Add custom nodes, Civitai loras (LFS), and vast.ai setup script

Includes 30 custom nodes committed directly, 7 Civitai-exclusive
loras stored via Git LFS, and a setup script that installs all
dependencies and downloads HuggingFace-hosted models on vast.ai.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

custom_nodes/rgthree-comfy/py/sdxl_power_prompt_postive.py

@@ -0,0 +1,178 @@
+import os
+import re
+from nodes import MAX_RESOLUTION
+from comfy_extras.nodes_clip_sdxl import CLIPTextEncodeSDXL
+
+from .log import log_node_warn, log_node_info, log_node_success
+from .constants import get_category, get_name
+from .power_prompt_utils import get_and_strip_loras
+from nodes import LoraLoader, CLIPTextEncode
+import folder_paths
+
+NODE_NAME = get_name('SDXL Power Prompt - Positive')
+
+
+class RgthreeSDXLPowerPromptPositive:
+  """The Power Prompt for positive conditioning."""
+
+  NAME = NODE_NAME
+  CATEGORY = get_category()
+
+  @classmethod
+  def INPUT_TYPES(cls):  # pylint: disable = invalid-name, missing-function-docstring
+    # Removed Saved Prompts feature; No sure it worked any longer. UI should fail gracefully,
+    # TODO: Rip out saved prompt input data
+    SAVED_PROMPTS_FILES=[]
+    SAVED_PROMPTS_CONTENT=[]
+    return {
+      'required': {
+        'prompt_g': ('STRING', {
+          'multiline': True,
+          'dynamicPrompts': True
+        }),
+        'prompt_l': ('STRING', {
+          'multiline': True,
+          'dynamicPrompts': True
+        }),
+      },
+      'optional': {
+        "opt_model": ("MODEL",),
+        "opt_clip": ("CLIP",),
+        "opt_clip_width": ("INT", {
+          "forceInput": True,
+          "default": 1024.0,
+          "min": 0,
+          "max": MAX_RESOLUTION
+        }),
+        "opt_clip_height": ("INT", {
+          "forceInput": True,
+          "default": 1024.0,
+          "min": 0,
+          "max": MAX_RESOLUTION
+        }),
+        'insert_lora': (['CHOOSE', 'DISABLE LORAS'] +
+                        [os.path.splitext(x)[0] for x in folder_paths.get_filename_list('loras')],),
+        'insert_embedding': ([
+          'CHOOSE',
+        ] + [os.path.splitext(x)[0] for x in folder_paths.get_filename_list('embeddings')],),
+        'insert_saved': ([
+          'CHOOSE',
+        ] + SAVED_PROMPTS_FILES,),
+        # We'll hide these in the UI for now.
+        "target_width": ("INT", {
+          "default": -1,
+          "min": -1,
+          "max": MAX_RESOLUTION
+        }),
+        "target_height": ("INT", {
+          "default": -1,
+          "min": -1,
+          "max": MAX_RESOLUTION
+        }),
+        "crop_width": ("INT", {
+          "default": -1,
+          "min": -1,
+          "max": MAX_RESOLUTION
+        }),
+        "crop_height": ("INT", {
+          "default": -1,
+          "min": -1,
+          "max": MAX_RESOLUTION
+        }),
+      },
+      'hidden': {
+        'values_insert_saved': (['CHOOSE'] + SAVED_PROMPTS_CONTENT,),
+      }
+    }
+
+  RETURN_TYPES = ('CONDITIONING', 'MODEL', 'CLIP', 'STRING', 'STRING')
+  RETURN_NAMES = ('CONDITIONING', 'MODEL', 'CLIP', 'TEXT_G', 'TEXT_L')
+  FUNCTION = 'main'
+
+  def main(self,
+           prompt_g,
+           prompt_l,
+           opt_model=None,
+           opt_clip=None,
+           opt_clip_width=None,
+           opt_clip_height=None,
+           insert_lora=None,
+           insert_embedding=None,
+           insert_saved=None,
+           target_width=-1,
+           target_height=-1,
+           crop_width=-1,
+           crop_height=-1,
+           values_insert_saved=None):
+
+    if insert_lora == 'DISABLE LORAS':
+      prompt_g, loras_g, _skipped, _unfound = get_and_strip_loras(prompt_g,
+                                                                  True,
+                                                                  log_node=self.NAME)
