Compare commits

..

1 Commits

Author SHA1 Message Date
a54504ba60 feat: add OAuth 2.1 + RFC 7591 DCR endpoints to openapi.yaml
Add the OAuth 2.1 authorization flow and RFC 7591 Dynamic Client
Registration endpoints to the shared spec, alongside the existing
auth-tagged operations (/api/auth/session, /api/auth/token,
/.well-known/jwks.json). All tagged x-runtime: [cloud] with a
[cloud-only] description prefix, following the established
convention for cloud-runtime-only operations.

Endpoints:

- GET  /.well-known/oauth-authorization-server  (RFC 8414 metadata)
- GET  /.well-known/oauth-protected-resource    (RFC 9728 metadata)
- GET  /oauth/authorize                         (consent challenge)
- POST /oauth/authorize                         (consent submission)
- POST /oauth/token                             (RFC 6749 §3.2)
- POST /oauth/register                          (RFC 7591 §3.1 DCR)

Component schemas added:

- OAuthAuthorizationServerMetadata
- OAuthProtectedResourceMetadata
- OAuthConsentChallenge, OAuthConsentChallengeWorkspace
- OAuthAuthorizeRedirectResponse
- OAuthTokenResponse, OAuthTokenError
- OAuthRegisterRequest, OAuthRegisterResponse, OAuthRegisterError

These endpoints are implemented in the cloud runtime today and
are called by browser frontends rendering the consent UI and by
MCP-spec-compliant clients (Claude Desktop, Cursor, etc.) doing
auto-discovery + self-registration. Documenting them in the
shared spec lets the cloud frontend generate types directly from
this spec instead of maintaining a parallel definition.

Spectral lints clean (0 errors). The hint-level findings on
OAuthTokenError / OAuthRegisterError ("standard error schema")
match the same hint on CloudError — these are protocol-specific
RFC-shaped errors, not generic application errors.
2026-05-20 21:05:26 -07:00
24 changed files with 168 additions and 1435 deletions

