Increase mcp version #759
Increase mcp version #759
Conversation
| "railtracks", | ||
| "watchdog >= 3.0.0" | ||
| ] | ||
| dependencies = ["watchdog >= 3.0.0"] |
There was a problem hiding this comment.
railtracks removed as dependency?
There was a problem hiding this comment.
He's investigating the test failures with circular import structures, I don't think the PRs ready for review yet
…outdated-versioning-requirement' of https://github.com/RailtownAI/rc into 758-bug-mcp-not-functioning-in-jupyter-notebook-due-to-outdated-versioning-requirement
|
Finally found out reason why updating mcp version is failing, so currently, new mcp requires nvicorn>1.31.0, while litellm require <1.30, so its not working, however, the feature levi implemented requires item introduced in recent mcp versions so we have a deadlock with no current solutions. |
|
Raised issue on litellm: BerriAI/litellm#15511 |
|
@Granine looks like we had a reversion on our originaal fix PR BerriAI/litellm#16300 we will need to address ^ |
|
Looks like we have updation. Thanks @Granine |
|
Thanks @Granine, I'll just add this here for tracking: BerriAI/litellm#16307 |
What does this add?
Increase mcp required base version from 1.9.0 to 1.16.0. All test and local repo are using 1.16.0 by default so there is no code impact.
The reason of the change is that some features we used only exist >= 1.12, (mcp.os), hence it is worth updating. The reason for bumping directly to 1.16.0 is that all versions of 1.12 and above are backward safe. We wil also enjoy extra feature and supporting starting with latest version
Type of changes
Please check the type of change your PR introduces:
\