269c0ed80f
Fix width/height extraction and embeddings detection
PXL8D2025-08-13 12:13:48 -07:00
7ddc19c1b4
Session 8 Priority 1: Simplify LoRA output format to real_name and weight only
PXL8D2025-08-13 11:44:36 -07:00
cc24d46b0a
Fix Tensor.AI JSON parsing by cleaning null bytes, improve de-obfuscation logging and error handling, ensure proper model/LoRA mapping by index
PXL8D2025-08-12 21:41:07 -07:00
eeedbebc45
Critical bug fixes: resolve NameError in metadata scope, fix model tracing to continue through node references, improve LoRA mapping by index for Tensor.AI de-obfuscation
PXL8D2025-08-12 21:35:16 -07:00
bc4f891ee7
Implement Tensor.AI de-obfuscation system with VAE extraction - decode EMS model codes using generation_data chunk, extract real model names with hashes, add VAE support to component extractor
PXL8D2025-08-12 21:26:59 -07:00
c426f24953
Implement intelligent model resolution system - trace node connections to resolve actual model names and detect model types via VAE analysis
PXL8D2025-08-12 19:59:37 -07:00
53d4cda2fe
Fix LoRAs and embeddings extraction - add missing return statement and embeddings assignment
PXL8D2025-08-12 19:33:23 -07:00
f94df71c22
Enhance negative prompt detection and add embeddings support
PXL8D2025-08-12 19:18:19 -07:00
dbf44ae1a0
CRITICAL FIX: Prevent LoRA-only content from being selected as positive prompt
PXL8D2025-08-12 19:08:26 -07:00