View File

@ -1,484 +0,0 @@
name: Backport Release
on:
workflow_dispatch:
inputs:
commit:
description: 'Full 40-char SHA of the tip commit of the backport source branch (the PR head commit that passed tests). The branch is resolved from this SHA and must be unique.'
required: true
type: string
permissions:
contents: read
pull-requests: read
checks: read
jobs:
backport-release:
name: Create backport release
runs-on: ubuntu-latest
environment: backport release
steps:
- name: Generate GitHub App token
id: app-token
uses: actions/create-github-app-token@bcd2ba49218906704ab6c1aa796996da409d3eb1
with:
app-id: ${{ secrets.FEN_RELEASE_APP_ID }}
private-key: ${{ secrets.FEN_RELEASE_PRIVATE_KEY }}
- name: Checkout repository
uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd
with:
token: ${{ steps.app-token.outputs.token }}
fetch-depth: 0
fetch-tags: true
- name: Configure git
run: |
git config user.name "fen-release[bot]"
git config user.email "fen-release[bot]@users.noreply.github.com"
- name: Resolve source branch from commit SHA
id: resolve
env:
SOURCE_COMMIT: ${{ inputs.commit }}
DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
run: |
set -euo pipefail
# Require a full 40-char lowercase-hex SHA. Short SHAs are ambiguous
# and we will be comparing this value against API responses (PR head
# SHA, ref tips) that always return the full form.
if [[ ! "${SOURCE_COMMIT}" =~ ^[0-9a-f]{40}$ ]]; then
echo "::error::Input commit '${SOURCE_COMMIT}' is not a full 40-char lowercase hex SHA."
exit 1
fi
# Fetch all remote branches so we can search for which one(s) point
# at this SHA. `actions/checkout` with fetch-depth: 0 fetches full
# history of the checked-out ref but does not necessarily populate
# every refs/remotes/origin/*, so do it explicitly.
git fetch --prune origin '+refs/heads/*:refs/remotes/origin/*'
# Verify the commit actually exists in this repo's object DB.
if ! git cat-file -e "${SOURCE_COMMIT}^{commit}" 2>/dev/null; then
echo "::error::Commit ${SOURCE_COMMIT} was not found in the repository."
exit 1
fi
# Find every remote branch whose tip == SOURCE_COMMIT. Exactly one
# branch must point at it. If zero, the commit isn't anyone's tip
# (likely stale, force-pushed past, or never the PR head). If more
# than one, the (branch -> SHA) mapping is ambiguous and we refuse
# to guess — the operator must give us a unique branch to release.
mapfile -t matching_branches < <(
git for-each-ref \
--format='%(refname:strip=3)' \
--points-at="${SOURCE_COMMIT}" \
refs/remotes/origin/ \
| grep -vx 'HEAD' || true
)
if [[ "${#matching_branches[@]}" -eq 0 ]]; then
echo "::error::No branch on origin has ${SOURCE_COMMIT} as its tip."
echo "::error::Either the branch was updated after you copied this SHA, or this commit was never the head of a branch."
exit 1
fi
if [[ "${#matching_branches[@]}" -gt 1 ]]; then
echo "::error::More than one branch on origin has ${SOURCE_COMMIT} as its tip; cannot pick one:"
for b in "${matching_branches[@]}"; do
echo "::error:: - ${b}"
done
echo "::error::Refusing to proceed with an ambiguous source branch."
exit 1
fi
source_branch="${matching_branches[0]}"
if [[ "${source_branch}" == "${DEFAULT_BRANCH}" ]]; then
echo "::error::Source branch must not be the default branch ('${DEFAULT_BRANCH}')."
exit 1
fi
echo "Resolved commit ${SOURCE_COMMIT} to branch '${source_branch}'."
echo "source_branch=${source_branch}" >> "$GITHUB_OUTPUT"
- name: Determine latest stable release
id: latest
env:
GH_TOKEN: ${{ steps.app-token.outputs.token }}
run: |
set -euo pipefail
# List all tags matching vMAJOR.MINOR.PATCH and pick the highest by numeric
# comparison of each component. We DO NOT use `sort -V` because it treats
# v0.19.99 as higher than v0.20.1.
latest_tag="$(
git tag --list 'v[0-9]*.[0-9]*.[0-9]*' \
| grep -E '^v[0-9]+\.[0-9]+\.[0-9]+$' \
| awk -F'[v.]' '{ printf "%010d %010d %010d %s\n", $2, $3, $4, $0 }' \
| sort -k1,1n -k2,2n -k3,3n \
| tail -n1 \
| awk '{print $4}'
)"
if [[ -z "${latest_tag}" ]]; then
echo "::error::No stable release tags (vMAJOR.MINOR.PATCH) were found."
exit 1
fi
# Parse components
ver="${latest_tag#v}"
major="${ver%%.*}"
rest="${ver#*.}"
minor="${rest%%.*}"
patch="${rest#*.}"
new_patch=$((patch + 1))
new_version="v${major}.${minor}.${new_patch}"
release_branch="release/v${major}.${minor}"
latest_sha="$(git rev-list -n 1 "refs/tags/${latest_tag}")"
echo "latest_tag=${latest_tag}" >> "$GITHUB_OUTPUT"
echo "latest_sha=${latest_sha}" >> "$GITHUB_OUTPUT"
echo "major=${major}" >> "$GITHUB_OUTPUT"
echo "minor=${minor}" >> "$GITHUB_OUTPUT"
echo "patch=${patch}" >> "$GITHUB_OUTPUT"
echo "new_version=${new_version}" >> "$GITHUB_OUTPUT"
echo "new_version_no_v=${major}.${minor}.${new_patch}" >> "$GITHUB_OUTPUT"
echo "release_branch=${release_branch}" >> "$GITHUB_OUTPUT"
echo "Latest stable release: ${latest_tag} (${latest_sha})"
echo "New version will be: ${new_version}"
echo "Release branch: ${release_branch}"
- name: Validate source branch is cut directly from the latest stable release
env:
SOURCE_BRANCH: ${{ steps.resolve.outputs.source_branch }}
SOURCE_COMMIT: ${{ inputs.commit }}
LATEST_TAG_SHA: ${{ steps.latest.outputs.latest_sha }}
LATEST_TAG: ${{ steps.latest.outputs.latest_tag }}
run: |
set -euo pipefail
# Use the user-provided SHA directly rather than re-resolving the branch
# tip — the resolve step already proved the branch tip equals SOURCE_COMMIT,
# and pinning to the SHA here makes the rest of the job TOCTOU-safe against
# someone pushing to the branch mid-run.
source_sha="${SOURCE_COMMIT}"
# Walking first-parent from the source tip must reach LATEST_TAG_SHA.
# We capture rev-list into a variable and grep against a here-string
# rather than piping `rev-list | grep -q`: under `set -o pipefail`,
# `grep -q` would exit on first match and SIGPIPE the still-streaming
# `rev-list`, propagating exit 141 as a spurious "not found".
first_parent_chain="$(git rev-list --first-parent "${source_sha}")"
if ! grep -Fxq "${LATEST_TAG_SHA}" <<< "${first_parent_chain}"; then
echo "::error::Source branch '${SOURCE_BRANCH}' is not cut from '${LATEST_TAG}'."
echo "::error::Its first-parent history does not include ${LATEST_TAG_SHA}."
exit 1
fi
# Additionally, every commit added on top of the tag (the set we are
# about to publish) must itself be a descendant of the tag along
# first-parent — i.e. no sibling commits from master sneak in via a
# non-first-parent path. Enforce by requiring that the symmetric
# difference is empty in one direction: commits in source that are
# NOT first-parent-reachable from source starting at the tag.
# We do this by intersecting:
# A = commits reachable from source but not from tag (full DAG)
# B = commits on the first-parent chain from source down to tag
# and requiring A == B.
all_added="$(git rev-list "${LATEST_TAG_SHA}..${source_sha}" | sort)"
first_parent_added="$(
git rev-list --first-parent "${LATEST_TAG_SHA}..${source_sha}" | sort
)"
if [[ "${all_added}" != "${first_parent_added}" ]]; then
echo "::error::Source branch '${SOURCE_BRANCH}' contains commits not on its first-parent chain from '${LATEST_TAG}'."
echo "::error::This usually means the branch was cut from master (not from the tag) or contains a merge from master."
echo "Commits reachable but not on first-parent chain:"
comm -23 <(printf '%s\n' "${all_added}") <(printf '%s\n' "${first_parent_added}") \
| while read -r sha; do
echo " $(git log -1 --format='%h %s' "${sha}")"
done
exit 1
fi
added_count="$(printf '%s\n' "${all_added}" | grep -c . || true)"
echo "Source branch is cut directly from ${LATEST_TAG} with ${added_count} commit(s) on top."
- name: Validate PR exists, is open, named correctly, has latest commit, and checks pass
env:
GH_TOKEN: ${{ steps.app-token.outputs.token }}
SOURCE_BRANCH: ${{ steps.resolve.outputs.source_branch }}
SOURCE_COMMIT: ${{ inputs.commit }}
NEW_VERSION: ${{ steps.latest.outputs.new_version }}
REPO: ${{ github.repository }}
run: |
set -euo pipefail
expected_title="ComfyUI backport release ${NEW_VERSION}"
# Find open PRs from this branch into master. The --state open filter
# is load-bearing: a closed/merged PR with passing checks must not be
# accepted as authorization for a new release.
pr_json="$(
gh pr list \
--repo "${REPO}" \
--state open \
--head "${SOURCE_BRANCH}" \
--base master \
--json number,title,headRefOid,state \
--limit 10
)"
pr_count="$(echo "${pr_json}" | jq 'length')"
if [[ "${pr_count}" -eq 0 ]]; then
echo "::error::No open PR found from '${SOURCE_BRANCH}' into 'master'. The PR must exist and be open."
exit 1
fi
# Pick the PR matching the expected title
pr_number="$(echo "${pr_json}" | jq -r --arg t "${expected_title}" '
map(select(.title == $t)) | .[0].number // empty
')"
pr_head_sha="$(echo "${pr_json}" | jq -r --arg t "${expected_title}" '
map(select(.title == $t)) | .[0].headRefOid // empty
')"
if [[ -z "${pr_number}" ]]; then
echo "::error::No open PR from '${SOURCE_BRANCH}' into 'master' is titled '${expected_title}'."
echo "Found PRs:"
echo "${pr_json}" | jq -r '.[] | " #\(.number): \(.title)"'
exit 1
fi
# The PR's current head commit must equal the SHA the operator gave us.
# This is what closes the door on releasing stale code: if anyone has
# pushed to the branch since the operator validated tests passed, the
# PR head will have advanced past SOURCE_COMMIT and we abort. (The
# resolve step already proved the branch tip == SOURCE_COMMIT; this
# ties that same SHA to the PR that authorizes the release.)
if [[ "${pr_head_sha}" != "${SOURCE_COMMIT}" ]]; then
echo "::error::PR #${pr_number} head commit is ${pr_head_sha}, but the operator-provided commit is ${SOURCE_COMMIT}."
echo "::error::The PR has new commits since this release was authorized. Re-run with the new head SHA after verifying its checks."
exit 1
fi
echo "Found open PR #${pr_number} titled '${expected_title}' at head ${pr_head_sha} (matches operator-provided commit)."
# Verify all check runs on the head commit have completed successfully.
# A check is considered passing if conclusion is success, neutral, or skipped.
checks_json="$(
gh api \
--paginate \
"repos/${REPO}/commits/${pr_head_sha}/check-runs" \
--jq '.check_runs[] | {name: .name, status: .status, conclusion: .conclusion}'
)"
if [[ -z "${checks_json}" ]]; then
echo "::error::No check runs found on PR head commit ${pr_head_sha}."
exit 1
fi
echo "Check runs on ${pr_head_sha}:"
echo "${checks_json}" | jq -s '.'
failing="$(echo "${checks_json}" | jq -s '
map(select(
.status != "completed"
or (.conclusion as $c
| ["success","neutral","skipped"]
| index($c) | not)
))
')"
failing_count="$(echo "${failing}" | jq 'length')"
if [[ "${failing_count}" -gt 0 ]]; then
echo "::error::One or more checks have not passed on PR head commit ${pr_head_sha}:"
echo "${failing}" | jq -r '.[] | " - \(.name): status=\(.status) conclusion=\(.conclusion)"'
exit 1
fi
echo "All checks have passed on ${pr_head_sha}."
- name: Prepare release branch
id: prepare
env:
GH_TOKEN: ${{ steps.app-token.outputs.token }}
REPO: ${{ github.repository }}
RELEASE_BRANCH: ${{ steps.latest.outputs.release_branch }}
LATEST_TAG: ${{ steps.latest.outputs.latest_tag }}
LATEST_TAG_SHA: ${{ steps.latest.outputs.latest_sha }}
PATCH: ${{ steps.latest.outputs.patch }}
run: |
set -euo pipefail
# Try to fetch the release branch. If patch == 0, it shouldn't exist yet
# and we'll create it from the latest stable tag. If patch > 0, it must
# already exist and its tip must equal the latest stable tag commit (i.e.
# the previous patch release).
if git ls-remote --exit-code --heads origin "${RELEASE_BRANCH}" >/dev/null 2>&1; then
echo "Release branch '${RELEASE_BRANCH}' already exists on origin."
git fetch origin "refs/heads/${RELEASE_BRANCH}:refs/remotes/origin/${RELEASE_BRANCH}"
git checkout -B "${RELEASE_BRANCH}" "refs/remotes/origin/${RELEASE_BRANCH}"
current_tip="$(git rev-parse HEAD)"
if [[ "${current_tip}" != "${LATEST_TAG_SHA}" ]]; then
echo "::error::Release branch '${RELEASE_BRANCH}' tip (${current_tip}) is not at the latest stable release '${LATEST_TAG}' (${LATEST_TAG_SHA})."
echo "::error::Refusing to release on top of a divergent branch."
exit 1
fi
echo "branch_existed=true" >> "$GITHUB_OUTPUT"
else
if [[ "${PATCH}" != "0" ]]; then
echo "::error::Release branch '${RELEASE_BRANCH}' does not exist on origin, but the latest stable release '${LATEST_TAG}' has patch=${PATCH} (>0). This is inconsistent."
exit 1
fi
echo "Release branch '${RELEASE_BRANCH}' does not exist. Creating from ${LATEST_TAG}."
git checkout -B "${RELEASE_BRANCH}" "refs/tags/${LATEST_TAG}"
echo "branch_existed=false" >> "$GITHUB_OUTPUT"
fi
- name: Fast-forward merge source branch into release branch
env:
SOURCE_BRANCH: ${{ steps.resolve.outputs.source_branch }}
SOURCE_COMMIT: ${{ inputs.commit }}
RELEASE_BRANCH: ${{ steps.latest.outputs.release_branch }}
run: |
set -euo pipefail
# --ff-only guarantees no merge commit is created. If a fast-forward is
# not possible (i.e. the release branch has commits the source branch
# doesn't), the merge will fail and we abort. Because we already validated
# that the source branch is rooted on the latest stable tag, and the
# release branch tip equals that same tag, this fast-forward should
# always succeed for a well-formed backport branch.
#
# We merge the operator-provided SHA, not the branch ref, so a push to
# the branch in the window between resolve and now cannot smuggle new
# commits into the release.
if ! git merge --ff-only "${SOURCE_COMMIT}"; then
echo "::error::Cannot fast-forward '${RELEASE_BRANCH}' to ${SOURCE_COMMIT} (tip of '${SOURCE_BRANCH}'). A merge commit would be required. Aborting."
exit 1
fi
echo "Fast-forwarded '${RELEASE_BRANCH}' to ${SOURCE_COMMIT} (tip of '${SOURCE_BRANCH}')."
- name: Bump version files
env:
NEW_VERSION_NO_V: ${{ steps.latest.outputs.new_version_no_v }}
run: |
set -euo pipefail
if [[ ! -f comfyui_version.py ]]; then
echo "::error::comfyui_version.py not found in repo root."
exit 1
fi
if [[ ! -f pyproject.toml ]]; then
echo "::error::pyproject.toml not found in repo root."
exit 1
fi
# Replace the version string in comfyui_version.py.
# Expected format: __version__ = "X.Y.Z"
python3 - "$NEW_VERSION_NO_V" <<'PY'
import re, sys, pathlib
new = sys.argv[1]
p = pathlib.Path("comfyui_version.py")
src = p.read_text()
new_src, n = re.subn(
r'(__version__\s*=\s*[\'"])[^\'"]+([\'"])',
lambda m: f'{m.group(1)}{new}{m.group(2)}',
src,
count=1,
)
if n != 1:
sys.exit("Could not find __version__ assignment in comfyui_version.py")
p.write_text(new_src)
p = pathlib.Path("pyproject.toml")
src = p.read_text()
# Replace the first `version = "..."` inside [project] or [tool.poetry].
new_src, n = re.subn(
r'(?m)^(version\s*=\s*")[^"]+(")',
lambda m: f'{m.group(1)}{new}{m.group(2)}',
src,
count=1,
)
if n != 1:
sys.exit("Could not find version assignment in pyproject.toml")
p.write_text(new_src)
PY
echo "Updated version to ${NEW_VERSION_NO_V} in comfyui_version.py and pyproject.toml."
git --no-pager diff -- comfyui_version.py pyproject.toml
- name: Commit version bump and tag release
env:
NEW_VERSION: ${{ steps.latest.outputs.new_version }}
run: |
set -euo pipefail
git add comfyui_version.py pyproject.toml
git commit -m "ComfyUI ${NEW_VERSION}"
if git rev-parse -q --verify "refs/tags/${NEW_VERSION}" >/dev/null; then
echo "::error::Tag ${NEW_VERSION} already exists locally."
exit 1
fi
git tag "${NEW_VERSION}"
- name: Verify tag does not already exist on origin
env:
NEW_VERSION: ${{ steps.latest.outputs.new_version }}
run: |
set -euo pipefail
if git ls-remote --exit-code --tags origin "refs/tags/${NEW_VERSION}" >/dev/null 2>&1; then
echo "::error::Tag ${NEW_VERSION} already exists on origin. Aborting."
exit 1
fi
- name: Push release branch and tag
env:
RELEASE_BRANCH: ${{ steps.latest.outputs.release_branch }}
NEW_VERSION: ${{ steps.latest.outputs.new_version }}
run: |
set -euo pipefail
# Push the branch first, then the tag. Atomic-ish: if the branch push
# fails we never publish the tag.
git push origin "refs/heads/${RELEASE_BRANCH}:refs/heads/${RELEASE_BRANCH}"
git push origin "refs/tags/${NEW_VERSION}"
echo "Released ${NEW_VERSION} on ${RELEASE_BRANCH}."
- name: Summary
if: always()
env:
NEW_VERSION: ${{ steps.latest.outputs.new_version }}
RELEASE_BRANCH: ${{ steps.latest.outputs.release_branch }}
LATEST_TAG: ${{ steps.latest.outputs.latest_tag }}
SOURCE_BRANCH: ${{ steps.resolve.outputs.source_branch }}
SOURCE_COMMIT: ${{ inputs.commit }}
run: |
# SOURCE_BRANCH is empty if the resolve step never produced an output
# (e.g. the workflow failed in or before that step). Show a placeholder
# in that case so the summary table still renders cleanly.
source_branch_display="${SOURCE_BRANCH:-(unresolved)}"
{
echo "## Backport release"
echo ""
echo "| Field | Value |"
echo "|---|---|"
echo "| Source commit | \`${SOURCE_COMMIT}\` |"
echo "| Source branch | \`${source_branch_display}\` |"
echo "| Previous stable | \`${LATEST_TAG}\` |"
echo "| New version | \`${NEW_VERSION}\` |"
echo "| Release branch | \`${RELEASE_BRANCH}\` |"
} >> "$GITHUB_STEP_SUMMARY"

