Meilisearch v1.7 release notes - Meilisearch Documentation

New Features

New OpenAI embedding models

When configuring OpenAI embedders, you can now specify two new models:

text-embedding-3-small with a default dimension of 1536
text-embedding-3-large with a default dimension of 3072

These new models are cheaper and improve search result relevancy.

Custom OpenAI model dimensions

You can configure dimensions for sources using the new OpenAI models: text-embedding-3-small and text-embedding-3-large. Dimensions must be greater than 0 and smaller than the model size:

{
  "embedders": {
    "new_model": {
      "source": "openAi",
      "model": "text-embedding-3-large",
      "dimensions": 512
    },
    "legacy_model": {
      "source": "openAi",
      "model": "text-embedding-ada-002"
    }
  }
}

You cannot customize dimensions for older OpenAI models such as text-embedding-ada-002. Setting dimensions to any value except the default size of these models will result in an error.

GPU support for Hugging Face embeddings

Activate CUDA to use Nvidia GPUs when computing Hugging Face embeddings. This can significantly improve embedding generation speeds. To enable GPU support through CUDA:

Install CUDA dependencies
Clone and compile Meilisearch with the cuda feature: cargo build --release --package meilisearch --features cuda
Launch your freshly compiled Meilisearch binary
Activate vector search
Add a Hugging Face embedder

Stabilized `showRankingScoreDetails`

The showRankingScoreDetails search parameter is now a stable feature. Use it with the /search endpoint to view detailed scores per ranking rule for each returned document:

curl \
 -X POST 'http://localhost:7700/indexes/movies/search' \
 -H 'Content-Type: application/json' \
 --data-binary '{"q": "Batman Returns", "showRankingScoreDetails": true}'

When showRankingScoreDetails is set to true, returned documents include a _rankingScoreDetails field with detailed scoring information for each ranking rule.

Experimental JSON log output

Configure Meilisearch to output logs in JSON format by passing json to the --experimental-logs-mode command-line option:

./meilisearch --experimental-logs-mode json

The --experimental-logs-mode option accepts two values:

human: default human-readable output
json: JSON structured logs

Experimental `/logs/stream` and `/logs/stderr` routes

Two new experimental API routes allow you to manage log output: Activate the routes using the /experimental-features endpoint:

curl \
 -X PATCH 'http://localhost:7700/experimental-features/' \
 -H 'Content-Type: application/json' \
 --data-binary '{"logsRoute": true}'

/logs/stream - Stream logs in real-time:

curl \
 -X POST http://localhost:7700/logs/stream \
 -H 'Content-Type: application/json' \
 --data-binary '{"mode": "human", "target": "actix=off,debug"}'

Parameters:

target: Defines log level and which part of the engine to apply it to. Format: code_part=log_level. Valid log levels: trace, debug, info, warn, error, or off
mode: Accepts fmt (basic) or profile (verbose trace)

Stop streaming with:

curl -X DELETE http://localhost:7700/logs/stream

You may only have one listener at a time. /logs/stderr - Configure default log output:

curl \
 -X POST http://localhost:7700/logs/stderr \
 -H 'Content-Type: application/json' \
 --data-binary '{"target": "debug"}'

Parameters:

target: Defines log level and which part of the engine to apply it to. Format: code_part=log_level. Valid log levels: trace, debug, info, warn, error, or off

Experimental cluster mode

New experimental feature to change the behavior of Meilisearch to run in a cluster by externalizing the task queue.

Improvements

Improved indexing speed and reduced memory usage

Auto-batch task deletion reduces indexing time
Hybrid search experimental feature indexing is now more than 10 times faster
Capped the maximum memory of grenade sorters to reduce memory usage
Multiple technical improvements to indexing pipeline
Enhanced facet incremental indexing
Improved threshold triggering incremental indexing

Improved logging

Log messages now follow a new pattern:

2024-02-06T14:54:11Z INFO actix_server::builder: 200: starting 10 workers

This replaces the previous format:

[2024-02-06T14:54:11Z INFO actix_server::builder] starting 10 workers

Multiple language support improvements

Expanded support for multiple languages, including improved Vietnamese normalization (Ð and Đ are now normalized to d). Updated to Charabia v0.8.7.

Additional improvements

Added content type to webhook requests
Skip reindexing when modifying unknown faceted fields
Added timeout to webhook requests
Enhanced Prometheus experimental feature with job variable in Grafana dashboard

Other

Breaking changes - Log output format

Log messages now follow a different pattern. If you have automated tasks based on log output parsing, you may need to update them to work with the new format. Find more information on GitHub

​New Features

​New OpenAI embedding models

​Custom OpenAI model dimensions

​GPU support for Hugging Face embeddings

​Stabilized showRankingScoreDetails

​Experimental JSON log output

​Experimental /logs/stream and /logs/stderr routes

​Experimental cluster mode

​Improvements

​Improved indexing speed and reduced memory usage

​Improved logging

​Multiple language support improvements

​Additional improvements

​Other

​Breaking changes - Log output format