Reroute image models to GenerateContent in interactions node

interactions_nodes.py

@@ -28,6 +28,7 @@ from .gemini_nodes import (
     GetKeyAPI,
     SSL_GeminiAPIKeyConfig,
     SSL_GeminiVideoConfig,
+    SSL_GeminiTextPrompt,
 )
 
 
@@ -657,6 +658,49 @@ class SSL_GeminiInteractionsTextPrompt(IO.ComfyNode):
                 response_schema: ResponseSchema.Type | None = None) -> IO.NodeOutput:
 
         print(f"[INFO] SSL_GeminiInteractionsTextPrompt execute called, model: {model}")
+
+        # If model is an image generation model, reroute to SSL_GeminiTextPrompt (GenerateContent)
+        if model in cls.IMAGE_MODELS:
+            print(f"[INFO] Model '{model}' is an image generation model; rerouting to GenerateContent engine (SSL_GeminiTextPrompt).")
+            gen_output = SSL_GeminiTextPrompt.execute(
+                config=config,
+                prompt=prompt,
+                system_instruction=system_instruction,
+                model=model,
+                temperature=temperature,
+                top_p=top_p,
+                top_k=top_k,
+                max_output_tokens=max_output_tokens,
+                include_images=include_images,
+                aspect_ratio=aspect_ratio,
+                bypass_mode=bypass_mode,
+                thinking_budget=thinking_budget,
+                use_proxy=use_proxy,
+                proxy_host=proxy_host,
+                proxy_port=proxy_port,
+                use_seed=use_seed,
+                seed=seed,
+                timeout=timeout,
+                include_thoughts=include_thoughts,
+                thinking_level=thinking_level,
+                media_resolution=media_resolution,
+                retry_pattern=retry_pattern,
+                max_retries=max_retries,
+                timeout_fallback_text=timeout_fallback_text,
+                image_size=image_size,
+                video=video,
+                image_inputs=image_inputs,
+                response_schema=response_schema,
+            )
+            return IO.NodeOutput(
+                gen_output[0],  # text
+                gen_output[1],  # image
+                gen_output[2],  # final_actual_seed
+                gen_output[3],  # structured_output
+                gen_output[4],  # thoughts
+                "",             # interaction_id
+            )
+
         use_cache = bool(config.get("use_cache", False))
         cache_ttl_minutes = int(config.get("cache_ttl_minutes", 60))
         cache_seed = int(config.get("cache_seed", 0))

tests/test_interactions_nodes.py

@@ -289,19 +289,15 @@ def test_structured_output_with_schema(mock_interactions_client):
     ]
 
 
-def test_image_generation(mock_interactions_client):
-    # Create small valid 16x16 PNG image
-    img = Image.new("RGB", (16, 16), color=(255, 0, 0))
-    buf = BytesIO()
-    img.save(buf, format="PNG")
-    b64_png = base64.b64encode(buf.getvalue()).decode("utf-8")
+def test_image_generation(mock_interactions_client, monkeypatch):
+    called_with = []
+    fake_img = torch.zeros((1, 16, 16, 3), dtype=torch.float32)
 
-    mock_interactions_client.responses = [{
-        "text": "Generated an image",
-        "image_data": b64_png,
-        "image_mime": "image/png",
-        "id": "int_img_gen_1",
-    }]
+    def fake_generate(**kwargs):
+        called_with.append(kwargs)
+        return interactions_nodes.IO.NodeOutput("Generated an image", fake_img, 0, "", "")
+
+    monkeypatch.setattr(interactions_nodes.SSL_GeminiTextPrompt, "execute", fake_generate)
 
     args = _execute_kwargs(
         model="gemini-3.1-flash-image",
@@ -311,14 +307,16 @@ def test_image_generation(mock_interactions_client):
     )
     output = interactions_nodes.SSL_GeminiInteractionsTextPrompt.execute(**args)
 
+    assert output[0] == "Generated an image"
     image_tensor = output[1]
     assert torch.is_tensor(image_tensor)
     assert image_tensor.shape == (1, 16, 16, 3)
-
-    call = mock_interactions_client.calls[0]
-    assert call["response_format"] == [
-        {"type": "image", "aspect_ratio": "16:9", "image_size": "1K"}
-    ]
+    assert output[5] == ""  # interaction_id is empty for rerouted image models
+    assert len(called_with) == 1
+    assert called_with[0]["model"] == "gemini-3.1-flash-image"
+    assert called_with[0]["include_images"] is True
+    assert called_with[0]["aspect_ratio"] == "16:9"
+    assert called_with[0]["image_size"] == "1K"
 
 
 def test_thinking_and_thoughts_extraction(mock_interactions_client):