🎉 Jina 1.3
⬆️ Major Features and Improvements
Query while Indexing
This is one of the most requested features from our community, so we are psyched to release it finally! With Jina 1.3, you are can now make use of our dump-reload feature. This allows your users to query your system while you simultaneously Index new data continuously. In addition, you can also now make use of new locations to store your indexed data, such as a Postgres database, which gives you better control over the scalability and reliability of your datastore. You can try this feature out in a prebuilt example, or check out our documentation for a deeper understanding. Related pull requests: #2224 #2331 #2307 #2260
Replicas to improve throughput
It seems that with every release note I write, we talk about how we've improved speed in Jina! In this month's speed section, we have added replicas. 😵. When you activate a Pod replica in Jina, you create identical copies of that Pod that can simultaneously handle data, thus increasing throughput. For example, four replicas of an encoder Pod could increase your encoding speed by up to x4! This is still a prototype feature, and we suggest using it with caution. If you discover an issue, please open a Github issue! For more details about this feature, check out our documentation.
Learning to rank
If you've been closely following our release notes, you may have noticed we mentioned this feature before... anyway, we now do it better!
Learning to rank is a machine learning technique that allows you to optimize your ranking on a metric, such as click-through rates. For example, if you have a product catalog search, a user queries for shoes. You can now continuously show them the most purchased shoe, likely increasing your sales rate. This feature is now even easier to implement in Jina; you can start by checking our documentation.
⚠️ Breaking Changes
📗 Documentation
- Improve the documentation formats for
zed.py#2358 - Fix broken links in
README.md. #2376 #2435 @akanz1
🐞 Bug Fixes and Other Changes
Flow
- Fix the blocking bug when sending empty requests via websocket #2368 @jacobowitz
- Refactor arguments handling logic for replicas #2335
- Fix the deadlocks when rolling update Pods. #2344 #2385
Drivers
- Refactor the batching support for Indexers #2356
Tests
- Add mock for the external calls. #1958
- Move integration tests of
jina hello fashionto system tests. #2379 - Fix the test for compound pod sockets #2347
HubIO
- Fix the dependency issue when using
git+https://for the hub images. #2365
Others
- Add variables for deploying cloud.jina.ai #2372
- Implement the offline version of learning-to-rank. #2366
🙏 Thanks to our Contributors
This release contains contributions from @maximilianwerk @deepankarm @davidbp @JoanFM @hanxiao @jacobowitz @mkhilai @slettner @akanz1 @bwanglzu @niuzs-alan @Kelton8Z @nan-wang @FionnD @lusloher @aga11313 @alexcg1 @pswu11 @BastinJafari @imsergiy @Roshanjossey
🙏 Thanks to our Community
And thanks to all of you out there as well! Without you, Jina couldn't do what we do. Your support means a lot to us.
🤝 Work with Jina
Want to work with Jina full-time? Check out our openings on our website.