ComfyUI backport release v0.22.3

[Kosinkadink]

Backport for v0.22.3. Cherry-picks:

• 0cce76d [Partner Nodes] feat: improve video references uploading for SeeDance 2 ([Partner Nodes] feat: improve video references uploading for SeeDance 2 #14098)
• 53eba22 chore: update workflow templates to v0.9.85 (Update workflow templates to v0.9.85 #14134)
• 175e854 [Partner Nodes] feat: add Krea2 nodes ([Partner Nodes] feat: add Krea2 Image nodes #14130)

API Node PR Checklist

Scope

• Is API Node Change

Pricing & Billing

• Need pricing update
• No pricing update

If Need pricing update:

• Metronome rate cards updated
• Auto‑billing tests updated and passing

QA

• QA done
• QA not required

Comms

• Informed Kosinkadink

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR introduces three new LLM/3D-generation node modules (Krea2 image generation, OpenRouter chat completions, and Rodin Gen-2.5 support), adds Claude thinking behavior with per-model configuration, refactors video resizing utilities to support both downscaling and upscaling, integrates the new utilities into ByteDance Seedance with virtual-library asset uploads, and bumps package versions to 0.22.2. The changes span infrastructure (video handling), feature additions (LLM nodes, 3D generation), and utility updates (quality-mode lookup tables, response validation).

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 25.40% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Title check	⚠️ Warning	The title states 'v0.22.3' but the actual version changes in the PR are to '0.22.2' (comfyui_version.py and pyproject.toml), not '0.22.3'.	Update the title to reflect the actual version being released, such as 'ComfyUI backport release v0.22.2' to match the version bumps in comfyui_version.py and pyproject.toml.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description check	✅ Passed	The PR description clearly identifies it as a backport for v0.22.3 and lists three specific cherry-picked commits with their purposes, directly corresponding to the changeset modifications.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 2

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@comfy_api_nodes/nodes_anthropic.py`:
- Around line 268-270: Validate the incoming reasoning_effort before indexing
_REASONING_BUDGET: check if reasoning_effort is a key in _REASONING_BUDGET and
if not raise a clear validation error (e.g., ValueError or a
NodeValidationError) with a message like "Invalid reasoning_effort: <value>";
only after that perform budget = _REASONING_BUDGET[reasoning_effort], apply the
min(...) cap, and construct AnthropicThinkingConfig as before (references:
_REASONING_BUDGET, reasoning_effort, AnthropicThinkingConfig, thinking_cfg).

In `@comfy_api_nodes/nodes_rodin.py`:
- Around line 581-588: The list comprehension that builds files opens file
handles via open(image, "rb") without closing them, causing a resource leak;
replace the comprehension with logic that uses a helper (e.g., _open_image_file)
to open string paths, append those open file objects to a local list, and ensure
all opened file handles are closed in a finally block after the upload/usage
completes; keep using tensor_to_filelike(image) for non-string inputs and
reference/modify the variable files and the helper _open_image_file so every
file opened is explicitly closed even on exceptions.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 7d9f3154-dfe0-43ad-a5f7-5f0c534e2091

📥 Commits

Reviewing files that changed from the base of the PR and between 175e854 and 0cc5934.

⛔ Files ignored due to path filters (5)

• comfy_api_nodes/apis/anthropic.py is excluded by !comfy_api_nodes/apis/**
• comfy_api_nodes/apis/bytedance.py is excluded by !comfy_api_nodes/apis/**
• comfy_api_nodes/apis/krea.py is excluded by !comfy_api_nodes/apis/**
• comfy_api_nodes/apis/openrouter.py is excluded by !comfy_api_nodes/apis/**
• comfy_api_nodes/apis/rodin.py is excluded by !comfy_api_nodes/apis/**

📒 Files selected for processing (10)

• comfy_api_nodes/nodes_anthropic.py
• comfy_api_nodes/nodes_bytedance.py
• comfy_api_nodes/nodes_krea.py
• comfy_api_nodes/nodes_openrouter.py
• comfy_api_nodes/nodes_rodin.py
• comfy_api_nodes/util/__init__.py
• comfy_api_nodes/util/conversions.py
• comfyui_version.py
• pyproject.toml
• requirements.txt

[coderabbitai]

⚠️ Potential issue | 🟡 Minor | ⚡ Quick win

Guard reasoning_effort before budget lookup.

_REASONING_BUDGET[reasoning_effort] can throw a KeyError if the incoming model payload is malformed (e.g., edited workflow JSON). Please fail with a clear validation error instead of an unhandled exception.

Suggested fix

-                budget = _REASONING_BUDGET[reasoning_effort]
+                budget = _REASONING_BUDGET.get(reasoning_effort)
+                if budget is None:
+                    raise ValueError(f"Invalid reasoning_effort: {reasoning_effort!r}")
                 budget = min(budget, max(1024, max_tokens - 1024))
                 thinking_cfg = AnthropicThinkingConfig(type="enabled", budget_tokens=budget)

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	budget = _REASONING_BUDGET[reasoning_effort]
	budget = min(budget, max(1024, max_tokens - 1024))
	thinking_cfg = AnthropicThinkingConfig(type="enabled", budget_tokens=budget)
	budget = _REASONING_BUDGET.get(reasoning_effort)
	if budget is None:
	raise ValueError(f"Invalid reasoning_effort: {reasoning_effort!r}")
	budget = min(budget, max(1024, max_tokens - 1024))
	thinking_cfg = AnthropicThinkingConfig(type="enabled", budget_tokens=budget)

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@comfy_api_nodes/nodes_anthropic.py` around lines 268 - 270, Validate the
incoming reasoning_effort before indexing _REASONING_BUDGET: check if
reasoning_effort is a key in _REASONING_BUDGET and if not raise a clear
validation error (e.g., ValueError or a NodeValidationError) with a message like
"Invalid reasoning_effort: <value>"; only after that perform budget =
_REASONING_BUDGET[reasoning_effort], apply the min(...) cap, and construct
AnthropicThinkingConfig as before (references: _REASONING_BUDGET,
reasoning_effort, AnthropicThinkingConfig, thinking_cfg).

[coderabbitai]

⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Resource leak: file handles opened without context manager.

When isinstance(image, str) is True, open(image, "rb") creates a file handle that's never explicitly closed. This can exhaust file descriptors under heavy use.

Proposed fix using a helper to manage file lifecycle

A minimal fix is to restructure so the files list is built with proper cleanup. One approach:

+    file_handles = []
+    try:
         files = None
         if images:
             files = [
                 (
                     "images",
-                    open(image, "rb") if isinstance(image, str) else tensor_to_filelike(image),
+                    _open_image_file(image, file_handles),
                 )
                 for image in images
                 if image is not None
             ]
 
         response = await sync_op(
             ...
         )
+    finally:
+        for fh in file_handles:
+            fh.close()

Where _open_image_file appends file handles to the list for later cleanup.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@comfy_api_nodes/nodes_rodin.py` around lines 581 - 588, The list
comprehension that builds files opens file handles via open(image, "rb") without
closing them, causing a resource leak; replace the comprehension with logic that
uses a helper (e.g., _open_image_file) to open string paths, append those open
file objects to a local list, and ensure all opened file handles are closed in a
finally block after the upload/usage completes; keep using
tensor_to_filelike(image) for non-string inputs and reference/modify the
variable files and the helper _open_image_file so every file opened is
explicitly closed even on exceptions.