Using multi-search to perform a federated search

Meilisearch allows you to make multiple search requests at the same time with the /multi-search endpoint. A federated search is a multi-search that returns results from multiple queries in a single list.

In this tutorial you will see how to create separate indexes containing different types of data from a CRM application. You will then perform a query searching all these indexes at the same time to obtain a single list of results.

• A running Meilisearch project
• A command-line console

Download the following datasets: crm-chats.json, crm-profiles.json, and crm-tickets.json containing data from a fictional CRM application.

Add the datasets to Meilisearch and create three separate indexes, profiles, chats, and tickets:

curl  -X POST 'http://localhost:7700/indexes/profiles'  -H 'Content-Type: application/json'  --data-binary @crm-profiles.json &&
curl  -X POST 'http://localhost:7700/indexes/chats'  -H 'Content-Type: application/json'  --data-binary @crm-chats.json &&
curl  -X POST 'http://localhost:7700/indexes/tickets'  -H 'Content-Type: application/json'  --data-binary @crm-tickets.json

Use the tasks endpoint to check the indexing status. Once Meilisearch successfully indexed all three datasets, you are ready to perform a federated search.

When you are looking for Natasha Nguyen's email address in your CRM application, you may not know whether you will find it in a chat log, among the existing customer profiles, or in a recent support ticket. In this situation, you can use federated search to search across all possible sources and receive a single list of results.

Use the /multi-search endpoint with the federation parameter to query the three indexes simultaneously:

curl \
  -X POST 'http://localhost:7700/multi-search' \
  -H 'Content-Type: application/json' \
  --data-binary '{
    "federation": {},
    "queries": [
      {
        "indexUid": "chats",
        "q": "natasha"
      },
      {
        "indexUid": "profiles",
        "q": "natasha"
      },
      {
        "indexUid": "tickets",
        "q": "natasha"
      }
    ]
  }'

Meilisearch should respond with a single list of search results:

{
  "hits": [
    {
      "id": 0,
      "client_name": "Natasha Nguyen",
      "message": "My email is natasha.nguyen@example.com",
      "time": 1727349362,
      "_federation": {
        "indexUid": "chats",
        "queriesPosition": 0
      }
    },
    …
  ],
  "processingTimeMs": 0,
  "limit": 20,
  "offset": 0,
  "estimatedTotalHits": 3,
  "semanticHitCount": 0
}

Since this is a CRM application, users have profiles with their preferred contact information. If you want to search for Riccardo Rotondo's preferred email, you can boost documents in the profiles index.

Use the weight property of the federation parameter to boost results coming from a specific query:

curl \
  -X POST 'http://localhost:7700/multi-search' \
  -H 'Content-Type: application/json' \
  --data-binary '{
    "federation": {},
    "queries": [
      {
        "indexUid": "chats",
        "q": "rotondo"
      },
      {
        "indexUid": "profiles",
        "q": "rotondo",
        "federationOptions": {
          "weight": 1.2 
        }
      },
      {
        "indexUid": "tickets",
        "q": "rotondo"
      }
    ]
  }'

This request will lead to results from the query targeting profile ranking higher than documents from other queries:

{
  "hits": [
  {
    "id": 1,
    "name": "Riccardo Rotondo",
    "email": "riccardo.rotondo@example.com",
      "_federation": {
        "indexUid": "profiles",
        "queriesPosition": 1
      }
    },
    …
  ],
  "processingTimeMs": 0,
  "limit": 20,
  "offset": 0,
  "estimatedTotalHits": 3,
  "semanticHitCount": 0
}

You have created three indexes, then performed a federated multi-index search to receive all results in a single list. You then used weight to boost results from the index most likely to contain the information you wanted.