News guides for WordPress & WooCommerce

Announcing new wpsolr.com series of self-signed certificate tutorials for OpenSearch Project, Elasticsearch, Apache Solr, Vespa.ai, Weaviate. Docker and Kubernetes. Because installing security on your self-hosted search engine should not be a daunting task.

For Nils Reimers (father of sentence transformers https://www.sbert.net/) the road is long: reranking, long texts, temporality (trends?), signals like popularity or recency, or even a new kind of LLMs are hot research and development topics!!! Me who was afraid to have gone around the subject. I’ll have to postpone my “Definitive AI retrieval encyclopedia” publication 😅 See it with your own eyes at wpsolr.com: WooCommerce + Weaviate + Cohere embeddings + Cohere reranking

Well, here is a good trick to get it for free 😎 1) Activate your free local Elasticsearch on your Cloudways WordPress application: https://lnkd.in/dibD-6UQ (the trick is to be hosted on Cloudways) 2) Install the free WPSOLR plugin: https://lnkd.in/dFueA7QY 3) Configure WPSOLR to create a local Elasticsearch index: https://lnkd.in/dX4jJDrM That’s it. You now get your first-class search experience, without breaking the bank. And it’s free, even with millions of visitors, posts, or products. And if Cloudways, in a near future, also pre-installs other search engines like Weaviate or Vespa.ai, you will be able to use their awesome semantic search capabilities for free too.

https://about.qwant.com/wp-content/uploads/2022/10/20221018_Qwant-x-Vespa-EN.pdf Qwant search is already powered by the Vespa.ai stack. And now, WooCommerce and WordPress owners can benefit from the same stack. With virtually no prior knowledge of Vespa.ai or vector search or deep learning. WPSOLR + WooCommerce + Vespa.ai demos: https://lnkd.in/dzucnPtZ https://lnkd.in/duBpAFgR #quant #vespasearch #woocommerce #wordpress #wpsolr #search

Small business WooCommerce shops finally have access to its world class vector and hybrid search. Here is the second preview of WooCommerce demo with Vespa.ai search, but this time with the bi-encoder embedder bi-encoder intfloat/multilingual-e5-small: https://demo-woocommerce-cloudways-2k-vespa-transformers.wpsolr.com/shop/ ! – Perfect for Vespa’s newbies, no requirement to know anything about Vespa. Vespa’s deployments are fully managed from WPSOLR’ dashboard, thanks to Vespa’s REST APis. No need of using the Vespa’s CLI anymore, but experts can continue to tune everything to their liking. – Changing settings (adding product attributes to index, changing ranking …) triggers a chain of events: create a Vespa session, download the index’s schema sd and services.xml, update both files (adding or removing an index in services.xml for instance), and upload both with the replace and deploy API. – This demo show facets,

In short: of course you can. Just ask in PM, and I’ll send you a link to schedule your personal Zoom presentation. If you ask yourself any of these questions, you are in for a treat: – What does a Kubernetes Weaviate / Vespa.ai cluster look like? – How to create and use tokens for Cohere, OpenAI, Hugging Face, Palm2 APIs? – Which hosting for Elasticsearch, Apache Solr should I use? – How to configure several indices for A/B tests? – My site is in French, Italian, Chinese, … what are my options and for which search engines? – Can I quickly test in my own WordPress / WooCommerce? – (and many more) Front-end demos: https://lnkd.in/dzucnPtZ Back-end demos (or your own site): ask in PM what you’d like to see…

At the time the article was written, WPSOLR was already delivering search from The Apache Software Foundation #Solr and Elastic #Elasticsearch. But since, we’ve integrated OpenSearch Project, Algolia, Google Cloud Retail search and even made a big leap towards AI search with Weaviate‘s vector database (and soon Vespa.ai). All those search engines can be accessed from hosted services, therefore no special installations are required on WP Engine’s hosting side. For WP Engine sites with lots of traffic, like WooCommerce shops during #blackfriday, using an external search engine compared to using MySQL leads to less CPU/memory usage and a far better scalability. Especially considering that caching is not an option for search.

