fix(inference): namespace model gateway endpoint

What changed

The model gateway is now mounted at /inference/v1 instead of the root /v1 path. The inference profile detail page now shows that namespaced base address in the client configuration, and the focused controller and LiveView tests assert the new path. The self-hosting guide now documents the same client endpoints.

Why

The public gateway path should make it clear that the route belongs to Hive’s inference surface. Downstream deployments that use Hive as their model gateway can now point clients at https://hive.tuist.dev/inference/v1 instead of sharing the root versioned path.

Root cause

The gateway was mounted directly at /v1, so dashboard guidance and downstream configuration naturally produced https://hive.tuist.dev/v1.

Approach

Keep upstream provider addresses unchanged and move only Hive’s client-facing gateway namespace. That preserves provider routing while making the public Hive address explicit.

Impact

Clients using Hive’s old /v1 base address need to update to /inference/v1. Provider configurations that already point to upstream services remain unchanged.

Verification screenshots

Before: The review screenshot showed a downstream base address of https://hive.tuist.dev/v1.

After: Captured through a headless Chrome browser at /tmp/hive-inference-v1-after.png, showing /inference/v1/models reaches Hive’s gateway and returns the expected missing-token response.

Validation

• mix format lib/hive_web/router.ex lib/hive_web/live/ops_live/inference_profile.ex test/hive_web/controllers/inference_controller_test.exs test/hive_web/live/ops_live/inference_test.exs
• mix test test/hive_web/controllers/inference_controller_test.exs test/hive_web/live/ops_live/inference_test.exs: 17 tests passed.
• curl -i --max-time 10 http://127.0.0.1:4876/inference/v1/models: returned the expected unauthorized gateway response.