The article introduction

Jina 3 was released yesterday. What are the major updates to Jina 3 compared to the previous two major releases? What new experience will neural search bring? Take a look!

The first big news of 2022: Jina 3 is here!

Compared to the previous two major releases, Jina 3 offers significant improvements in cloud computing readiness and aggregation, making it easier for users to move projects from local to cloud production environments.

At the same time, the scalability and stability of Jina 3 have been significantly improved by reconfiguring the overall architecture and communication modules to make network communication between executors more stable and fully support asynchronous communication.

Jina 3 updates include:

1, Executors: No download, support cloud operation

Publish the Jina sandbox simultaneously. Instead of downloading executors from Jina Hub or using Docker, developers can try out cloud-running executors locally with Jinahub + Sandbox ://ExecutorName.

This allows developers to evaluate and find the right Executor more quickly, without downloading or hogging local computing.

from docarray import Document
from jina import Flow

flow = Flow().add(uses='jinahub+sandbox://Hello')

with flow:
docs = flow.post('/', inputs=Document(text='world'))
print(docs.texts)
Copy the code

Flows: Flows support K8S and Docker-compose

Provide Kubernetes and Docker-compose support, generate corresponding configuration files to help developers quickly deploy:

Kubernetes:

* flow.to_k8s_yaml(‘./k8s_flow’) is used to generate the Kubernetes deployment configuration file. With the generated configuration files, developers can quickly deploy Jina using Kubernetes.

Docker-compose:

* flow.to_docker_compose_yaml() is used to generate the docker-compose configuration file. By running Docker-compose up, developers can easily deploy Jina in docker-compose environment.

3,Everything can DocArray

Document and DocumentArray will be split into separate Python installation package DocArray. Developers can express and encapsulate unstructured data structures easily and efficiently in any project.

In addition, DocArray provides powerful visualization capabilities and support for multiple file storage schemes.

Before:

from jina import Document, DocumentArray, Flow
Copy the code

After:

from docarray import Document, DocumentArray
from jina import Flow
Copy the code

To learn more

For more details, please visit:

Release notes

Jina 3 Migration Guide

The related documents

README