View File

@ -20,7 +20,7 @@
[website-url]: https://www.comfy.org/
<!-- Workaround to display total user from https://github.com/badges/shields/issues/4500#issuecomment-2060079995 -->
[discord-shield]: https://img.shields.io/badge/dynamic/json?url=https%3A%2F%2Fdiscord.com%2Fapi%2Finvites%2Fcomfyorg%3Fwith_counts%3Dtrue&query=%24.approximate_member_count&logo=discord&logoColor=white&label=Discord&color=green&suffix=%20total
[discord-url]: https://discord.com/invite/comfyorg
[discord-url]: https://www.comfy.org/discord
[twitter-shield]: https://img.shields.io/twitter/follow/ComfyUI
[twitter-url]: https://x.com/ComfyUI

View File

@ -62,8 +62,6 @@ def get_comfy_package_versions():
def check_comfy_packages_versions():
"""Warn for every comfy* package whose installed version is below requirements.txt."""
from packaging.version import InvalidVersion, parse as parse_pep440
outdated_packages = []
for pkg in get_comfy_package_versions():
installed_str = pkg["installed"]
required_str = pkg["required"]
@ -75,26 +73,19 @@ def check_comfy_packages_versions():
logging.error(f"Failed to check {pkg['name']} version: {e}")
continue
if outdated:
outdated_packages.append((pkg["name"], installed_str, required_str))
else:
logging.info("{} version: {}".format(pkg["name"], installed_str))
if outdated_packages:
package_warnings = "\n".join(
f"Installed {name} version {installed} is lower than the recommended version {required}."
for name, installed, required in outdated_packages
)
app.logger.log_startup_warning(
f"""
app.logger.log_startup_warning(
f"""
________________________________________________________________________
WARNING WARNING WARNING WARNING WARNING
{package_warnings}
Installed {pkg["name"]} version {installed_str} is lower than the recommended version {required_str}.
{get_missing_requirements_message()}
________________________________________________________________________
""".strip()
)
)
else:
logging.info("{} version: {}".format(pkg["name"], installed_str))
REQUEST_TIMEOUT = 10 # seconds

View File

@ -1613,16 +1613,6 @@ class ModelPatcherDynamic(ModelPatcher):
#use all ModelPatcherDynamic this is ignored and its all done dynamically.
return super().memory_required(input_shape=input_shape) * 1.3 + (1024 ** 3)
def restore_loaded_backups(self):
restored = self.model.model_loaded_weight_memory
for key in list(self.backup.keys()):
bk = self.backup.pop(key)
comfy.utils.set_attr_param(self.model, key, bk.weight)
for key in list(self.backup_buffers.keys()):
comfy.utils.set_attr_buffer(self.model, key, self.backup_buffers.pop(key))
self.model.model_loaded_weight_memory = 0
return restored
def load(self, device_to=None, lowvram_model_memory=0, force_patch_weights=False, full_load=False, dirty=False):
@ -1639,7 +1629,7 @@ class ModelPatcherDynamic(ModelPatcher):
num_patches = 0
allocated_size = 0
self.restore_loaded_backups()
self.model.model_loaded_weight_memory = 0
with self.use_ejected():
self.unpatch_hooks()
@ -1726,9 +1716,6 @@ class ModelPatcherDynamic(ModelPatcher):
force_load=True
if force_load:
if hasattr(m, "_v"):
comfy_aimdo.model_vbar.vbar_unpin(m._v)
delattr(m, "_v")
force_load_param(self, "weight", device_to)
force_load_param(self, "bias", device_to)
else:
@ -1786,7 +1773,13 @@ class ModelPatcherDynamic(ModelPatcher):
freed = 0 if vbar is None else vbar.free_memory(memory_to_free)
if freed < memory_to_free:
freed += self.restore_loaded_backups()
for key in list(self.backup.keys()):
bk = self.backup.pop(key)
comfy.utils.set_attr_param(self.model, key, bk.weight)
for key in list(self.backup_buffers.keys()):
comfy.utils.set_attr_buffer(self.model, key, self.backup_buffers.pop(key))
freed += self.model.model_loaded_weight_memory
self.model.model_loaded_weight_memory = 0
return freed