Cloudways already does it for #Elasticsearch, with a free package that can be activated on every subscription. Doing the same for vector databases like Weaviate and Vespa.ai would drastically lower the technical barrier for small businesses that would like to participate to the AI search revolution. This would be a great solution for WordPress / WooCommerce sites with simple needs. For even more scalability, specialized hosting like Weaviate Cloud Services https://weaviate.io/developers/weaviate/installation/weaviate-cloud-services or Vespa Cloud https://cloud.vespa.ai/ are probably more suited. WPSOLR + Elasticsearch free package on Cloudways: https://www.wpsolr.com/guide/configuration-step-by-step-schematic/configure-your-indexes/create-a-cloudways-elasticsearch-index/ WPSOLR + (Weaviate on Kubernetes) + (Vespa on Kubernetes) demos: https://www.wpsolr.com/wpsolr-demos/

Answer: cross-language search In a globalised world, no one should ignore visitors from other countries or continents. But while automated pages translation is widely available with plugins like GTranslate https://wordpress.org/plugins/gtranslate/, how do we manage multilingual search? After all, your content is written into one language, yours, and eventually translated into one or two pore languages. Maximum. To search from a hundred countries/languages, you need something different, and this is where vector search and LLMs shine. Specially trained LLM models on tens of languages will transform your content and your visitors query into embeddings, which a vector database will use to match. Try our multilingual WooCommerce demo with WPSOLR + Weaviate + Cohere: https://demo-woocommerce-flatsome-cloudways-2k-cohere.wpsolr.com/shop/

1) Upload a model https://opensearch.org/docs/latest/ml-commons-plugin/ml-framework/#example-request 2) Embeddings with the model https://opensearch.org/docs/latest/search-plugins/neural-search/#example-request 3) Hybrid semantic search with filters and aggregations. WPSOLR + WooCommerce + #OpenSearch (vectors soon?): https://wpsolr.com

Multivalued attributes are fundamental to e-commerce: think about product variations with several prices per combination of sizes and colours for instance. Being able to sort, filter, and aggregate (faceting) on product attributes with keywords, vectors or hybrid mode is not only important but mandatory. You can monitor our progress on this roadmap: https://lnkd.in/eA6hCN3M WPSOLR + Vespa (soon!): https://wpsolr.com #wpsolr #vespasearch #woocommerce

My 5 cents feeling from https://lnkd.in/ekvimz4k: 1. 😀 E-Commerce search is all about faceting. And vector search is no exception. “A rich framework of filtering and faceting capabilities that Elasticsearch developers rely on, available for vector search.” 2. 😀 Most small businesses cannot afford to spend time and money on building external integrations to produce embeddings. “Capture meaning, context, and associations of data in dense vectors, with flexibility of picking embedding models. Run machine learning inference as you index data.” 3. 😥 The catch is that the embeddings generation is not free !! You have to buy a license or pay for a cloud hosting subscription “Generate embeddings: full support (paid)” The licensing for embeddings is what prevented WPSOLR from integrating Elasticsearch vector search: it

2 years ago, who could imagine a free WordPress vector search? Powered by free WPSOLR, free Weaviate, free bi-encoders similarity and free cross-encoders reranking 🙂 (and free WordPress!!) Small businesses can enjoy a search stack that many large corporations can only dream of 😂 Download WPSOLR free at wordpress.org/plugins: https://lnkd.in/dFueA7QY #wpsolr #wordpress #plugin #search #vectorsearch #biencoder #crossencoder #reranking

Demo time ! – Vespa deployment is fully managed from WPSOLR’ dashboard, thanks to Vespa’s REST APis. No need of using the Vespa’s cli anymore, but you still can of course. – Changing settings (adding product attributes to index, changing ranking …) triggers a chain of events: create a Vespa session, download the index’s schema sd and services.xml, update both files (adding or removing an index in services.xml for instance), and upload both with the replace and deploy API. – This demo show facets, filters, and suggestions, with keyword search. Sorting is not operational, yet. – Vespa.ai is installed on a Google Cloud Kubernetes cluster. – More demos with Vespa.ai‘s vector/hybrid/reranked search will follow soon… WPSOLR + Vespa.ai (soon): https://wpsolr.com #wpsolr #vespaai #woocommerce #wordpress #search