+      prompt_l, loras_l, _skipped, _unfound = get_and_strip_loras(prompt_l,
+                                                                  True,
+                                                                  log_node=self.NAME)
+      loras = loras_g + loras_l
+      log_node_info(
+        NODE_NAME,
+        f'Disabling all found loras ({len(loras)}) and stripping lora tags for TEXT output.')
+    elif opt_model is not None and opt_clip is not None:
+      prompt_g, loras_g, _skipped, _unfound = get_and_strip_loras(prompt_g, log_node=self.NAME)
+      prompt_l, loras_l, _skipped, _unfound = get_and_strip_loras(prompt_l, log_node=self.NAME)
+      loras = loras_g + loras_l
+      if len(loras) > 0:
+        for lora in loras:
+          opt_model, opt_clip = LoraLoader().load_lora(opt_model, opt_clip, lora['lora'],
+                                                       lora['strength'], lora['strength'])
+          log_node_success(NODE_NAME, f'Loaded "{lora["lora"]}" from prompt')
+        log_node_info(NODE_NAME, f'{len(loras)} Loras processed; stripping tags for TEXT output.')
+    elif '<lora:' in prompt_g or '<lora:' in prompt_l:
+      _prompt_g, loras_g, _skipped, _unfound = get_and_strip_loras(prompt_g,
+                                                                   True,
+                                                                   log_node=self.NAME)
+      _prompt_l, loras_l, _skipped, _unfound = get_and_strip_loras(prompt_l,
+                                                                   True,
+                                                                   log_node=self.NAME)
+      loras = loras_g + loras_l
+      if len(loras):
+        log_node_warn(
+          NODE_NAME, f'Found {len(loras)} lora tags in prompt but model & clip were not supplied!')
+        log_node_info(NODE_NAME, 'Loras not processed, keeping for TEXT output.')
+
+    conditioning = self.get_conditioning(prompt_g, prompt_l, opt_clip, opt_clip_width,
+                                         opt_clip_height, target_width, target_height, crop_width,
+                                         crop_height)
+
+    return (conditioning, opt_model, opt_clip, prompt_g, prompt_l)
+
+  def get_conditioning(self, prompt_g, prompt_l, opt_clip, opt_clip_width, opt_clip_height,
+                       target_width, target_height, crop_width, crop_height):
+    """Checks the inputs and gets the conditioning."""
+    conditioning = None
+    if opt_clip is not None:
+      do_regular_clip_text_encode = opt_clip_width and opt_clip_height
+      if do_regular_clip_text_encode:
+        target_width = target_width if target_width and target_width > 0 else opt_clip_width
+        target_height = target_height if target_height and target_height > 0 else opt_clip_height
+        crop_width = crop_width if crop_width and crop_width > 0 else 0
+        crop_height = crop_height if crop_height and crop_height > 0 else 0
+        try:
+          conditioning = CLIPTextEncodeSDXL().encode(opt_clip, opt_clip_width, opt_clip_height,
+                                                     crop_width, crop_height, target_width,
+                                                     target_height, prompt_g, prompt_l)[0]
+        except Exception:
+          do_regular_clip_text_encode = True
+          log_node_info(
+            self.NAME,
+            'Exception while attempting to CLIPTextEncodeSDXL, will fall back to standard encoding.'
+          )
+      else:
+        log_node_info(
+          self.NAME,
+          'CLIP supplied, but not CLIP_WIDTH and CLIP_HEIGHT. Text encoding will use standard ' +
+          'encoding with prompt_g and prompt_l concatenated.')
+
+      if not do_regular_clip_text_encode:
+        conditioning = CLIPTextEncode().encode(
+          opt_clip, f'{prompt_g if prompt_g else ""}\n{prompt_l if prompt_l else ""}')[0]
+    return conditioning