View File

@ -1019,11 +1019,10 @@ def bislerp(samples, width, height):
def lanczos(samples, width, height):
#the below API is strict and expects grayscale to be squeezed
if samples.ndim == 4:
samples = samples.squeeze(1) if samples.shape[1] == 1 else samples.movedim(1, -1)
samples = samples.squeeze(1) if samples.shape[1] == 1 else samples.movedim(1, -1)
images = [Image.fromarray(np.clip(255. * image.cpu().numpy(), 0, 255).astype(np.uint8)) for image in samples]
images = [image.resize((width, height), resample=Image.Resampling.LANCZOS) for image in images]
images = [torch.from_numpy(t).movedim(-1, 0) if (t := np.array(image).astype(np.float32) / 255.0).ndim == 3 else torch.from_numpy(t) for image in images]
images = [torch.from_numpy(np.array(image).astype(np.float32) / 255.0).movedim(-1, 0) for image in images]
result = torch.stack(images)
return result.to(samples.device, samples.dtype)

View File

@ -35,19 +35,6 @@ class AnthropicMessage(BaseModel):
content: list[AnthropicTextContent | AnthropicImageContent] = Field(...)
class AnthropicThinkingConfig(BaseModel):
type: Literal["enabled", "disabled", "adaptive"] = Field(...)
budget_tokens: int | None = Field(
None, ge=1024,
description="Reasoning budget in tokens. Used when type is 'enabled'. Must be less than max_tokens.",
)
class AnthropicOutputConfig(BaseModel):
"""Used with `thinking.type='adaptive'` on models like Opus 4.7."""
effort: Literal["low", "medium", "high"] | None = Field(None)
class AnthropicMessagesRequest(BaseModel):
model: str = Field(...)
messages: list[AnthropicMessage] = Field(...)
@ -57,8 +44,6 @@ class AnthropicMessagesRequest(BaseModel):
top_p: float | None = Field(None, ge=0.0, le=1.0)
top_k: int | None = Field(None, ge=0)
stop_sequences: list[str] | None = Field(None)
thinking: AnthropicThinkingConfig | None = Field(None)
output_config: AnthropicOutputConfig | None = Field(None)
class AnthropicResponseTextBlock(BaseModel):
@ -66,14 +51,6 @@ class AnthropicResponseTextBlock(BaseModel):
text: str = Field(...)
class AnthropicResponseThinkingBlock(BaseModel):
type: Literal["thinking"] = "thinking"
thinking: str = Field(...)
AnthropicResponseBlock = AnthropicResponseTextBlock | AnthropicResponseThinkingBlock
class AnthropicCacheCreationUsage(BaseModel):
ephemeral_5m_input_tokens: int | None = Field(None)
ephemeral_1h_input_tokens: int | None = Field(None)
@ -92,7 +69,7 @@ class AnthropicMessagesResponse(BaseModel):
type: str | None = Field(None)
role: str | None = Field(None)
model: str | None = Field(None)
content: list[AnthropicResponseBlock] | None = Field(None)
content: list[AnthropicResponseTextBlock] | None = Field(None)
stop_reason: str | None = Field(None)
stop_sequence: str | None = Field(None)
usage: AnthropicMessagesUsage | None = Field(None)

View File

@ -1,93 +0,0 @@
"""Pydantic models for the OpenRouter chat completions API.
See: https://openrouter.ai/docs/api/api-reference/chat/send-chat-completion-request
"""
from typing import Literal
from pydantic import BaseModel, Field
class OpenRouterTextContent(BaseModel):
type: Literal["text"] = "text"
text: str = Field(...)
class OpenRouterImageUrl(BaseModel):
url: str = Field(...)
class OpenRouterImageContent(BaseModel):
type: Literal["image_url"] = "image_url"
image_url: OpenRouterImageUrl = Field(...)
class OpenRouterVideoUrl(BaseModel):
url: str = Field(...)
class OpenRouterVideoContent(BaseModel):
type: Literal["video_url"] = "video_url"
video_url: OpenRouterVideoUrl = Field(...)
OpenRouterContentBlock = OpenRouterTextContent | OpenRouterImageContent | OpenRouterVideoContent
class OpenRouterMessage(BaseModel):
role: Literal["system", "user", "assistant"] = Field(...)
content: str | list[OpenRouterContentBlock] = Field(...)
class OpenRouterReasoningConfig(BaseModel):
effort: str | None = Field(None)
exclude: bool | None = Field(None, description="If true, model reasons but reasoning is excluded from response.")
class OpenRouterWebSearchOptions(BaseModel):
search_context_size: str | None = Field(None)
class OpenRouterChatRequest(BaseModel):
model: str = Field(...)
messages: list[OpenRouterMessage] = Field(...)
seed: int | None = Field(None)
reasoning: OpenRouterReasoningConfig | None = Field(None)
web_search_options: OpenRouterWebSearchOptions | None = Field(None)
stream: bool = Field(False)
class OpenRouterUsage(BaseModel):
prompt_tokens: int | None = Field(None)
completion_tokens: int | None = Field(None)
total_tokens: int | None = Field(None)
cost: float | None = Field(None, description="Server-side authoritative USD cost of the call.")
class OpenRouterResponseMessage(BaseModel):
role: str | None = Field(None)
content: str | None = Field(None)
reasoning: str | None = Field(None)
refusal: str | None = Field(None)
class OpenRouterChoice(BaseModel):
index: int | None = Field(None)
message: OpenRouterResponseMessage | None = Field(None)
finish_reason: str | None = Field(None)
class OpenRouterError(BaseModel):
code: int | str | None = Field(None)
message: str | None = Field(None)
metadata: dict | None = Field(None)
class OpenRouterChatResponse(BaseModel):
id: str | None = Field(None)
model: str | None = Field(None)
object: str | None = Field(None)
provider: str | None = Field(None)
choices: list[OpenRouterChoice] | None = Field(None)
usage: OpenRouterUsage | None = Field(None)
error: OpenRouterError | None = Field(None)

View File

@ -9,11 +9,8 @@ from comfy_api_nodes.apis.anthropic import (
AnthropicMessage,
AnthropicMessagesRequest,
AnthropicMessagesResponse,
AnthropicOutputConfig,
AnthropicResponseTextBlock,
AnthropicRole,
AnthropicTextContent,
AnthropicThinkingConfig,
)
from comfy_api_nodes.util import (
ApiEndpoint,
@ -35,29 +32,15 @@ CLAUDE_MODELS: dict[str, str] = {
"Haiku 4.5": "claude-haiku-4-5-20251001",
}
_THINKING_UNSUPPORTED = {"Haiku 4.5"}
# Models that use the newer "adaptive" thinking mode (Opus 4.7 requires it; older models keep the explicit budget API).
# Anthropic decides the actual budget when adaptive is used, based on the `output_config.effort` hint.
_ADAPTIVE_THINKING_MODELS = {"Opus 4.7", "Opus 4.6", "Sonnet 4.6"}
# Budget mode (Sonnet 4.5): effort -> reasoning budget in tokens. Must be < max_tokens.
# Sized so even the "high" budget fits comfortably under the default max_tokens=32768.
_REASONING_BUDGET: dict[str, int] = {
"low": 2048,
"medium": 8192,
"high": 16384,
}
_REASONING_EFFORTS = ["off", "low", "medium", "high"]
def _claude_model_inputs(model_label: str):
inputs: list = [
def _claude_model_inputs():
return [
IO.Int.Input(
"max_tokens",
default=32768,
min=4096,
max=64000,
tooltip="Maximum number of tokens to generate (includes reasoning tokens when enabled).",
default=16000,
min=32,
max=32000,
tooltip="Maximum number of tokens to generate before stopping.",
advanced=True,
),
IO.Float.Input(
@ -66,24 +49,10 @@ def _claude_model_inputs(model_label: str):
min=0.0,
max=1.0,
step=0.01,
tooltip=(
"Controls randomness. 0.0 is deterministic, 1.0 is most random. "
"Ignored for Opus 4.7 and any model when reasoning_effort is set."
),
tooltip="Controls randomness. 0.0 is deterministic, 1.0 is most random. Ignored for Opus 4.7.",
advanced=True,
),
]
if model_label not in _THINKING_UNSUPPORTED:
inputs.append(
IO.Combo.Input(
"reasoning_effort",
options=_REASONING_EFFORTS,
default="off",
tooltip="Extended thinking effort. 'off' disables reasoning.",
advanced=True,
)
)
return inputs
def _model_price_per_million(model: str) -> tuple[float, float] | None:
@ -126,11 +95,7 @@ def calculate_tokens_price(response: AnthropicMessagesResponse) -> float | None:
def _get_text_from_response(response: AnthropicMessagesResponse) -> str:
if not response.content:
return ""
# Thinking blocks are silently dropped — we never want reasoning in the output.
return "\n".join(
block.text for block in response.content
if isinstance(block, AnthropicResponseTextBlock) and block.text
)
return "\n".join(block.text for block in response.content if block.text)
async def _build_image_content_blocks(
@ -168,10 +133,7 @@ class ClaudeNode(IO.ComfyNode):
),
IO.DynamicCombo.Input(
"model",
options=[
IO.DynamicCombo.Option(label, _claude_model_inputs(label))
for label in CLAUDE_MODELS
],
options=[IO.DynamicCombo.Option(label, _claude_model_inputs()) for label in CLAUDE_MODELS],
tooltip="The Claude model used to generate the response.",
),
IO.Int.Input(
@ -245,29 +207,8 @@ class ClaudeNode(IO.ComfyNode):
) -> IO.NodeOutput:
validate_string(prompt, strip_whitespace=True, min_length=1)
model_label = model["model"]
max_tokens = model.get("max_tokens", 32768)
reasoning_effort = model.get("reasoning_effort", "off")
thinking_enabled = reasoning_effort not in ("off", None) and model_label not in _THINKING_UNSUPPORTED
# Anthropic requires temperature to be unset (defaults to 1.0) when thinking is enabled.
# Opus 4.7 also rejects user-supplied temperature.
if thinking_enabled or model_label == "Opus 4.7":
temperature = None
else:
temperature = model.get("temperature", 1.0)
thinking_cfg: AnthropicThinkingConfig | None = None
output_cfg: AnthropicOutputConfig | None = None
if thinking_enabled:
if model_label in _ADAPTIVE_THINKING_MODELS:
# Adaptive mode - Anthropic chooses the budget based on effort hint
thinking_cfg = AnthropicThinkingConfig(type="adaptive")
output_cfg = AnthropicOutputConfig(effort=reasoning_effort)
else:
# Budget mode (Sonnet 4.5). Leave at least 1024 tokens for the actual response
budget = _REASONING_BUDGET[reasoning_effort]
budget = min(budget, max(1024, max_tokens - 1024))
thinking_cfg = AnthropicThinkingConfig(type="enabled", budget_tokens=budget)
max_tokens = model["max_tokens"]
temperature = None if model_label == "Opus 4.7" else model["temperature"]
image_tensors: list[Input.Image] = [t for t in (images or {}).values() if t is not None]
if sum(get_number_of_images(t) for t in image_tensors) > CLAUDE_MAX_IMAGES:
@ -288,8 +229,6 @@ class ClaudeNode(IO.ComfyNode):
messages=[AnthropicMessage(role=AnthropicRole.user, content=content)],
system=system_prompt or None,
temperature=temperature,
thinking=thinking_cfg,
output_config=output_cfg,
),
price_extractor=calculate_tokens_price,
)