=> WPSOLR 23.3 is now offering Weaviate‘s two new reranking modules: Cohere‘s reranker API, but also cross-encoder transformers. => Weaviate’s retrieval first phase is based on highly efficient bi-encoder sentence transformers. => Reranking is more CPU consuming, especially with cross-encoders, hence is used on a second phase, and only on the first phase retrieved results. WooCommerce demo comparing a plain bi-encoder retrieval with a cross-encoder reranker: https://lnkd.in/eBnnd4wj WPSOLR 23.3: https://lnkd.in/ej2mmbRc WPSOLR + Weaviate rerankers: https://lnkd.in/emg3QUsb #reranking #cohere #weaviate #wpsolr #crossencoder #woocommerce

You can try and compare results from 2 suggestion boxes, one with reranking from a sentence transformer cross-encoder, the other without. (Search is powered without reranking) Demo: https://lnkd.in/e8TNX_8q Weaviate search rerank: https://lnkd.in/e9V8kw_y WPSOLR + WooCommerce + Weaviate + reranking(soon): https://wpsolr.com #ranking #search #woocommerce #weaviate #cohere #crossencoder

We had to upgrade our production Weaviate, used for our WooCommerce demos, several times to add new features : sentence transformers, OpenAI, Cohere, Google Palm2, and last but not least Rerank. At first, we did what most of us do in that case: clone the cluster, upgrade it, test it, switch to it, and delete the old cluster. Quite a challenge. But now we simply upgrade the production cluster, configured with a stateful replica of 2. And incredibly, during the upgrade, both querying and indexing suffer no interruption. 😎 Nice!! WPSOLR + Weaviate: https://wpsolr.com

Cohere‘s reranking API is indeed a very neat way of improving search quality without downgrading performance. I’m wondering if it can also improve Cohere‘s own multi-language embedding-based vector search? There is nowadays many reasons to integrate reranking to existing symbolic and semantic search results. — Reranking for recommenders like Metarank Labs — User signals are not the only way to feed recommender systems. For small businesses like WooCommerce or WordPress, semantic reranking is a great alternative to the lack of historical data. — Reranking for performance boost — SolR, Elasticsearch, OpenSearch Project, Algolia, or Weaviate are super efficient at retrieving documents, thanks to fast BM25 (statistics) or bi-encoder sentence transformer models (semantic). More efficient algorithms exist, like cross-encoder models, at a performance cost that can only be acceptable when used after the retrieval phase, as

— Sorting arrays — An important feature of E-Commerce search is the ability to sort on arrays of strings, or arrays of numbers: product variations can hold several colours, sizes, or prices. — Vespas’s new release — Today, Vespa.ai released the multi-valued sort feature that WPSOLR will use to power attributes and price sorting. Faceting and filtering on multi-values has already been integrated to WPSOLR. WPSOLR + https://vespa.ai (soon?): https://wpsolr.com Integration roadmap: https://lnkd.in/eA6hCN3M #wpsolr #vespasearch #woocommerce #wordpress

The image shows two sets of results from the same WooCommerce demo contents. — Image search https://lnkd.in/ez_zrH95 — Image search is powered by Weaviate‘s CLIP embedding module https://lnkd.in/eEhCqiem, with a 50% weight on images. A visitor will immediately notice the coherence of images with the query: they are all yellow. This is very important, as the first impression is visual. Only during a second phase the visitor will check the result titles and description. — Text search https://lnkd.in/eHjBEJy6 — Text search is powered by Weaviate‘s OpenAI embeddings module https://lnkd.in/erxxhdqS. A visitor will immediately notice the incoherence of images with the query: they are not all yellow. This is very important, as the first impression is visual. There is a good chance that the visitor leaves without getting a second chance to evaluate titles and descriptions. WPSOLR

