- Removed deprecated features
create_editandget_usage_status
- Fixed a bug in
request_bodythat preventedqueryargument from being passed toHTTP.jl. - Updated completion model in the unit tests suite (
adaseries has been deprecated). - Added warnings to
create_editthat it's deprecated by OpenAI. Disabled tests forget_usage_statusandcreate_editfunctions, as they cannot be tested via API.
- Added OpenAI Assistants API
- disable
test/usage.jlintest/runtests.jlto close issue: #46
- bugfix: http_kwargs (ie, kwargs to be provided to HTTP.jl requests) are not being passed, because of a missing semicolon inside of request_body() call.
- bugfix: created a separate test file for chat completion: despite the naming similarity, the /v1/chat/completions has nothing to do with /v1/completions. If a test fails, we need to be able to differentiate between the two easily.
- added JET code quality test
- add attributions for contributors via authors field in Project.toml
- feature: get api usage via get_usage_status(), add test
- refactor: rewrite methods related to api
- bugfix: issue with HTTP.request kwargs propagation, enables the user to pass an arbitrarily named tuple to http_kwargs
- bugfix: error in chat streaming
- Refactor to remove Downloads.jl in favor of HTTP.jl
- feature: add chat streaming
- feature: introduce provider abstraction to support arbitrary OpenAI deployments
- feature: add image generation endpoint
- add create_edit example
- update README
- feature: add API for creating embeddings
- feature: add chat/completion
- bugfix: avoid issues with quoted strings in kwargs
- Replace HTTP/JSON with Downloads/JSON3
- removed Manifest.toml from tracking
- updated HTTP compat
- feature: added "create edit"
- added GitHub Actions workflow for running tests and parameterizing secrets
- adding completions
- engine endpoints deprecated in favor of models
- engine endpoints working