docs: clarify general no-LoRA runtime

MODEL_BOM.md

@@ -22,7 +22,7 @@ NSFW classifier, safety checker, or semantic blocklist is added in H2.
 
 The source Space exposed all 19 candidates below, and H2 pinned their revisions
 for provenance. The current public runtime enables none of them: requests run
-the pinned replacement transformer directly in base edit mode without loading,
+the pinned replacement transformer directly in general edit mode without loading,
 selecting, or applying a LoRA adapter.
 
 | API ID | Repository | Revision | Recorded status |
@@ -59,4 +59,4 @@ selecting, or applying a LoRA adapter.
 - Produce an exact Linux transitive dependency lock after a controlled build.
 - A previous `Anime-V2` smoke does not validate the new no-LoRA general edit path.
   Run a separate, explicitly authorized synthetic-image smoke before describing
-  base-mode image generation as end-to-end verified.
+  general no-LoRA image generation as end-to-end verified.

README.md

@@ -28,7 +28,7 @@ compatibility and accepts only the reserved internal value `__base__`.
 - Input shape: the upstream seven fields
 - Output shape: `{ "image": "data:image/png;base64,...", "seed": number }`
 - Technical validation: one PNG/JPEG/WebP data URL, at most 10 MiB and 16 MP,
-  Prompt length, numeric ranges, the reserved base edit mode, and one
+  Prompt length, numeric ranges, the reserved general edit mode, and one
   concurrent pipeline operation
 - Error output: stable low-cardinality codes; raw provider exceptions and user
   content are not returned or logged

app.py

@@ -125,7 +125,7 @@ def _clear_cuda_cache() -> None:
 
 @spaces.GPU(size="xlarge")
 def _run_gpu_edit(request: ValidatedEditRequest) -> tuple[bytes, int]:
-    """Run one base-transformer edit with no LoRA adapter loaded."""
+    """Run one general edit with the pinned replacement transformer and no LoRA."""
 
     _clear_cuda_cache()
     try:
@@ -177,7 +177,7 @@ def edit_image(
     """Validate on CPU, then call the internal ZeroGPU edit function.
 
     The public seven-field contract is retained for compatibility. The legacy
-    lora_adapter field must contain the reserved base-mode value; no LoRA is
+    lora_adapter field must contain the reserved general-mode value; no LoRA is
     loaded. Validation is technical only, with no semantic content moderation.
     """
 

tests/test_space_static.py

@@ -78,7 +78,7 @@ class SpaceStaticTests(unittest.TestCase):
         self.assertNotIn("set_adapters", source)
         self.assertNotIn("unload_lora_weights", source)
 
-    def test_public_runtime_is_explicitly_base_mode(self):
+    def test_public_runtime_is_explicitly_general_no_lora_mode(self):
         source = (ROOT / "app.py").read_text(encoding="utf-8")
         self.assertIn('"edit_mode": "general"', source)
         self.assertIn('"uses_lora": False', source)