Azure Cognitive Search添加多个Suggester报错的技术咨询

First, let's break down why you're seeing that error:

"An index cannot have more than one suggester with searchMode='analyzingInfixMatching'"

Azure Cognitive Search currently enforces a hard limit of one suggester using the analyzingInfixMatching mode per index. Even though the suggesters property is structured as an array (a clear nod to future extensibility), right now this is the only supported search mode for suggesters—and you can't have multiple instances of it in a single index.

Is the API reserved for future features? #

Absolutely. The array structure of suggesters was intentionally built to support additional search modes down the line. The Azure team designed the API with future functionality in mind, but as of now, no other modes are available. So for the time being, multiple suggesters in one index aren't a viable option.

How to handle your multilingual autocomplete needs #

Since you want to serve language-specific suggestions based on the user's selected language, here are the most practical workarounds:

• Option 1: Create language-specific indexes
  Split your data into separate indexes for each language (e.g., products-en, products-fr, products-de). Each index can have its own suggester pointing only to that language's fields. When a user selects a language, your app simply queries the corresponding index. This is the cleanest approach for strict multilingual isolation.

• Option 2: Use a single index with targeted query parameters
  Keep all language fields in one index, but assign language-specific analyzers to each field (e.g., en.microsoft for English fields, es.microsoft for Spanish). Create a single suggester that includes all language fields, then when making a suggestion request, use the searchFields parameter to restrict results to the current language's fields. For example:

  POST /indexes/products/docs/suggest?api-version=2023-11-01
  {
    "search": "app",
    "suggesterName": "mySuggester",
    "searchFields": "title_en,description_en"
  }
  

  This way, even though the suggester includes all fields, the query only pulls suggestions from the English fields when needed.

• Option 3: App-level filtering
  Build logic into your application layer that dynamically constructs the suggestion request based on the user's language selection. This is similar to Option 2, but the field targeting happens in your code before sending the request to Azure Search.

Whichever option you choose, make sure each language field uses an appropriate language analyzer—this ensures that suggestions are contextually relevant to the user's language (e.g., handling accented characters, stemming, or stop words correctly).

内容的提问来源于stack exchange，提问作者Jacco