0.39.0

• This release removes deprecated APIs (See details in #600). Use feed_iterable instead of feed_batch and update_batch.
• Adds support for using streaming mode #629
• Adds support for configuring alias for fields #633 (Thanks maxice8)

0.38.0

This release fixes multiple issues and adds new functionality to pyvespa.

• Several APIs around batch are now deprecated. See #600 and #599.
• Add a new feed_iterable API for stream feeding instead of batch
• Add support for global-phase ranking expression
• Add retry logic for improved stability for control API to Vespa Cloud (#608)
• Read errors from Vespa in case of error response instead of throwing exception (#599)
• Support **kwargs for Vespa query API so that users can use native parameters
• Reduce default number of connections and clarify documentation on using context managers

0.37.1

Bugfix release:

• Add macros to templates in setup.py: #551

0.37.0

The features added make it possible to run the multi-vector-indexing sample application in pyvespa:

• Add support for Component config: #548
• Add support for synthetic fields: #547

v0.36.0

• Allow to specify match-features in RankProfile: #525
• Add utility to create a vespa feed: #536

0.35.0

• Require minimum Python 3.8
• Support setting default stemming of Schema: #510
• Add support for first phase ranking: #512
• Support using key/cert pair generated by Vespa CLI: #513
• Implement deploy_from_disk for Vespa Cloud: #514

0.34.0

• Improved error handling: #498
• Use allowed_methods in urllib3.util.Retry: #501

0.33.0

• Support for rank profile inputs: #482

Version 0.32.0

• A local Docker deployment can now open the debug port (default 5005) into the container: #399
• Field additions:
  • The match block for fields is now supported: #403
  • Support for the weight and bolding fields: #406
  • Implement Struct Field, and allow Field to use it: #413
• Schema Document Summary is added: #408
• Add support for stemming, query-command, rank, and implement Struct for Documents: #416
• ApplicationConfiguration is added: #421
• rank-profile additions #424:
  • weight
  • type-hints for eq and repr
  • rank_type
  • rank-properties
• Support for validation-overrides.xml: #426
• Throw HTTPError on failures: #466

Version 0.31.0

• Remove parse_labeled_data. Used now only on the learntorank library for evaluation and data collection. (#384)
• Moved vespa.experimental.ranking to learntorank.ranking. (#386)
• Fixed CI pipeline by temporarily downgrading torch version from 1.13 to 1.12. (#387)
• Moved vespa.ml to learntorank.ml. (#388)
• Moved bits of ML code that were mixed in vespa.package to learntorank.ml. (#390)
• Removed unused serialization code from pyvespa. (#391)