View File

@ -43,16 +43,15 @@ from comfy_api_nodes.util import (
ApiEndpoint,
download_url_to_image_tensor,
download_url_to_video_output,
downscale_video_to_max_pixels,
get_number_of_images,
image_tensor_pair_to_batch,
poll_op,
resize_video_to_pixel_budget,
sync_op,
upload_audio_to_comfyapi,
upload_image_to_comfyapi,
upload_images_to_comfyapi,
upload_video_to_comfyapi,
upscale_video_to_min_pixels,
validate_image_aspect_ratio,
validate_image_dimensions,
validate_string,
@ -111,13 +110,12 @@ def _validate_ref_video_pixels(video: Input.Video, model_id: str, resolution: st
max_px = limits.get("max")
if min_px and pixels < min_px:
raise ValueError(
f"Reference video {index} is too small: {w}x{h} = {pixels:,} total pixels. "
f"Minimum for this model is {min_px:,} total pixels."
f"Reference video {index} is too small: {w}x{h} = {pixels:,}px. " f"Minimum is {min_px:,}px for this model."
)
if max_px and pixels > max_px:
raise ValueError(
f"Reference video {index} is too large: {w}x{h} = {pixels:,} total pixels. "
f"Maximum for this model is {max_px:,} total pixels. Try downscaling the video."
f"Reference video {index} is too large: {w}x{h} = {pixels:,}px. "
f"Maximum is {max_px:,}px for this model. Try downscaling the video."
)
@ -1678,14 +1676,14 @@ class ByteDance2FirstLastFrameNode(IO.ComfyNode):
"first_frame_asset_id",
default="",
tooltip="Seedance asset_id to use as the first frame. "
"Mutually exclusive with the first_frame image input.",
"Mutually exclusive with the first_frame image input.",
optional=True,
),
IO.String.Input(
"last_frame_asset_id",
default="",
tooltip="Seedance asset_id to use as the last frame. "
"Mutually exclusive with the last_frame image input.",
"Mutually exclusive with the last_frame image input.",
optional=True,
),
IO.Int.Input(
@ -1867,20 +1865,11 @@ def _seedance2_reference_inputs(resolutions: list[str], default_ratio: str = "16
IO.Boolean.Input(
"auto_downscale",
default=False,
advanced=True,
optional=True,
tooltip="Automatically downscale reference videos that exceed the model's pixel budget "
"for the selected resolution. Aspect ratio is preserved; videos already within limits are untouched.",
),
IO.Boolean.Input(
"auto_upscale",
default=False,
advanced=True,
optional=True,
tooltip="Automatically upscale reference videos that are below the model's minimum pixel count "
"for the selected resolution. Aspect ratio is preserved; videos already meeting the minimum are "
"untouched. Note: upscaling a low-resolution source does not add real detail and may produce "
"lower-quality generations.",
),
IO.Autogrow.Input(
"reference_assets",
template=IO.Autogrow.TemplateNames(
@ -2041,13 +2030,7 @@ class ByteDance2ReferenceNode(IO.ComfyNode):
max_px = SEEDANCE2_REF_VIDEO_PIXEL_LIMITS.get(model_id, {}).get(model["resolution"], {}).get("max")
if max_px:
for key in reference_videos:
reference_videos[key] = downscale_video_to_max_pixels(reference_videos[key], max_px)
if model.get("auto_upscale") and reference_videos:
min_px = SEEDANCE2_REF_VIDEO_PIXEL_LIMITS.get(model_id, {}).get(model["resolution"], {}).get("min")
if min_px:
for key in reference_videos:
reference_videos[key] = upscale_video_to_min_pixels(reference_videos[key], min_px)
reference_videos[key] = resize_video_to_pixel_budget(reference_videos[key], max_px)
total_video_duration = 0.0
for i, key in enumerate(reference_videos, 1):