Very nice paper https://lnkd.in/eyfhuyAi from Metarank Labs Similar to Vespa papers, Metarank papers are based on hard benchmark figures, which is the best/only way to understand facts. — Summary — > BM25 was a tough contender 2 years ago, when the BEIR benchmark and SBERT models reigned supreme. (Probably explaining why Elasticsearch is now fully commited to vector search) > Nowadays, new MTEB benchmark https://lnkd.in/exGBk3ND is dominared by new models like Microsoft E5. The trend is that keeping your current vector search and waiting for model embeddings improvement is a winning strategy ! WPSOLR + WooCommerce + Weaviate + SBERT embeddings (waiting for E5?): https://wpsolr.com #wpsolr #bert #sbert #wordpress #woocommerce #huggingface #weaviate #metarank

** Why Metarank ? ** – Their great presentation at Berlin Buzzwords: https://lnkd.in/eqifGvjH – Recommendations, personalization, and re-ranking – Modern with ML models like cross-encoders – Plenty of algorithms ready-to-use – Open-source https://www.metarank.ai/ – Well documented – Can work with or without user signals – Self-hosted, which is great for privacy – Production ready on Kubernetes with Redis – REST API and real-time integrations to send data https://lnkd.in/e4eHCMnx – Automatic model retraining – Already works with Opensearch https://lnkd.in/ebmuDHFb ** Benefits for WPSOLR users ** We can make Metarank work with all the search engine already integrated in WPSOLR. Metarank could therefore, with a single setup, re-rank/personalize search results for Elasticsearch, Opensearch, Apache Solr, Google Retail, and Weaviate. WPSOLR: https://wpsolr.com Metarank: https://www.metarank.ai/ Metarank live demo: https://demo.metarank.ai/ #wpsolr #metarank #recommendationsystems #reranking #elasticsearch #opensearch #apachesolr #weaviate

Personal playlist of Youtube tutorials for Sentence transformers from my journey to add automated fine-tuning to WooCommerce with Weaviate and Metarank Labs: https://lnkd.in/eRHshF_E Sentence transformers produce new embeddings from BERT embeddings trained to match 2 text inputs. They are the foundation of text vector similarity search. Sentence transformers site: https://lnkd.in/enET_iSj Hugging Face Models: https://lnkd.in/e3KG8hGj I’ll add more about training stacks like Anyscale and MosaicML later. Stay tuned: https://wpsolr.com #wpsolr #sentencetransformer #huggingface #weaviate #metarank

– Classic workflow – Usually, Chatbots are build from intents and actions. An agent is trained to match user inputs to a predefined intent. Then the intent triggers an action, for instance by calling an external Webhook. – An example – 1) User input: “Hi, I’m xyz@abcd.com, can you tell me the status of my order XYZ? It looks like stuck.” 2) Bot: “Of course. Your order XYZ is waiting for the bank’s approval. I’ve asked a sales rep to call you back.” – Explanation – In this example, the input is matched to the intent “order status”. The parameters {client_email} and {order_number} are extracted and used as input to a call to WooCommerce orders API to get the status and staff comments. – New kid

Recently, Carrefour announced a beta release of its own OpenAI generative ChatBot https://lnkd.in/dHVDUNSr It must be worth it, is it not? 🙂 WPSOLR already powers WooCommerce with vector search, which means all the data stack of the ChatBot is available. Does that mean a ChatBot is on our roadmap? We’ll see…. WPSOLR + WooCommerce + ChatBot(?): https://wpsolr.com #wpsolr #weaviate #vespasearch #chatbot #woocommerce

