Update providers/together_provider.py

providers/together_provider.py

@@ -0,0 +1,33 @@
+# providers/together_provider.py
+from typing import List
+from together import Together
+from .base import ImageProvider
+
+_DEFAULT_MODEL = "black-forest-labs/FLUX.1-dev"
+
+class TogetherImageProvider(ImageProvider):
+    """
+    Together Images API (FLUX family).
+    Docs: https://docs.together.ai/docs/images-overview
+          https://docs.together.ai/reference/images
+    """
+    def __init__(self, api_key: str, model: str = _DEFAULT_MODEL):
+        # The SDK reads TOGETHER_API_KEY from env automatically,
+        # but we also pass it explicitly to be safe.
+        self.client = Together(api_key=api_key)
+        self.model = model
+
+    def generate_image(self, prompt: str, size: str = "1024x1024", n: int = 1) -> List[str]:
+        width, height = (int(x) for x in size.split("x"))
+        # steps 24–30 are a good sweet spot for interiors; dev is fast, pro is premium.
+        resp = self.client.images.generate(
+            prompt=prompt,
+            model=self.model,
+            width=width,
+            height=height,
+            steps=28,
+            n=n,
+            response_format="b64_json",  # return base64 in JSON
+        )
+        # Each item exposes .b64_json
+        return [d.b64_json for d in resp.data]