Allow Custom Client to be Passed

[PvH-SPC]

Hi,

If a user passes a custom client in via the constructor, it gets overwritten with the 'default' instance. Could we please enable passing of a custom build constructor.

databricks-ai-bridge/integrations/langchain/src/databricks_langchain/chat_models.py

Line 255 in b9afb57

self.client = get_deployment_client(self.target_uri)

[sunishsheth2009]

Hey @PvH-SPC, thank you for reporting this issue. I am curious what client would users want to use that support predict out of the box?
Also we are happy to take PRs if you have one with example usage as tests :) Thank you

[shahrukh-shaik]

Can you see if below helps

from mlflow.deployments import get_deploy_client
client = get_deploy_client("databricks")  # it uses databricks authentication environmental variables for auth, 
# such as DATABRICKS_HOST, DATABRICKS_TOKEN, DATABRICKS_CLIENT_ID/SECRET
chat_model = ChatDatabricks(
    endpoint="databricks-claude-3-7-sonnet",
    client=client,
    temperature=0.1,
    max_tokens=2560,
    extra_params={
        "thinking": {
            "type": "enabled",
            "budget_tokens": 1024
        }
    }
    # See https://python.langchain.com/api_reference/community/chat_models/langchain_community.chat_models.databricks.ChatDatabricks.html for other supported parameters
)