Releases: weaviate/weaviate-python-client
0.4.0-rc0 Add actions
This release adds actions.
This is a breaking change for reference methods. They require a sematic type now which states if the reference is from a thing or an action to a thing or an action.
Add pytnon 2.7 support
Add vector boosting
client.create_thing
has a new optional argument vector_weights
, that allows the boosting of single words. See the official release notes for more information.
Python 2.7
To support legacy code python 2.7 is now also supported. We encourage to use a python version >= 2.6 if possible though.
Add vector word boosting
client.create_thing
has a new optional argument vector_weights
, that allows the boosting of single words. See the official release notes for more information.
Allow querying the GraphQL endpoint
The client now offers a function query
to execute GraphQL queries on weaviate.
Support of vectorization masks
- Creation of schemas has changed to support the vectorization masks introduced in weaviate
0.21.11
. - Adding things or references in batch now returns a list containing the status for every thing/reference.
- The is_ready function now uses weaviates ready endpoint introduced in
0.21.2
- Authentication can now additionally be done using a user name and password
Contextual classification and filtered classifications
New Features:
- Contextual classification
- Filter on classifications
Breaking changes:
All logic connected to classification has been moved to Client.classification
.
Example:
Classify products according to their description into a category.
Only use products from the same country to classify.
# Create a configuration for a contextual classification
context_configuration = client.classification.get_contextual_config("Product", ["description"], ["hasCategory"])
# Create a filter (assumes that both the source and target class have a property inCountry)
country_filter = {
"operator": "Equal",
"path": ["inCountry"],
"valueString": "Germany"
}
# Add the filters
germany_config = client.classification.add_filter_to_config(
weaviate.SOURCE_WHERE_FILTER, country_filter, context_configuration)
germany_config = client.classification.add_filter_to_config(
weaviate.TARGET_WHERE_FILTER, data_filter, germany_config)
# Classify
classification_info = client.classification.start(germany_config)
while not client.classification.is_classification_complete(classification_info["id"]):
print("Classifying ...")
time.sleep(3.0)
# Change filter
country_filter["valueString"] = "Estonia"
# Add the filter
estonia_config = client.classification.add_filter_to_config(
weaviate.SOURCE_WHERE_FILTER, country_filter, context_configuration)
estonia_config = client.classification.add_filter_to_config(
weaviate.TARGET_WHERE_FILTER, data_filter, estonia_config)
# Classify
classification_info = client.classification.start(estonia_config)
while not client.classification.is_classification_complete(classification_info["id"]):
print("Classifying ...")
time.sleep(3.0)
0.1.3 Adding delete functionality
New features:
- Schemas specifying index can now be loaded.
- Things can now be deleted
0.1.3-rc-0 - Adding index to schema
The index keyword is now recognized by the client.
0.1.2 - Add contains_schema
Add function to know weather a schema was already loaded.
0.1.1 - Add patch
Adds the patch merge function to partially change a thing within weaviate.
Improves the execution of is_reachable
to also include testing for backend components running.