the detected object framed in a black box, otherwise only the parts of the stream corresponding to detected objects are displayed.
Next step is determining the roots of the characteristic polynomial of a matrix. Before working on extracting the roots using Newton’s method I wanted to be able to visualize the function.
I’m currently reading Deep Learning, from Ian Goodfellow, Yoshua Bengio and Aaron Courville. While reading I got interested in calculating the determinant of a square matrix, I know there are many existing libraries to do that but the algorithm looked simply enough to give it a try.
A friend of mine recently asked if I can create an animation out of this nice example; the first thing I did was wrap the render function around a requestAnimationFrame and add a random speed factor to all the vertices at every iteration.
Angular.js does an excellent job on preventing a form submission when invalid but only if the action attribute is not present, otherwise we have to implement a few tricks to achieve the same behaviour.
I spent part of the last weekend on this experiment that startet with a question: it is possible to convert a video to a stream of CSS properties ? Surprisingly the answer is yes. We can dump each video frame to a canvas and then create a special (and very long) multiple background property (one background for each video row) where each pixel is created using a 1px size gradient.
Recently I had to deal with a WordPress multisite instance, the task was moving the contents of one of the sites hosted on the network to a local copy, update them and the re-move them to the production server.
Lately I’ve been involved into a fascinating project called PepperTweet which consist in extract and visualize streams of pics from twitter starting from some search keywords.
