ci: deploy pro_gradio_demo (b76bf27af8b8)

app.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 from nutonic_pro_gradio_demo.gradio_app import build_demo
+from nutonic_pro_gradio_demo.gradio_app import LEAFLET_JS
 
 # Hugging Face ZeroGPU Spaces require at least one `@spaces.GPU` function to be
 # defined at import time in the Space entry file (`app.py`).
@@ -18,5 +19,5 @@ except ImportError:
 demo = build_demo()
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch(ssr_mode=False, js=LEAFLET_JS)
 

nutonic_pro_gradio_demo/client.py

@@ -10,7 +10,7 @@ from urllib.parse import urljoin, urlparse
 import httpx
 
 from nutonic_pro_gradio_demo.hmac_signing import nutonic_hmac_headers
-from nutonic_pro_gradio_demo.models import ProJobCreateIn, ProJobCreateOut, ProJobStatusOut, ProVlmModelManifest
+from nutonic_pro_gradio_demo.models import ProJobCreateIn, ProJobCreateOut, ProJobStatusOut, ProVlmModelManifest, TokenResponse
 from nutonic_pro_gradio_demo.settings import Settings
 
 
@@ -52,6 +52,8 @@ class NutonicServerClient:
                 "User-Agent": "nutonic-pro-gradio-demo/0.1 (+https://huggingface.co/spaces/Tonic/nutonic-pro-demo)",
             },
         )
+        self._session_token: str | None = None
+        self._session_token_expires_at: float = 0.0
 
     @property
     def origin(self) -> str:
@@ -70,15 +72,24 @@ class NutonicServerClient:
         return self._origin + path
 
     def post_pro_job(self, body: ProJobCreateIn) -> ProJobCreateOut:
-        r = self._request_with_backoff("POST", "/api/v1/pro/jobs", json_body=body.model_dump(mode="json"))
+        r = self._request_with_backoff(
+            "POST",
+            "/api/v1/pro/jobs",
+            json_body=body.model_dump(mode="json"),
+            require_bearer=True,
+        )
         return ProJobCreateOut.model_validate(r.json())
 
     def get_pro_job(self, job_id: str) -> ProJobStatusOut:
-        r = self._request_with_backoff("GET", f"/api/v1/pro/jobs/{job_id}")
+        r = self._request_with_backoff("GET", f"/api/v1/pro/jobs/{job_id}", require_bearer=True)
         return ProJobStatusOut.model_validate(r.json())
 
     def get_artifact(self, *, job_id: str, artifact_id: str) -> bytes:
-        r = self._request_with_backoff("GET", f"/api/v1/pro/jobs/{job_id}/artifacts/{artifact_id}")
+        r = self._request_with_backoff(
+            "GET",
+            f"/api/v1/pro/jobs/{job_id}/artifacts/{artifact_id}",
+            require_bearer=True,
+        )
         return r.content
 
     def get_bytes_by_url(self, url_or_path: str) -> bytes:
@@ -87,9 +98,23 @@ class NutonicServerClient:
         return r.content
 
     def get_vlm_model_manifest(self) -> ProVlmModelManifest:
-        r = self._request_with_backoff("GET", "/api/v1/pro/vlm/model-manifest")
+        r = self._request_with_backoff("GET", "/api/v1/pro/vlm/model-manifest", require_bearer=True)
         return ProVlmModelManifest.model_validate(r.json())
 
+    def post_auth_token(self) -> TokenResponse:
+        r = self._request_with_backoff("POST", "/api/v1/auth/token")
+        return TokenResponse.model_validate(r.json())
+
+    def _ensure_bearer(self) -> str:
+        now = time.time()
+        if self._session_token and now < self._session_token_expires_at:
+            return self._session_token
+        issued = self.post_auth_token()
+        self._session_token = issued.access_token
+        # Refresh a bit early to avoid edge-of-expiry 401s.
+        self._session_token_expires_at = now + max(10, int(issued.expires_in) - 15)
+        return self._session_token
+
     def _signed_headers(self, *, method: str, url: str, body: bytes) -> dict[str, str]:
         if not self._hmac_secret:
             return {}
@@ -103,6 +128,7 @@ class NutonicServerClient:
         absolute: bool = False,
         max_retries: int = 4,
         json_body: Any | None = None,
+        require_bearer: bool = False,
     ) -> httpx.Response:
         """
         Hugging Face Spaces can return 429 under load. Handle 429 (and some transient 5xx)
@@ -123,6 +149,8 @@ class NutonicServerClient:
             attempt += 1
             url = path_or_url if absolute else self._url(path_or_url)
             headers: dict[str, str] = self._signed_headers(method=method, url=url, body=body_bytes)
+            if require_bearer:
+                headers["Authorization"] = f"Bearer {self._ensure_bearer()}"
             request_kwargs: dict[str, Any] = {}
             if json_body is not None:
                 # Send the same exact bytes we hashed; otherwise httpx may re-serialize

nutonic_pro_gradio_demo/gradio_app.py

@@ -14,6 +14,8 @@ from nutonic_pro_gradio_demo.settings import get_settings
 from nutonic_pro_gradio_demo.vlm_parse import parse_vlm_output
 from nutonic_pro_gradio_demo.vlm_runtime import ensure_model_loaded, infer_caption_and_boxes
 
+LEAFLET_JS = ""
+
 
 def _bbox_half_km_for_zoom(zoom: int) -> float:
     z = int(zoom)
@@ -221,6 +223,8 @@ def _run_full_pipeline(
 
 def build_demo() -> gr.Blocks:
     settings = get_settings()
+    # Used by app.py (Gradio 6 moved js= from Blocks(...) to launch()).
+    # Keep it as a local value and return it via a module constant below.
     leaflet_js = r"""
 function() {
   // Load Leaflet (CSS + JS) dynamically (OSM tiles; no Mapbox).
@@ -300,7 +304,10 @@ function() {
 }
 """
 
-    with gr.Blocks(title="NU:TONIC PRO (ZeroGPU demo)", js=leaflet_js) as demo:
+    global LEAFLET_JS
+    LEAFLET_JS = leaflet_js
+
+    with gr.Blocks(title="NU:TONIC PRO (ZeroGPU demo)") as demo:
         if settings.require_server_origin and not settings.nutonic_server_origin.strip():
             gr.Markdown(
                 "## Configuration required\n\n"

nutonic_pro_gradio_demo/models.py

@@ -5,6 +5,12 @@ from typing import Any, Literal
 from pydantic import BaseModel, Field
 
 
+class TokenResponse(BaseModel):
+    access_token: str
+    token_type: str = "bearer"
+    expires_in: int
+
+
 ProJobProfile = Literal[
     "wildfire",
     "oceanscout_ship_detection",