relax pinning on importlib_metadata, typing_extensions #363
Conversation
|
Thanks for submitting your first pull request! You are awesome! 🤗 |
|
@minrk |
|
@minrk |
== pinning is almost always to be avoided in packages, and tight pinning on backports like importlib_metadata is extremely unlikely to be desirable
|
Congrats on your first merged pull request in this project! 🎉 |
== pinning is almost always to be avoided in packages, and tight pinning on backports like importlib_metadata is extremely unlikely to be desirable
* Update README, docs (#347) * Copy edits for brevity, adds large screenshot * Adds screenshot, adds principles to contributor docs * Uses screenshot class for home page image * Fix reference to screenshot * Copy edits * Update index.md Styles magic command * Copy edits based on @dlqqq's feedback * fix newline typo in improve_code (#364) * Remove frontend js unit tests as not planned (#371) * Added alias for bedrock titan model (#368) * Added alias for bedrock titan model * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * relax pinning on importlib_metadata, typing_extensions (#363) == pinning is almost always to be avoided in packages, and tight pinning on backports like importlib_metadata is extremely unlikely to be desirable * Loads vector store index lazily (#374) * Upgrades LangChain to 0.0.277 * Pinned Pydantic version, updated pydantic references * add .yarn to .gitignore in 1.x for local dev --------- Co-authored-by: michaelchia <michael_chia7@msn.com> Co-authored-by: Andrii Ieroshenko <aieroshe@amazon.com> Co-authored-by: Piyush Jain <piyushjain@duck.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Min RK <benjaminrk@gmail.com>
== pinning is almost always to be avoided in packages, and tight pinning on backports like importlib_metadata is extremely unlikely to be desirable
== pinning is almost always to be avoided in packages, and tight pinning on backports like importlib_metadata is extremely unlikely to be desirable
==pinning is ~always to be avoided in exported package dependencies, and tight pinning on backports like importlib_metadata is extremely unlikely to be desirable, since these prohibit users from installing later, compatible versions of these dependencies for their own use alongside jupyter-ai.I've replaced some very stable dependencies with strict pinnings (which cause many installation conflicts without any actual compatibility issues) to lower bounds, and replaced some strict
~=x.y.zwith semantic>=x.y.z,<x+1, where~=x.y.zproduces>=x.y.z,<x.y+1.I didn't update strict pinnings on dependencies like
langchain, which I strongly suspect should also be relaxed, but I can't speak to that one.In general, I'd encourage upper bounds to be on the major version at most, if upper bounds should be used at all (usually the answer is no unless compatibility breakage is likely, e.g. jupyterlab).