fix(models): Add compatibility shim for mistralai v2.x import path (#…

[shivamnegi1705]

Description

mistralai v2.0 changed the import path for the Mistral client class from mistralai.Mistral to mistralai.client.Mistral. Since pyproject.toml specifies mistralai>=1.8.2 with no upper bound, fresh installs pull v2.0+ and fail at runtime with AttributeError: module 'mistralai' has no attribute 'Mistral' when stream() or structured_output() is called.

This PR adds a compatibility shim using importlib.import_module that tries the v2.x path first (mistralai.client.Mistral) and falls back to v1.x (mistralai.Mistral). All three mistralai.Mistral() call sites in stream() and structured_output() are replaced with the resolved MistralClient alias.

Related Issues

Closes #1924

Documentation PR

N/A — no documentation changes needed.

Type of Change

Bug fix

Testing

All 35 unit tests pass (33 existing + 2 new import shim tests).

Reproducing the issue

pip install 'strands-agents[mistral]'
# Pulls mistralai 2.0.4

Import succeeds but runtime usage fails:

python -c "from strands.models.mistral import MistralModel; print('OK')"
# OK

python -c "
import asyncio
from strands.models.mistral import MistralModel

m = MistralModel(model_id='mistral-small-latest')
messages = [{'role': 'user', 'content': [{'text': 'hi'}]}]

async def test():
    async for chunk in m.stream(messages):
        print(chunk)

asyncio.run(test())
"
# AttributeError: module 'mistralai' has no attribute 'Mistral'

After applying the fix

The same script with mistralai==2.0.4 no longer throws AttributeError. Instead it proceeds to the API call and succeeds.

Also verified with mistralai==1.12.4 (v1.x) — works as before with no regressions.

New unit tests

• test_mistral_client_import_v2: verifies shim resolves MistralClient from mistralai.client.Mistral (v2.x path)

• test_mistral_client_import_v1_fallback: verifies fallback to mistralai.Mistral when mistralai.client is unavailable (v1.x path)

• I ran hatch run prepare

Checklist

• I have read the CONTRIBUTING document
• I have added any necessary tests that prove my fix is effective or my feature works
• I have updated the documentation accordingly
• I have added an appropriate example to the documentation to outline the feature, or no new docs are needed
• My changes generate no new warnings
• Any dependent changes have been merged and published

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[mkmeral]

Hi @shivamnegi1705 , thank you for your PR.

However, I will close this PR in favor of #1935 We have not tested or gone through Mistral changes, so we cannot reliably say we support Mistral 2.0

We will release a patch change right now to make sure Mistral works as is (forces to v1, through upper bound). We can iterate on 2.0 client of MistralAI in another ticket