View File

@ -1,374 +0,0 @@
"""API Nodes for OpenRouter LLM chat completions."""
from dataclasses import dataclass
from typing import Literal
from typing_extensions import override
from comfy_api.latest import IO, ComfyExtension, Input
from comfy_api_nodes.apis.openrouter import (
OpenRouterChatRequest,
OpenRouterChatResponse,
OpenRouterContentBlock,
OpenRouterImageContent,
OpenRouterImageUrl,
OpenRouterMessage,
OpenRouterReasoningConfig,
OpenRouterTextContent,
OpenRouterVideoContent,
OpenRouterVideoUrl,
OpenRouterWebSearchOptions,
)
from comfy_api_nodes.util import (
ApiEndpoint,
get_number_of_images,
sync_op,
upload_images_to_comfyapi,
upload_video_to_comfyapi,
validate_string,
)
OPENROUTER_CHAT_ENDPOINT = "/proxy/openrouter/api/v1/chat/completions"
Profile = Literal["standard", "reasoning", "frontier_reasoning", "perplexity", "perplexity_reasoning"]
@dataclass(frozen=True)
class _ModelSpec:
slug: str # exact OpenRouter model id
profile: Profile
price_in: float # USD per token (prompt)
price_out: float # USD per token (completion)
max_images: int = 0 # 0 = no image input; otherwise max URL-passed images supported
max_videos: int = 0 # 0 = no video input; otherwise max URL-passed videos supported
MODELS: list[_ModelSpec] = [
_ModelSpec("anthropic/claude-opus-4.7", "frontier_reasoning", 0.000005, 0.000025, max_images=20),
_ModelSpec("openai/gpt-5.5-pro", "frontier_reasoning", 0.00003, 0.00018, max_images=20),
_ModelSpec("openai/gpt-5.5", "frontier_reasoning", 0.000005, 0.00003, max_images=20),
_ModelSpec("google/gemini-3.5-flash", "reasoning", 0.0000015, 0.000009, max_images=20, max_videos=4),
_ModelSpec("x-ai/grok-4.20", "reasoning", 0.00000125, 0.0000025, max_images=20),
_ModelSpec("x-ai/grok-4.3", "reasoning", 0.00000125, 0.0000025, max_images=20),
_ModelSpec("deepseek/deepseek-v4-pro", "reasoning", 0.000000435, 0.00000087),
_ModelSpec("deepseek/deepseek-v4-flash", "reasoning", 0.000000112, 0.000000224),
_ModelSpec("deepseek/deepseek-v3.2", "reasoning", 0.000000252, 0.000000378),
_ModelSpec("qwen/qwen3.6-max-preview", "reasoning", 0.00000104, 0.00000624),
_ModelSpec("qwen/qwen3.6-plus", "reasoning", 0.000000325, 0.00000195, max_images=10, max_videos=4),
_ModelSpec("qwen/qwen3.6-flash", "reasoning", 0.0000001875, 0.000001125, max_images=10, max_videos=4),
_ModelSpec("mistralai/mistral-large-2512", "standard", 0.0000005, 0.0000015, max_images=8),
_ModelSpec("mistralai/mistral-medium-3-5", "reasoning", 0.0000015, 0.0000075, max_images=8),
_ModelSpec("z-ai/glm-4.6", "reasoning", 0.00000043, 0.00000174),
_ModelSpec("z-ai/glm-5", "reasoning", 0.0000006, 0.00000192),
_ModelSpec("moonshotai/kimi-k2.6", "reasoning", 0.00000073, 0.00000349, max_images=10),
_ModelSpec("moonshotai/kimi-k2-thinking", "reasoning", 0.0000006, 0.0000025),
_ModelSpec("perplexity/sonar-pro", "perplexity", 0.000003, 0.000015),
_ModelSpec("perplexity/sonar-reasoning-pro", "perplexity_reasoning", 0.000002, 0.000008),
_ModelSpec("perplexity/sonar-deep-research", "perplexity_reasoning", 0.000002, 0.000008),
]
_MODELS_BY_SLUG: dict[str, _ModelSpec] = {m.slug: m for m in MODELS}
_REASONING_EFFORTS = ["off", "low", "medium", "high"]
_SEARCH_CONTEXT_SIZES = ["low", "medium", "high"]
def _reasoning_extra_inputs() -> list:
return [
IO.Combo.Input(
"reasoning_effort",
options=_REASONING_EFFORTS,
default="off",
tooltip="Reasoning effort. 'off' disables reasoning entirely.",
advanced=True,
),
]
def _perplexity_extra_inputs() -> list:
return [
IO.Combo.Input(
"search_context_size",
options=_SEARCH_CONTEXT_SIZES,
default="medium",
tooltip="How much web search context to retrieve. Larger = more grounded but slower/pricier.",
advanced=True,
),
]
def _profile_inputs(profile: Profile) -> list:
if profile == "standard":
return []
if profile in ("reasoning", "frontier_reasoning"):
return _reasoning_extra_inputs()
if profile == "perplexity":
return _perplexity_extra_inputs()
if profile == "perplexity_reasoning":
return _perplexity_extra_inputs() + _reasoning_extra_inputs()
raise ValueError(f"Unknown profile: {profile}")
def _media_inputs(spec: _ModelSpec) -> list:
extras: list = []
if spec.max_images > 0:
extras.append(
IO.Autogrow.Input(
"images",
template=IO.Autogrow.TemplateNames(
IO.Image.Input("image"),
names=[f"image_{i}" for i in range(1, spec.max_images + 1)],
min=0,
),
tooltip=f"Optional reference image(s) — up to {spec.max_images}. Sent as URLs.",
)
)
if spec.max_videos > 0:
extras.append(
IO.Autogrow.Input(
"videos",
template=IO.Autogrow.TemplateNames(
IO.Video.Input("video"),
names=[f"video_{i}" for i in range(1, spec.max_videos + 1)],
min=0,
),
tooltip=f"Optional reference video(s) — up to {spec.max_videos}. Sent as URLs.",
)
)
return extras
def _inputs_for_model(spec: _ModelSpec) -> list:
return _profile_inputs(spec.profile) + _media_inputs(spec)
def _build_model_options() -> list[IO.DynamicCombo.Option]:
return [IO.DynamicCombo.Option(spec.slug, _inputs_for_model(spec)) for spec in MODELS]
def _calculate_price(response: OpenRouterChatResponse) -> float | None:
if response.usage and response.usage.cost is not None:
return float(response.usage.cost)
return None
def _price_badge_jsonata() -> str:
rates_pairs = []
for spec in MODELS:
prompt_per_1k = spec.price_in * 1000
completion_per_1k = spec.price_out * 1000
rates_pairs.append(f' "{spec.slug}": [{prompt_per_1k:.8g}, {completion_per_1k:.8g}]')
rates_block = ",\n".join(rates_pairs)
return (
"(\n"
" $rates := {\n"
f"{rates_block}\n"
" };\n"
" $r := $lookup($rates, widgets.model);\n"
" $r ? {\n"
' "type": "list_usd",\n'
' "usd": $r,\n'
' "format": { "approximate": true, "separator": "-", "suffix": " per 1K tokens" }\n'
' } : {"type": "text", "text": "Token-based"}\n'
")"
)
async def _build_image_blocks(
cls: type[IO.ComfyNode], spec: _ModelSpec, images: list[Input.Image]
) -> list[OpenRouterImageContent]:
urls = await upload_images_to_comfyapi(
cls,
images,
max_images=spec.max_images,
total_pixels=2048 * 2048,
mime_type="image/png",
wait_label="Uploading reference images",
)
return [OpenRouterImageContent(image_url=OpenRouterImageUrl(url=url)) for url in urls]
async def _build_video_blocks(cls: type[IO.ComfyNode], videos: list[Input.Video]) -> list[OpenRouterVideoContent]:
blocks: list[OpenRouterVideoContent] = []
total = len(videos)
for idx, video in enumerate(videos):
label = "Uploading reference video"
if total > 1:
label = f"{label} ({idx + 1}/{total})"
url = await upload_video_to_comfyapi(cls, video, wait_label=label)
blocks.append(OpenRouterVideoContent(video_url=OpenRouterVideoUrl(url=url)))
return blocks
def _user_message(prompt: str, media_blocks: list[OpenRouterContentBlock]) -> OpenRouterMessage:
if not media_blocks:
return OpenRouterMessage(role="user", content=prompt)
blocks: list[OpenRouterContentBlock] = list(media_blocks)
blocks.append(OpenRouterTextContent(text=prompt))
return OpenRouterMessage(role="user", content=blocks)
def _build_messages(
system_prompt: str, prompt: str, media_blocks: list[OpenRouterContentBlock]
) -> list[OpenRouterMessage]:
messages: list[OpenRouterMessage] = []
if system_prompt:
messages.append(OpenRouterMessage(role="system", content=system_prompt))
messages.append(_user_message(prompt, media_blocks))
return messages
def _build_request(
slug: str,
system_prompt: str,
prompt: str,
media_blocks: list[OpenRouterContentBlock],
*,
seed: int,
reasoning_effort: str | None,
search_context_size: str | None,
) -> OpenRouterChatRequest:
reasoning_cfg: OpenRouterReasoningConfig | None = None
if reasoning_effort and reasoning_effort != "off":
# exclude=True asks providers to reason internally but not return the trace
reasoning_cfg = OpenRouterReasoningConfig(effort=reasoning_effort, exclude=True)
web_search_cfg: OpenRouterWebSearchOptions | None = None
if search_context_size:
web_search_cfg = OpenRouterWebSearchOptions(search_context_size=search_context_size)
return OpenRouterChatRequest(
model=slug,
messages=_build_messages(system_prompt, prompt, media_blocks),
seed=seed if seed > 0 else None,
reasoning=reasoning_cfg,
web_search_options=web_search_cfg,
)
def _extract_text(response: OpenRouterChatResponse) -> str:
if response.error:
code = response.error.code if response.error.code is not None else "unknown"
raise ValueError(f"OpenRouter error ({code}): {response.error.message or 'no message'}")
if not response.choices:
raise ValueError("Empty response from OpenRouter (no choices).")
message = response.choices[0].message
if not message:
raise ValueError("Empty response from OpenRouter (no message).")
if message.refusal:
raise ValueError(f"Model refused to respond: {message.refusal}")
return message.content or ""
class OpenRouterLLMNode(IO.ComfyNode):
@classmethod
def define_schema(cls):
return IO.Schema(
node_id="OpenRouterLLMNode",
display_name="OpenRouter LLM",
category="api node/text/OpenRouter",
essentials_category="Text Generation",
description=(
"Generate text responses through OpenRouter. Routes to a curated set of popular "
"models from xAI, DeepSeek, Qwen, Mistral, Z.AI (GLM), Moonshot (Kimi), and "
"Perplexity Sonar."
),
inputs=[
IO.String.Input(
"prompt",
multiline=True,
default="",
tooltip="Text input to the model.",
),
IO.DynamicCombo.Input(
"model",
options=_build_model_options(),
tooltip="The OpenRouter model used to generate the response.",
),
IO.Int.Input(
"seed",
default=0,
min=0,
max=2147483647,
control_after_generate=True,
tooltip="Seed for sampling. Set to 0 to omit. Most models treat this as a hint only.",
),
IO.String.Input(
"system_prompt",
multiline=True,
default="",
optional=True,
advanced=True,
tooltip="Foundational instructions that dictate the model's behavior.",
),
],
outputs=[IO.String.Output()],
hidden=[
IO.Hidden.auth_token_comfy_org,
IO.Hidden.api_key_comfy_org,
IO.Hidden.unique_id,
],
is_api_node=True,
price_badge=IO.PriceBadge(
depends_on=IO.PriceBadgeDepends(widgets=["model"]),
expr=_price_badge_jsonata(),
),
)
@classmethod
async def execute(
cls,
prompt: str,
model: dict,
seed: int,
system_prompt: str = "",
) -> IO.NodeOutput:
validate_string(prompt, strip_whitespace=True, min_length=1)
slug: str = model["model"]
spec = _MODELS_BY_SLUG.get(slug)
if spec is None:
raise ValueError(f"Unknown OpenRouter model: {slug}")
reasoning_effort: str | None = model.get("reasoning_effort")
search_context_size: str | None = model.get("search_context_size")
image_tensors: list[Input.Image] = [t for t in (model.get("images") or {}).values() if t is not None]
if image_tensors and sum(get_number_of_images(t) for t in image_tensors) > spec.max_images:
raise ValueError(f"Up to {spec.max_images} images are supported for {slug}.")
video_inputs: list[Input.Video] = [v for v in (model.get("videos") or {}).values() if v is not None]
if video_inputs and len(video_inputs) > spec.max_videos:
raise ValueError(f"Up to {spec.max_videos} videos are supported for {slug}.")
media_blocks: list[OpenRouterContentBlock] = []
if image_tensors:
media_blocks.extend(await _build_image_blocks(cls, spec, image_tensors))
if video_inputs:
media_blocks.extend(await _build_video_blocks(cls, video_inputs))
request = _build_request(
slug,
system_prompt,
prompt,
media_blocks,
seed=seed,
reasoning_effort=reasoning_effort,
search_context_size=search_context_size,
)
response = await sync_op(
cls,
ApiEndpoint(path=OPENROUTER_CHAT_ENDPOINT, method="POST"),
response_model=OpenRouterChatResponse,
data=request,
price_extractor=_calculate_price,
)
return IO.NodeOutput(_extract_text(response))
class OpenRouterExtension(ComfyExtension):
@override
async def get_node_list(self) -> list[type[IO.ComfyNode]]:
return [OpenRouterLLMNode]
async def comfy_entrypoint() -> OpenRouterExtension:
return OpenRouterExtension()

