A checklist for WooCommerce

[eostis]

Being at the end of the integration process of Vespa with WooCommerce, I hereby share my discovery journey to those interested in investigating the subject.
It's a way to contribute back for all the time the Vespa team spent on my many questions 👍

I will continue to add more content as soon as new features are added.

You can ask for more details on specific points in the comments!

Pre-release

• https://www.wpsolr.com/forums/topic/vespa-integration/

Showstoppers:

• Dynamic summary not working with query on input field #29219
• Add exclusions to aggregation #26627
• bge-small-en ONNX component fails to load with "Opset 19 is under development" #28161
• HuggingFaceEmbedder model brings down the application APIs, making recovery impossible #28383

Features coded:

• Deploying an empty application on a docker instance
• Creating a schema: index, attribute, fieldset, basic ranking
• Management of several schemas per application
• Subclassing the default schema to let a user add/override settings (fields or ranking)
• Updating and redeploying a schema with new fields and the services.xml
• Deleting a schema and redeploying the services.xml
• CRUD on documents
• Batch indexing with HTTP2 multiplexing (Batch mode for document APIs #26579 )
• Keywords matching (userInput)
• Partial matching (N-Grams)
• Facets on numeric/strings and arrays: list, range, slider layouts (aggregations)
• Filters: AND, OR, !, range, and contains with annotations
• Sort on strings and numeric
• Pagination
• Autocompletion of documents, eventually grouped by document types
• Dynamic summary (Adding TRUE as an additional condition prevents dynamic summary generation #26693)
• Sort on arrays (Add sort on multivalued attributes #26681)
• Keywords demo deployed on Kubernetes
• Deploying ML models
• Semantic demo with intfloat/multilingual-e5-small deployed on Kubernetes
• Show Vespa deployment warnings
• Show/edit Vespa files from the dashboard
• Add embedding(s) choice to Vespa index settings
• Multiple ranking predefined in the schema, selectable from the UI: natural, BM25, vector, hybrid
• Multi-phase ranking

Features to be coded:

• Remove Vespa component unused from services.xml
• Deploying on Vespa cloud (multi tenant, multi application, ...) (elliptic encryption with https://github.com/phpseclib/phpseclib)
• Authentication
• Keyword autocompletion based on User shopping demo
• SSL on Docker

Features on standby:

• text-to-Image (CLIP) embedders ([FR] Support of more HuggingFace embedders for multimodality #28090 )
• Download application zip files (New API to download application packages #26487)
• Quickly use any HuggingFace Sentence transformer with easy configuration (Plug & Play download & configuration of HuggingFace sentence transformer models #26696 )
• External LLM APIs (Pre-built Integration to external embedding and generative APIs #26695 )
• Deploy prod on Vespa Cloud (API to deploy PROD applications on Vespa Cloud #26709)

[eostis]

Thanks for fixing: Dynamic summary (#26693)

[eostis]

Index setup with local Vespa docker

[eostis]

Dynamic fields added to schema during indexing

[eostis]

• onnx model generated with Pytorch 2.0

# https://huggingface.co/docs/optimum/installation
# https://blog.vespa.ai/simplify-search-with-multilingual-embeddings/
brew install python
pip3 install torch
pip3 install --upgrade-strategy eager install optimum[onnxruntime]

• onnx model deployed from a cloud object storage

[eostis]

Vespa Keywords demo on Kubernetes

[eostis]

Vespa multilingual vector search demo on Kubernetes

[eostis]

View/edit an application deployed file:

[eostis]

Automatic mapping and deployment of Vespa files

[eostis]

Choice of text embedder: