Fix ComfyUI prompt validation errors for lm_top_k, repainting_start, repainting_end

[Copilot]

ComfyUI's prompt validation was failing before any node code ran because the widget positional order in INPUT_TYPES didn't match what the workflow files had saved, and VALIDATE_INPUTS didn't cover the three fields ComfyUI was choking on.

Two distinct bugs

1. Widget order mismatch — the primary crash

src_audio (STRING) was defined at widget position 18 in INPUT_TYPES, but all four example workflows were serialized with audio_cover_strength=0.5 at position 18 and src_audio='' at position 22. This shifted the slot mapping so that position 21 — a '' string saved for the lego dropdown — was assigned to repainting_end (FLOAT), causing float('') to blow up in ComfyUI's validation layer, before generate() or _coerce_float were ever called.

Fix: moved src_audio in INPUT_TYPES (and generate() signature) to after lego, aligning positions with the saved workflow files:

# Before (broken)          # After (correct)
18: src_audio  ← STRING   18: audio_cover_strength ← FLOAT
19: audio_cover_strength   19: repainting_start
20: repainting_start       20: repainting_end
21: repainting_end ← ''   21: lego              ← '' ✓
22: lego                   22: src_audio

2. VALIDATE_INPUTS incomplete — empty-string bypass missing

ComfyUI skips its own int()/float() conversion only for inputs explicitly named in VALIDATE_INPUTS. lm_top_k, repainting_start, and repainting_end were not listed, so older user workflows with "" serialized for those fields still hit int('')/float('') directly. Added all three alongside the existing lm_top_p / audio_cover_strength entries.

Tests that weren't catching this

• Workflow tests only verified indices 14 (lm_top_p) and 18 (audio_cover_strength) were numeric — missing 15, 19, 20
• No test on total widget count (25) or src_audio position relative to lego
• No tests for VALIDATE_INPUTS at all

Added: TestValidateInputs (22 cases), test_src_audio_after_lego_in_widget_order, test_generate_widget_count, and numeric-type assertions for all three previously unchecked indices across all four workflow files.

README

Added a "Updating an existing installation" section (git pull + restart) with a note that users whose workflows show validation errors after an update should delete and re-add the Generate node to reset stale widget values.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[lmangani]

The issue was not resolved:

Failed to validate prompt for output 3:

• AcestepCPPGenerate 2:
  • Failed to convert an input value to a FLOAT value: repainting_end, , could not convert string to float: ''
  • Failed to convert an input value to a INT value: lm_top_k, , invalid literal for int() with base 10: ''
    Output will be ignored
    invalid prompt: {'type': 'prompt_outputs_failed_validation', 'message': 'Prompt outputs failed validation', 'details': '', 'extra_info': {}}