View File

@ -16,17 +16,16 @@ from .conversions import (
convert_mask_to_image,
downscale_image_tensor,
downscale_image_tensor_by_max_side,
downscale_video_to_max_pixels,
image_tensor_pair_to_batch,
pil_to_bytesio,
resize_mask_to_image,
resize_video_to_pixel_budget,
tensor_to_base64_string,
tensor_to_bytesio,
tensor_to_pil,
text_filepath_to_base64_string,
text_filepath_to_data_uri,
trim_video,
upscale_video_to_min_pixels,
video_to_base64_string,
)
from .download_helpers import (
@ -89,17 +88,16 @@ __all__ = [
"convert_mask_to_image",
"downscale_image_tensor",
"downscale_image_tensor_by_max_side",
"downscale_video_to_max_pixels",
"image_tensor_pair_to_batch",
"pil_to_bytesio",
"resize_mask_to_image",
"resize_video_to_pixel_budget",
"tensor_to_base64_string",
"tensor_to_bytesio",
"tensor_to_pil",
"text_filepath_to_base64_string",
"text_filepath_to_data_uri",
"trim_video",
"upscale_video_to_min_pixels",
"video_to_base64_string",
# Validation utilities
"get_image_dimensions",

View File

@ -415,48 +415,14 @@ def trim_video(video: Input.Video, duration_sec: float) -> Input.Video:
raise RuntimeError(f"Failed to trim video: {str(e)}") from e
def downscale_video_to_max_pixels(video: Input.Video, max_pixels: int) -> Input.Video:
"""Downscale a video to fit within ``max_pixels`` (w * h), preserving aspect ratio.
def resize_video_to_pixel_budget(video: Input.Video, total_pixels: int) -> Input.Video:
"""Downscale a video to fit within ``total_pixels`` (w * h), preserving aspect ratio.
Returns the original video object untouched when it already fits. Preserves frame rate, duration, and audio.
Aspect ratio is preserved up to a fraction of a percent (even-dim rounding).
"""
src_w, src_h = video.get_dimensions()
scale_dims = _compute_downscale_dims(src_w, src_h, max_pixels)
if scale_dims is None:
return video
return _apply_video_scale(video, scale_dims)
def _compute_upscale_dims(src_w: int, src_h: int, total_pixels: int) -> tuple[int, int] | None:
"""Return upscaled (w, h) with even dims meeting at least ``total_pixels``, or None if already large enough.
Source aspect ratio is preserved; output may drift by a fraction of a percent because both dimensions
are rounded up to even values (many codecs require divisible-by-2). The result is guaranteed to be at
least ``total_pixels``.
"""
pixels = src_w * src_h
if pixels >= total_pixels:
return None
scale = math.sqrt(total_pixels / pixels)
new_w = math.ceil(src_w * scale)
new_h = math.ceil(src_h * scale)
if new_w % 2:
new_w += 1
if new_h % 2:
new_h += 1
return new_w, new_h
def upscale_video_to_min_pixels(video: Input.Video, min_pixels: int) -> Input.Video:
"""Upscale a video to meet at least ``min_pixels`` (w * h), preserving aspect ratio.
Returns the original video object untouched when it already meets the minimum. Preserves frame rate,
duration, and audio. Aspect ratio is preserved up to a fraction of a percent (even-dim rounding).
Note: upscaling a low-resolution source does not add real detail; downstream model quality may suffer.
"""
src_w, src_h = video.get_dimensions()
scale_dims = _compute_upscale_dims(src_w, src_h, min_pixels)
scale_dims = _compute_downscale_dims(src_w, src_h, total_pixels)
if scale_dims is None:
return video
return _apply_video_scale(video, scale_dims)

View File

@ -543,7 +543,7 @@ class AudioConcat(IO.ComfyNode):
return IO.Schema(
node_id="AudioConcat",
search_aliases=["join audio", "combine audio", "append audio"],
display_name="Concatenate Audio",
display_name="Audio Concat",
description="Concatenates the audio1 to audio2 in the specified direction.",
category="audio",
inputs=[
@ -597,7 +597,7 @@ class AudioMerge(IO.ComfyNode):
return IO.Schema(
node_id="AudioMerge",
search_aliases=["mix audio", "overlay audio", "layer audio"],
display_name="Merge Audio",
display_name="Audio Merge",
description="Combine two audio tracks by overlaying their waveforms.",
category="audio",
inputs=[
@ -667,9 +667,8 @@ class AudioAdjustVolume(IO.ComfyNode):
return IO.Schema(
node_id="AudioAdjustVolume",
search_aliases=["audio gain", "loudness", "audio level"],
display_name="Adjust Audio Volume",
display_name="Audio Adjust Volume",
category="audio",
description="Adjust the volume of the audio by a specified amount in decibels (dB).",
inputs=[
IO.Audio.Input("audio"),
IO.Int.Input(

