[DOC]: OpenAILLM base_url is the /v1 root for OpenAI-compat hosts

Suggested Changes

Document OpenAILLM against an OpenAI-compatible Chat Completions host: base_url is the /v1 root, model_name is that host's catalog id.

A host whose public catalog is GET https://api.pzero.studio/v1/models (no key) fits:

from neo4j_graphrag.llm import OpenAILLM

llm = OpenAILLM(
    model_name="deepseek-v4-flash",
    api_key="<Bearer key for that host>",
    base_url="https://api.pzero.studio/v1",
)

Do not use AzureOpenAILLM (that constructs AzureOpenAI). OpenAILLM posts Chat Completions (client.chat.completions.create). Keep OpenAIEmbeddings on a real embedding host. Do not paste /v1/chat/completions as base_url. Override example gpt-5 with a catalog id from that host.

Not asking for a first-party provider. OpenAILLM plus base_url is enough.

Additional Info

Verified against src/neo4j_graphrag/llm/openai_llm.py (OpenAILLM.__init__ forwards base_url to both SDK clients).

Start by reading src/neo4j_graphrag/llm/openai_llm.py, especially OpenAILLM.init, to confirm how base_url is forwarded. Update the relevant documentation with the OpenAI-compatible Chat Completions example, explaining that base_url is the host's /v1 root and model_name is its catalog id; done when the documented example and cautions match the verified behavior.