Update app.py

app.py

@@ -120,51 +120,36 @@ def call_inference_api(model_id: str, prompt: str, width: int, height: int) -> I
         raise RuntimeError(f"{model_id}:{r.status_code}")
     return Image.open(BytesIO(r.content)).convert("RGB")
 
-def call_public_space_dynamic(prompt: str, width: int, height: int) -> Image.Image:
-    # Discover an api_name and feed only fields we can infer.
+def call_public_space(prompt: str, width: int, height: int) -> Image.Image:
+    """Use the FLUX public Space directly via its /infer endpoint."""
     from gradio_client import Client
-    client = Client(PUBLIC_SPACE_ID)
-    info = client.view_api(all_endpoints=True)
-    last_err = None
-    for ep in info:
-        api = ep.get("api_name")
-        params = ep.get("parameters", [])
-        # Build arg list by index
-        args = [None] * len(params)
-        bad = False
-        for idx, p in enumerate(params):
-            label = (p.get("label") or "").lower()
-            # Skip endpoints that clearly require an image upload
-            if "image" in label and "prompt" not in label:
-                bad = True
-                break
-            if "prompt" in label or label in ("prompt", "text", "caption"):
-                args[idx] = prompt
-            elif "width" in label:
-                args[idx] = int(width)
-            elif "height" in label:
-                args[idx] = int(height)
-            elif "negative" in label:
-                args[idx] = ""
-            elif "steps" in label or "num_inference_steps" in label:
-                args[idx] = 4
-            elif "guidance" in label or "cfg" in label:
-                args[idx] = 3.5
-            elif "seed" in label:
-                args[idx] = 42
-            else:
-                # leave None; gradio_client will fill defaults
-                pass
-        if bad:  # requires an image input etc.
-            continue
-        try:
-            res = client.predict(*args, api_name=api)
-            if isinstance(res, list): res = res[0]
-            return Image.open(res).convert("RGB")
-        except Exception as e:
-            last_err = e
-            continue
-    raise RuntimeError(f"space-no-endpoint:{last_err}")
+    client = Client("black-forest-labs/FLUX.1-schnell")
+    # order: prompt, seed, randomize_seed, width, height, num_inference_steps
+    result, _seed = client.predict(
+        prompt,
+        0,            # seed (0 = let Space choose unless randomize_seed=False)
+        True,         # randomize_seed
+        int(width),
+        int(height),
+        4,            # num_inference_steps (keep tiny for speed on mobile)
+        api_name="/infer"
+    )
+    # result is a dict with path/url
+    path = None
+    if isinstance(result, dict):
+        path = result.get("path") or result.get("url")
+    elif isinstance(result, list) and result:
+        item = result[0]
+        if isinstance(item, dict):
+            path = item.get("path") or item.get("url")
+        else:
+            path = item
+    else:
+        path = result
+    if not path:
+        raise RuntimeError("public-space returned empty result")
+    from PIL import Image
+    return Image.open(path).convert("RGB")
 
 def generate_image_auto(prompt: str, width: int, height: int):
     tried = []
@@ -179,8 +164,8 @@ def generate_image_auto(prompt: str, width: int, height: int):
                 continue
     # 2) Public Space dynamic
     try:
-        img = call_public_space_dynamic(prompt, width, height)
-        return img, f"✅ Public Space: **{PUBLIC_SPACE_ID}** (discovered endpoint)"
+        img = call_public_space(prompt, width, height)
+return img, "✅ Public Space: FLUX /infer"
     except Exception as e:
         tried.append(f"{PUBLIC_SPACE_ID}→{str(e)}")
     # 3) Gradient