Loading Insight...
Insights
Insights
4-min read • 11 Mar 2026
AI
Banking
From Keywords to Context: Improving Strapi with AI-Powered Semantic Search
- Out-of-the-box search in Strapi relies on exact keyword matching, which can produce inaccurate results when searching for content.
- Semantic search addresses the limitation by using embedding models to understand the intent and meaning behind a query.
- Implementing semantic search in Strapi enables efficient and accurate discovery of expert-insights articles as part of the Synpulse and Synpulse8 website revamps.
Written by Jason Tsang and Anthony Chan
Imagine if Google only returned results that exactly matched your search keywords. A simple query like “Best ramen shops in Hong Kong” would return many irrelevant results, such as “ramen recipes,” “clothing shops,” or generic pages about Hong Kong.
Semantic search[1] overcomes this limitation by capturing the intent and meaning behind a query, ranking results based on relevance rather than exact wording. This leads to more accurate results and a smoother, more satisfying search experience.
During the Synpulse website revamp, we wanted to add search functionality for our expert insights articles using Strapi as the content management system. But Strapi’s out-of-the-box keyword search often failed to return the most relevant articles. To fix this, we developed a Strapi plugin that brings semantic search to the platform.
How semantic search works
At the core of semantic search is the embedding model. In our implementation, we use a vector-search approach, converting each piece of text into a numerical vector that captures its meaning in high-dimensional space.
Texts with similar meanings are represented by vectors pointing in roughly the same direction. By comparing the angle between two vectors using cosine similarity, we can determine how closely the underlying meanings of two texts align.
Configuring the plugin in Strapi is simple. You only need credentials for your AI provider and a list of content types and fields you want to make searchable. The plugin is model-agnostic and can work with any embedding model provider that supports an OpenAI-compatible API format, including OpenAI and OpenRouter.
In addition, the plugin supports both manual embedding generation and automatic embedding generation whenever content is added, removed, or updated.
Smarter search, better user experience
Our semantic search plugin and its features have significantly enhanced the search experience on our Synpulse and Synpulse8 websites. Articles that were previously missed by keyword search now appear in semantic search results, enabling users to quickly find content relevant to their interests.
The plugin’s flexibility, which allows easy switching between different embedding models and providers, also makes it straightforward to test and refine search accuracy. This adaptability ensures continuous improvement in the quality of search results.
Recognising the value that semantic search brings to organisations and projects using Strapi, we decided to open-source the plugin and publish it to the npm registry. The source code is available in our GitHub repository, giving developers a ready-to-use solution to enhance search functionality on their own sites.
Transform how you search
Semantic search is a simple yet effective way to improve the accuracy of text-based search. Our plugin offers a flexible, configurable way to add semantic search to Strapi, delivering faster, more accurate results, as successfully demonstrated on our Synpulse and Synpulse8 websites.
Contact us to learn how our solutions can streamline your workflows and help transform your business!
What is semantic search? (Google, January 2026).
References:
Our Experts In This Topic
OUR EXPERT
Jason Tsang
Senior Analyst, Software Engineer
Jason Tsang is a Senior Frontend Developer with experience in building web applications in both enterprise and startup environments. Specializing in React, he is dedicated to delivering high quality designs and engaging user experiences that enhance usability.
OUR EXPERT
Jieke Pan
Chief Technology Officer
Jieke, CTO at Synpulse8, leads the global technology vision and engineering strategy. With over 20 years of experience in digital transformation and enterprise architecture, he drives the PULSE8 platform's evolution, delivering scalable, AI-enabled solutions that accelerate innovation and create transformative, client-centric digital experiences.
OUR EXPERT
Anthony Chan
Director, Software Engineer
He leads engineering teams to deliver bespoke software solutions, core banking systems, fintech integrations, and digital platforms that empower banks, and wealth managers.
He oversees architecture, agile delivery, and ecosystem partnerships to create secure, scalable, regulation-compliant outcomes.