Today, we were contacted by a client to try a pre-release of WPSOLR with Vespa.ai. The client is already using WPSOLR with Elasticsearch for the backend orders search, and Weaviate for the 200K front-end products search. The client wants to rank search results with a custom trained XGBoost model, and plans also to build recommendations. He’d like to use a single solution for search, ML ranking, and recommendations. Rather than build a custom solution from scratch, using WPSOLR + Vespa makes sense. WPSOLR will hide lots of technical details required when deploying Vespa application by hand, and let the client concentrate on schemas and ML ranking. Stay tuned for more informations on the POC… You can follow the POC progress on our forums: https://lnkd.in/dnmhb-zG WPSOLR: https://wpsolr.com #wpsolr #woocommerce #vespaengine #vectorsearch #xgboost #weaviate

Introduction In today’s digital age, e-commerce has become a thriving industry, and online retailers are constantly striving to enhance their customers’ shopping experience. One crucial aspect of any successful online store is the ability to provide efficient and accurate search functionality. For WooCommerce-powered stores, Elasticsearch is a powerful search engine that can significantly improve search performance and relevance. In this post, we will explore various Elasticsearch query types that can be used for advanced WooCommerce searches, along with code examples using the PHP client. Understanding Elasticsearch Queries Elasticsearch offers a wide range of query types to retrieve specific data from an index. These queries are designed to handle different use cases and provide flexibility in defining search criteria. Let’s explore some of the commonly used

Introduction When it comes to running an e-commerce website on the WooCommerce platform, search engine optimization (SEO) plays a vital role in attracting organic traffic and driving sales. One powerful tool that can be leveraged for WooCommerce SEO optimization is Elasticsearch. Elasticsearch is a highly scalable and flexible search engine that can be integrated into your WooCommerce store to improve search functionality and enhance SEO performance. In this post, we will explore how Elasticsearch can be utilized to optimize SEO for your WooCommerce store. We will also include some code snippets using the PHP client for Elasticsearch, demonstrating how you can implement these optimizations. Leveraging Elasticsearch for WooCommerce SEO Optimization Elasticsearch provides several features that can significantly enhance the search experience and improve SEO performance

Introduction Elasticsearch is a powerful and versatile search and analytics engine that can be utilized for a wide range of applications. One such application is monitoring and alerting for WooCommerce stores. WooCommerce is a popular e-commerce platform built on top of WordPress, and Elasticsearch can enhance its capabilities by providing real-time monitoring and alerting features. In this post, we will explore how to leverage Elasticsearch for WooCommerce store monitoring and alerts, along with some example code using the PHP Elasticsearch client. Setting Up Elasticsearch Before we dive into using Elasticsearch for monitoring and alerts, we need to set up an Elasticsearch cluster. You can either set up your own cluster or use a managed Elasticsearch service provided by a cloud provider like AWS or Elasticsearch

Introduction Conversion rates are a vital aspect of any online business, and WooCommerce, being a popular e-commerce platform, provides numerous tools and features to help improve those rates. One powerful tool that can significantly enhance the search functionality and, consequently, the conversion rates in WooCommerce is Elasticsearch. By integrating Elasticsearch into your WooCommerce store, you can offer your customers a fast and accurate search experience, leading to higher engagement and increased conversions. In this post, we will explore how Elasticsearch can be integrated into WooCommerce to optimize the search functionality. We will also provide some code examples using the PHP client for Elasticsearch to help you get started with the implementation. Chapter 1: Elasticsearch Integration To begin with, let’s discuss how to integrate Elasticsearch into

Introduction Integrating Elasticsearch with WooCommerce can greatly enhance the functionality and performance of your online store. Elasticsearch is a powerful search and analytics engine that allows for real-time indexing and searching of large amounts of data. By integrating Elasticsearch with WooCommerce, you can provide your customers with faster and more accurate search results, as well as real-time product updates. In this post, we will explore how to integrate Elasticsearch with WooCommerce and achieve real-time product updates. We will also discuss how WPSOLR, a popular WordPress plugin, can assist in simplifying the integration process. Integrating Elasticsearch with WooCommerce To integrate Elasticsearch with WooCommerce, we need to perform the following steps: 1. Install and configure Elasticsearch: Start by installing Elasticsearch on your server or using a cloud-based