Releases: pinecone-io/locust-pinecone
Releases · pinecone-io/locust-pinecone
v0.2.0
New Features
- Add
--pinecone-dataset-limit
option to only use a subset of a Dataset (#44) - Add
--pinecone-throughput-per-user
option to limit the number of requests per second each user issues. This helps to simulate real-world workloads which don't simply issue requests as fast as they possibly can :) (#48 - Add
--pinecone-recall
option to report Recall metrics (instead of latency) (#50) - Disable destructive requests (delete) by default (#53)
Bugfixes
- Adjust Type column to align 'Pinecone gRPC' correctly (#49)
- Dataset.recall(): Fix div-by-zero if zero matches (#52)
Miscellaneous
- Migrate deps management from pip to poetry by @daverigby in #39
- Add pre-commit, configure to run poetry pre-commit hooks by @daverigby in #41
- Move Dataset into its own file by @daverigby in #43
- Dataset.load(): Additional logging by @daverigby in #46
- Migrate from deprecated Blob.download_as_string() by @daverigby in #47
- Skip loading documents if not needed by @daverigby in #51
- README.md: Document measuring recall by @daverigby in #54
- README.md: Add examples by @daverigby in #55
Full Changelog: v0.1.0...v0.2.0
v0.1.0
What's Changed
This is the first formal release of locust-pinecone, after recent changes to improve the functionality and reliability of load-testing Pinecone indexes. These include:
- Add support for connecting using the Pinecone Python Client, in addition to the existing locust-based HTTP client. Connection mode can be set via
--pinecone-mode=
. Default connection mode is changed tosdk+grpc
. - A set of integration tests have been created to ensure locust-pinecone works as expected and for future regression testing. These are triggered via GitHub actions on each PR, and include pod-based and Serverless variants.
- Add support for Pinecone public datasets. When specified via
--pinecone-dataset=
, a dataset is downloaded from cloud storage (if not previously downloaded), and can be used for initial index population, and/or for query generation. Invoke with--pinecone-dataset=list
(orlist-details
to print information on the set of available datasets. - Update various 3rd party dependancies to latest versions, to address stability and security issues.
Full Changelog: https://github.com/pinecone-io/locust-pinecone/commits/v0.1.0
v0.1.0-rc2
What's Changed
- Change default pinecone-mode from 'rest' to 'sdk+grpc' by @daverigby in #35
- Bump pinecone-client[grpc] from 3.0.3 to 3.1.0 by @dependabot in #36
- Update README.md for v0.1.0 by @daverigby in #38
- pyarrow.jemalloc_set_decay_ms(): Ignore if jemalloc not present by @daverigby in #37
Full Changelog: v0.1.0-rc1...v0.1.0-rc2
v0.1.0-rc1
What's Changed
This is the first formal release of locust-pinecone, after recent changes to improve the functionality and reliability of load-testing Pinecone indexes. These include:
- Add support for connecting using the Pinecone Python Client, in addition to the existing locust-based HTTP client. Connection mode can be set via
--pinecone-mode=
. Default connection mode is changed tosdk+grpc
. - A set of integration tests have been created to ensure locust-pinecone works as expected and for future regression testing. These are triggered via GitHub actions on each PR, and include pod-based and Serverless variants.
- Add support for Pinecone public datasets. When specified via
--pinecone-dataset=
, a dataset is downloaded from cloud storage (if not previously downloaded), and can be used for initial index population, and/or for query generation. Invoke with--pinecone-dataset=list
(orlist-details
to print information on the set of available datasets. - Update various 3rd party dependancies to latest versions, to address stability and security issues.
Full Changelog: https://github.com/pinecone-io/locust-pinecone/commits/v0.1.0-rc1