Related Contents
Loading insights...
4-min read • 11 Mar 2026
AI
Banking
From Keywords to Context: Improving Strapi with AI-Powered Semantic Search
- Out-of-the-box search in Strapi relies on exact keyword matching, which can produce inaccurate results when searching for content.
- Semantic search addresses the limitation by using embedding models to understand the intent and meaning behind a query.
- Implementing semantic search in Strapi enables efficient and accurate discovery of expert-insights articles as part of the Synpulse and Synpulse8 website revamps.
Written by Jason Tsang and Anthony Chan
Imagine if Google only returned results that exactly matched your search keywords. A simple query like “Best ramen shops in Hong Kong” would return many irrelevant results, such as “ramen recipes,” “clothing shops,” or generic pages about Hong Kong.
Semantic search[1] overcomes this limitation by capturing the intent and meaning behind a query, ranking results based on relevance rather than exact wording. This leads to more accurate results and a smoother, more satisfying search experience.
During the Synpulse website revamp, we wanted to add search functionality for our expert insights articles using Strapi as the content management system. But Strapi’s out-of-the-box keyword search often failed to return the most relevant articles. To fix this, we developed a Strapi plugin that brings semantic search to the platform.
How semantic search works
At the core of semantic search is the embedding model. In our implementation, we use a vector-search approach, converting each piece of text into a numerical vector that captures its meaning in high-dimensional space.
Texts with similar meanings are represented by vectors pointing in roughly the same direction. By comparing the angle between two vectors using cosine similarity, we can determine how closely the underlying meanings of two texts align.
Configuring the plugin in Strapi is simple. You only need credentials for your AI provider and a list of content types and fields you want to make searchable. The plugin is model-agnostic and can work with any embedding model provider that supports an OpenAI-compatible API format, including OpenAI and OpenRouter.
In addition, the plugin supports both manual embedding generation and automatic embedding generation whenever content is added, removed, or updated.
Smarter search, better user experience
Our semantic search plugin and its features have significantly enhanced the search experience on our Synpulse and Synpulse8 websites. Articles that were previously missed by keyword search now appear in semantic search results, enabling users to quickly find content relevant to their interests.
The plugin’s flexibility, which allows easy switching between different embedding models and providers, also makes it straightforward to test and refine search accuracy. This adaptability ensures continuous improvement in the quality of search results.
Recognising the value that semantic search brings to organisations and projects using Strapi, we decided to open-source the plugin and publish it to the npm registry. The source code is available in our GitHub repository, giving developers a ready-to-use solution to enhance search functionality on their own sites.
Transform how you search
Semantic search is a simple yet effective way to improve the accuracy of text-based search. Our plugin offers a flexible, configurable way to add semantic search to Strapi, delivering faster, more accurate results, as successfully demonstrated on our Synpulse and Synpulse8 websites.
Contact us to learn how our solutions can streamline your workflows and help transform your business!
What is semantic search? (Google, January 2026).
References:
Our Experts In This Topic
OUR EXPERTJason Tsang
Senior Analyst, Software Engineer
Jason Tsang is a Senior Frontend Developer with experience in building web applications in both enterprise and startup environments. Specializing in React, he is dedicated to delivering high quality designs and engaging user experiences that enhance usability.
OUR EXPERTJieke Pan
Chief Technology Officer
Jieke, CTO at Synpulse8, leads the global technology vision and engineering strategy. With over 20 years of experience in digital transformation and enterprise architecture, he drives the PULSE8 platform's evolution, delivering scalable, AI-enabled solutions that accelerate innovation and create transformative, client-centric digital experiences.
OUR EXPERTAnthony Chan
Director, Software Engineer
He leads engineering teams to deliver bespoke software solutions, core banking systems, fintech integrations, and digital platforms that empower banks, and wealth managers.
He oversees architecture, agile delivery, and ecosystem partnerships to create secure, scalable, regulation-compliant outcomes.