Env var `OLLAMA_BASE_URL` is inconsistent with actual Ollama-held env var name

[sealad886]

Graphfy uses OLLAMA_BASE_URL to configure a custom API target here:

graphify/graphify/llm.py

Line 86 in cb96bda

"base_url": os.environ.get("OLLAMA_BASE_URL", "http://localhost:11434/v1"),

However, Ollama defines the OLLAMA_HOST env var and doesn't actually have a OLLAMA_BASE_URL concept at all:

Ollama Documentation How Do I Configure Ollama Server

---

Expected behavior:
Graphify should re-use the Ollama env vars so that custom configuration is automatically picked up.

[SinghAman21]

OLLAMA_HOST is the real variable and graphify doesn't read it.

I'd suggest adding it as a fallback rather than renaming: OLLAMA_BASE_URL still wins if set (it matches our other *_BASE_URL vars, and renaming would break existing users), and we fall back to OLLAMA_HOST when it's unset.

Two gotchas for to pick this up:

1. It's not a drop-in. OLLAMA_HOST is a bare host:port (0.0.0.0:11434), but we need a full OpenAI-compat URL — our default is http://localhost:11434/v1. So it needs a scheme added and /v1 appended.
2. 0.0.0.0 is a bind address, not a destination. That's the value the docs tell people to set, and connecting to http://0.0.0.0:11434 is flaky outside Linux. Should map to 127.0.0.1.

Also worth deciding: should a bare OLLAMA_HOST auto-select the ollama backend the way OLLAMA_BASE_URL does today? It's a lot more likely to just be lying around in someone's env, so I'd keep it config-only for now and leave detection alone.

Small related thing: README says the default is http://localhost:11434, but the code says http://localhost:11434/v1. Worth fixing too.

Happy to PR this if the fallback approach sounds good.

@safishamsi