[Integration] Add Cleanlab + Strands Agent integration #107
Conversation
| @@ -0,0 +1,9 @@ | |||
| # Requirements for cleanlab_codex strands integration | |||
| # Core dependencies | |||
| cleanlab-tlm>=1.1.14 | |||
There was a problem hiding this comment.
we need it for form_response_string_chat_completions_api like functions from cleanlab_tlm.utils.chat
| @@ -0,0 +1,9 @@ | |||
| # Requirements for cleanlab_codex strands integration | |||
There was a problem hiding this comment.
can you explicitly confirm that you've tested using the cleanlab-codex in an environment where strands and these other optional dependencies are NOT installed, and the rest of codex package still works as usual? I.e. that these dependencies truly are optional
There was a problem hiding this comment.
I tested this in an env without strands + with strands. We are not adding any dependencies to the cleanlab package here, just in a requirements.txt for users to manually install
| # Requirements for cleanlab_codex strands integration | ||
| # Core dependencies | ||
| cleanlab-tlm>=1.1.14 | ||
| openai |
There was a problem hiding this comment.
cleanlab.validate() expects messages in OpenAI format, to pass typechecker I had to typecast which means we now require types during runtime here.
There was a problem hiding this comment.
just had small suggestions and things to confirm.
I'm inclined not to invest other engineers' time on this experimental module. But if there's anything you are concerned about in terms of affecting the overall Codex experience for developers who are not using this strands module, then please do ask engineer to review that aspect.
|
I fixed the typecasting issue. This should be good to go. |
2 participants
Key Info
Note: typeignore is added to the types because this is experimental we do not want to require importing any strands code into the real package
Implementation plan: link
Priority:
What changed?
This adds an experimental folder with a strands integration
What do you want the reviewer(s) to focus on?
Checklist