⚠️ Since this is a release candidate (RC), we do NOT recommend using it in a production environment. Is something not working as expected? We welcome bug reports and feedback about new features.
Meilisearch v1.5 introduces improvements in indexing speed and the possibility of triggering snapshots on-demand, as it's already possible with dumps.
New features and improvements 🔥
Indexing speed improvements
For datasets containing a lot of words (and/or with a large variety of words), we improved the indexing speed between 5% and 20%, depending on your use cases and your documents. The more your documents contain a lot of words, the more the improvements will be visible.
The relevancy can be slightly different for queries containing 4 words or more. Let us know here if you have any issues or unexpected behavior.
For datasets containing only a few words per field, even if you have a huge number of fields, the improvement might not be visible. Don't worry, we keep working on the indexing speed in the next sprint to address other use cases.
Done by @ManyTheFish in #4131
Snapshots on-demand
We have introduced a new route, similar to the /dumps
one, creating snapshots on-demand.
To create a snapshot:
curl -X POST HTTP://localhost:7700/snapshots
This route will return a task.
The snapshot file will be created on your machine in the directory defined with --snapshot-dir
flag (default: snapshots/
)
Experimental feature: export Puffin Reports
Enabling this experimental feature will make Meilisearch automatically export .puffin
file reports. Analyzing those files can give great information on what's happening inside the engine and how the Meilisearch team can improve the indexing performances.
How to activate the feature:
curl \
-X PATCH 'http://localhost:7700/experimental-features/' \
-H 'Content-Type: application/json' \
--data-binary '{
"exportPuffinReports": true
}'
📣 To get more information about the feature or to give any feedback, please refer to this discussion.
Done by @Kerollmops in #4073.
Other improvements
- Make the experimental
/metrics
route activable via HTTP, and not only via CLI flag. Done by @braddotcoffee with the review of @vivek-26 in #4126.
⚠️ We don't recommend using CLI flag and HTTP calls at the same time to manage this feature.
Misc
- Dependencies upgrade for security reasons
- CIs and tests
- Add CI to trigger benchmarks in PR (#4102) @Kerollmops
- Improve test-suite.yml for CI failing when disabling tokenization (#4005) @harshau007
- Add more integrations to SDK CI (#4044) @curquiza
- Dependency issue is now created every 6 months (#4065) @curquiza
- Rename benchmark CI file to find it easily in the manifest list (#4125) @curquiza
- Update CI dependencies
- Misc
❤️ Thanks again to our external contributors: