feat: add Xinference LLM support

[euxx]

Hi, Vanna is great. We have been integrating Vanna and Xinference recently.

Xinference is an open-source platform that can manage many open-source AI models.

We can use all the popular LLMs with Xinference, including: LLaMA and Qwen.

So the implementation may be helpful to others.

Project pages: https://github.com/xorbitsai/inference https://github.com/xorbitsai/inference-client

---

Basic Use:

$ pip install 'vanna[xinference-client]'

from vanna.vannadb import VannaDB_VectorStore
from vanna.xinference import Xinference

class VannaXinference(VannaDB_VectorStore, Xinference):
    def __init__(self, config=None):
        VannaDB_VectorStore.__init__(self, vanna_model=MY_VANNA_MODEL, vanna_api_key=MY_VANNA_API_KEY, config=config)
        Xinference.__init__(self, config=config)

vn_xinference = VannaXinference(config={'base_url': 'http://localhost:9997', 'model_uid': 'qwen2.5-instruct-14B'})

I verified it locally, and the following test passed:

vn_xinference = VannaXinference(config={'base_url': 'http://localhost:9997', 'model_uid': 'qwen2.5-instruct-14B'})
vn_xinference.connect_to_sqlite('https://vanna.ai/Chinook.sqlite')

def test_vn_xinference():
    sql = vn_xinference.generate_sql("What are the top 4 customers by sales?")
    df = vn_xinference.run_sql(sql)
    assert len(df) == 4