PEP 420 namespace support, widen otel upper bound#346
Conversation
|
CI fails because |
2500f70 to
9a71b31
Compare
7bd332a to
02bffb1
Compare
README.md
Outdated
|
|
||
| ```bash | ||
| pip install mistralai[gcp] | ||
| uv add "mistralai[gcp]" |
There was a problem hiding this comment.
@Nelson-PROIA apologies for contradictory review, but I think we should use pip in doc (😶🌫️) (like other big SDK do).
Actually I don't think we should do this in this order. The main SDK (on which the workflow SDK depends) should first be updated no? Then, the workflow SDK can be updated and published with a new major version? |
@HHK1 yes your are correct. I was incorrectly understanding how the whole the workflow SDK fitted with the rest. |
louis-sanna-dev
left a comment
There was a problem hiding this comment.
@Nelson-PROIA looks good, let's just move back to pip in doc and merge.
- Widen opentelemetry-semantic-conventions upper bound from <0.60 to <0.61 - Add "Additional packages" section to README for mistralai-* namespace packages Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
be055f3 to
5fb625c
Compare
Summary
opentelemetry-semantic-conventionsupper bound from<0.60to<0.61to allowmistralaiandmistralai-workflowsto coexist (fixes [BUG CLIENT]:opentelemetry-semantic-conventionsupper bound<0.60blocks compatibility withopentelemetry-api>=1.39.0#341)mistralai-*namespace packagesValidation
Built wheels from both repos and verified the PEP 420 namespace merge in a clean venv using the following script:
Known issue: CI examples failing
The
run_example_scriptsCI job fails on fine-tuning examples (async_jobs.py,jobs.py,dry_run_job.py) because no models are available for completion fine-tuning on the CI API key. This is a pre-existing issue unrelated to this PR.Builds on #340.