Skip to content

Advanced Search for Awesome GEE Community Catalog

The Advanced Search feature of the Community Catalog leverages Vertex AI to enhance documentation search. This search is powered by both the catalog pages and Data JSON tabular structures. As more people use the feature, the indexed results improve over time. The feature benefits from grounded search, meaning that its summaries are always linked to the source material, ensuring transparency and trustworthiness.

This is made possible by RAG (Retrieval Augmented Generation), which not only improves search relevancy but also minimizes misinformation by grounding the output in reliable sources. Learn more about RAGs here in an earlier blog post to understand why this is so powerful and why grounded search can be effective. Currently, this search is in Beta, allowing us to test and refine it further with your participation. You can still use the embedded search within the Community Catalog for basic keyword or text-match searches.

Find the Advanced Search here.

Note

If the search doesn't work as expected, try disabling any widget blocker extensions or using an incognito window to troubleshoot.

catalog_search

Salient Features

  • Utilizes two data stores: the Community Catalog itself and the JSON data store. This dual approach enables quicker and more effective parsing across the data.
  • The search widget supports multiple languages. You can ask questions in your preferred language, and the results will be summarized in the same language.
  • As with all generative AI search engines, prompt formulation matters. Slight changes in phrasing can often yield better results.
  • Built on RAG (Retrieval Augmented Generation) technology, meaning summaries come with links to the source material. This boosts trust and minimizes inaccuracies.

Limitations

  • The final summary is generated from up to 10 search results. If multiple results meet the same criteria, not all may be included in the summary.
  • Summaries may evolve over time as the context window changes. Therefore, your summary might vary with each search.
  • Current output is part of a server-side controlled widget, which may affect customization or fine-tuning on the user end.

Changelog

  • Added a secondary data store (JSON) to speed up parsing of details like licenses and links.
  • Introduced a "Back to the Community Catalog" button for easy navigation back to the main catalog.
  • Implemented custom system instructions to improve output quality and relevance.
Logo Ask AI