lablink_allocator_service.providers.connectivity.mesh_overlay¶

`lablink_allocator_service.providers.connectivity.mesh_overlay` ¶

Mesh-overlay connectivity: browser -> allocator nginx proxy -> client KasmVNC, reached over a Tailscale overlay instead of a routable LAN/VPC address. Not Run:AI-specific — any client that isn't on the allocator's own network fits this connectivity strategy; Run:AI-hosted client workloads are simply the motivating case.

Classes¶

`MeshOverlayClientConnectivity` ¶

Attributes¶

`name = 'mesh_overlay'` `class-attribute` `instance-attribute` ¶

`requires_tailscale_check = True` `class-attribute` `instance-attribute` ¶

Methods:¶

`make_join_material(*, allocator_url, client_image, register_token)` ¶

Source code in packages/allocator/src/lablink_allocator_service/providers/connectivity/mesh_overlay.py

def make_join_material(
    self,
    *,
    allocator_url: str,
    client_image: str,
    register_token: str,
) -> ClientJoinMaterial:
    return ClientJoinMaterial(
        register_token=register_token,
        allocator_url=allocator_url,
        connectivity=self.name,
        client_image=client_image,
    )

`prepare_browser_session(**kwargs)` ¶

Source code in packages/allocator/src/lablink_allocator_service/providers/connectivity/mesh_overlay.py

def prepare_browser_session(self, **kwargs) -> BrowserSessionTarget:
    kwargs.setdefault("fallback_fn", _resolve_overlay_host)
    return prepare_browser_session(**kwargs)

lablink_allocator_service.providers.connectivity.mesh_overlay¶