View File

@ -47,10 +47,8 @@ class LoadImageDataSetFromFolderNode(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="LoadImageDataSetFromFolder",
search_aliases=["load folder", "load from folder", "load dataset", "load images", "import dataset"],
display_name="Load Image (from Folder)",
category="image",
description="Load a dataset of images from a specified folder and return a list of images. Supported formats: PNG, JPG, JPEG, WEBP.",
display_name="Load Image Dataset from Folder",
category="dataset",
is_experimental=True,
inputs=[
io.Combo.Input(
@ -86,16 +84,14 @@ class LoadImageTextDataSetFromFolderNode(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="LoadImageTextDataSetFromFolder",
search_aliases=["load folder", "load from folder", "load dataset", "load images", "import dataset"],
display_name="Load Image-Text (from Folder)",
category="image",
description="Load a dataset of pairs of images and text captions from a specified folder and return them as a list. Supported formats: PNG, JPG, JPEG, WEBP.",
display_name="Load Image and Text Dataset from Folder",
category="dataset",
is_experimental=True,
inputs=[
io.Combo.Input(
"folder",
options=folder_paths.get_input_subfolders(),
tooltip="The folder to load images and text captions from.",
tooltip="The folder to load images from.",
)
],
outputs=[
@ -210,10 +206,8 @@ class SaveImageDataSetToFolderNode(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="SaveImageDataSetToFolder",
search_aliases=["save folder", "save to folder", "save dataset", "save images", "export dataset"],
display_name="Save Image (to Folder) (DEPRECATED)",
category="image",
description="Save a dataset of images to a specified folder. Supported formats: PNG.",
display_name="Save Image Dataset to Folder",
category="dataset",
is_experimental=True,
is_output_node=True,
is_input_list=True, # Receive images as list
@ -232,7 +226,6 @@ class SaveImageDataSetToFolderNode(io.ComfyNode):
),
],
outputs=[],
is_deprecated=True, # This node is redundant and superseded by existing Save Image nodes where the target folder can be specified in the filename_prefix
)
@classmethod
@ -253,20 +246,14 @@ class SaveImageTextDataSetToFolderNode(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="SaveImageTextDataSetToFolder",
search_aliases=["save folder", "save to folder", "save dataset", "save images", "save text", "export dataset"],
display_name="Save Image-Text (to Folder)",
category="image",
description="Save a dataset of pairs of images and text captions to a specified folder. Images are saved as PNG files and captions are saved as TXT files with the same filename_prefix.",
display_name="Save Image and Text Dataset to Folder",
category="dataset",
is_experimental=True,
is_output_node=True,
is_input_list=True, # Receive both images and texts as lists
inputs=[
io.Image.Input("images", tooltip="List of images to save."),
io.String.Input("texts",
optional=True,
force_input=True,
tooltip="List of text captions to save."
),
io.String.Input("texts", tooltip="List of text captions to save."),
io.String.Input(
"folder_name",
default="dataset",
@ -283,7 +270,7 @@ class SaveImageTextDataSetToFolderNode(io.ComfyNode):
)
@classmethod
def execute(cls, images, folder_name, filename_prefix, texts=None):
def execute(cls, images, texts, folder_name, filename_prefix):
# Extract scalar values
folder_name = folder_name[0]
filename_prefix = filename_prefix[0]
@ -292,12 +279,11 @@ class SaveImageTextDataSetToFolderNode(io.ComfyNode):
saved_files = save_images_to_folder(images, output_dir, filename_prefix)
# Save captions
if texts:
for idx, (filename, caption) in enumerate(zip(saved_files, texts)):
caption_filename = filename.replace(".png", ".txt")
caption_path = os.path.join(output_dir, caption_filename)
with open(caption_path, "w", encoding="utf-8") as f:
f.write(caption)
for idx, (filename, caption) in enumerate(zip(saved_files, texts)):
caption_filename = filename.replace(".png", ".txt")
caption_path = os.path.join(output_dir, caption_filename)
with open(caption_path, "w", encoding="utf-8") as f:
f.write(caption)
logging.info(f"Saved {len(saved_files)} images and captions to {output_dir}.")
return io.NodeOutput()
@ -328,13 +314,11 @@ class ImageProcessingNode(io.ComfyNode):
Child classes should set:
node_id: Unique node identifier (required)
search_aliases: List of search aliases (optional)
display_name: Display name (optional, defaults to node_id)
description: Node description (optional)
extra_inputs: List of additional io.Input objects beyond "images" (optional)
is_group_process: None (auto-detect), True (group), or False (individual) (optional)
is_output_list: True (list output) or False (single output) (optional, default True)
is_deprecated: True if the node is deprecated (optional, default False)
Child classes must implement ONE of:
_process(cls, image, **kwargs) -> tensor (for single-item processing)
@ -342,13 +326,12 @@ class ImageProcessingNode(io.ComfyNode):
"""
node_id = None
search_aliases = []
display_name = None
description = None
extra_inputs = []
is_group_process = None # None = auto-detect, True/False = explicit
is_output_list = None # None = auto-detect based on processing mode
is_deprecated = False
@classmethod
def _detect_processing_mode(cls):
"""Detect whether this node uses group or individual processing.
@ -419,10 +402,8 @@ class ImageProcessingNode(io.ComfyNode):
return io.Schema(
node_id=cls.node_id,
search_aliases=cls.search_aliases,
display_name=cls.display_name or cls.node_id,
category=cls.category,
description=cls.description,
category="dataset/image",
is_experimental=True,
is_input_list=is_group, # True for group, False for individual
inputs=inputs,
@ -491,13 +472,11 @@ class TextProcessingNode(io.ComfyNode):
Child classes should set:
node_id: Unique node identifier (required)
search_aliases: List of search aliases (optional)
display_name: Display name (optional, defaults to node_id)
description: Node description (optional)
extra_inputs: List of additional io.Input objects beyond "texts" (optional)
is_group_process: None (auto-detect), True (group), or False (individual) (optional)
is_output_list: True (list output) or False (single output) (optional, default True)
is_deprecated: True if the node is deprecated (optional, default False)
Child classes must implement ONE of:
_process(cls, text, **kwargs) -> str (for single-item processing)
@ -505,13 +484,12 @@ class TextProcessingNode(io.ComfyNode):
"""
node_id = None
search_aliases = []
display_name = None
description = None
extra_inputs = []
is_group_process = None # None = auto-detect, True/False = explicit
is_output_list = None # None = auto-detect based on processing mode
is_deprecated = False
@classmethod
def _detect_processing_mode(cls):
"""Detect whether this node uses group or individual processing.
@ -649,17 +627,15 @@ class TextProcessingNode(io.ComfyNode):
class ResizeImagesByShorterEdgeNode(ImageProcessingNode):
node_id = "ResizeImagesByShorterEdge"
display_name = "Resize Images by Shorter Edge (DEPRECATED)"
category = "image/transform"
description = "Resize images so that the shorter edge matches the specified dimension while preserving aspect ratio."
is_deprecated = True # This node is superseded by Resize Image/Mask with resize_type = scale shorter dimension
display_name = "Resize Images by Shorter Edge"
description = "Resize images so that the shorter edge matches the specified length while preserving aspect ratio."
extra_inputs = [
io.Int.Input(
"shorter_edge",
default=512,
min=1,
max=8192,
tooltip="Target dimension for the shorter edge.",
tooltip="Target length for the shorter edge.",
),
]
@ -679,17 +655,15 @@ class ResizeImagesByShorterEdgeNode(ImageProcessingNode):
class ResizeImagesByLongerEdgeNode(ImageProcessingNode):
node_id = "ResizeImagesByLongerEdge"
display_name = "Resize Images by Longer Edge (DEPRECATED)"
category = "image/transform"
description = "Resize images so that the longer edge matches the specified dimension while preserving aspect ratio."
is_deprecated = True # This node is superseded by Resize Image/Mask with resize_type = scale longer dimension
display_name = "Resize Images by Longer Edge"
description = "Resize images so that the longer edge matches the specified length while preserving aspect ratio."
extra_inputs = [
io.Int.Input(
"longer_edge",
default=1024,
min=1,
max=8192,
tooltip="Target dimension for the longer edge.",
tooltip="Target length for the longer edge.",
),
]
@ -712,10 +686,8 @@ class ResizeImagesByLongerEdgeNode(ImageProcessingNode):
class CenterCropImagesNode(ImageProcessingNode):
node_id = "CenterCropImages"
search_aliases=["crop", "cut", "trim"]
display_name="Crop Image (Center)"
category="image/transform"
description = "Center crop an image to the specified dimensions."
display_name = "Center Crop Images"
description = "Center crop all images to the specified dimensions."
extra_inputs = [
io.Int.Input("width", default=512, min=1, max=8192, tooltip="Crop width."),
io.Int.Input("height", default=512, min=1, max=8192, tooltip="Crop height."),
@ -734,11 +706,10 @@ class CenterCropImagesNode(ImageProcessingNode):
class RandomCropImagesNode(ImageProcessingNode):
node_id = "RandomCropImages"
search_aliases=["crop", "cut", "trim"]
display_name = "Crop Image (Random)"
category="image/transform"
description = "Randomly crop an image to the specified dimensions."
display_name = "Random Crop Images"
description = (
"Randomly crop all images to the specified dimensions (for data augmentation)."
)
extra_inputs = [
io.Int.Input("width", default=512, min=1, max=8192, tooltip="Crop width."),
io.Int.Input("height", default=512, min=1, max=8192, tooltip="Crop height."),
@ -763,9 +734,7 @@ class RandomCropImagesNode(ImageProcessingNode):
class NormalizeImagesNode(ImageProcessingNode):
node_id = "NormalizeImages"
search_aliases=["normalize", "normalize colors"]
display_name = "Normalize Image Colors"
category = "image/color"
display_name = "Normalize Images"
description = "Normalize images using mean and standard deviation."
extra_inputs = [
io.Float.Input(
@ -793,10 +762,8 @@ class NormalizeImagesNode(ImageProcessingNode):
class AdjustBrightnessNode(ImageProcessingNode):
node_id = "AdjustBrightness"
search_aliases=["brightness"]
display_name = "Adjust Brightness"
category="image/adjustments"
description = "Adjust the brightness of an image."
description = "Adjust brightness of all images."
extra_inputs = [
io.Float.Input(
"factor",
@ -814,10 +781,8 @@ class AdjustBrightnessNode(ImageProcessingNode):
class AdjustContrastNode(ImageProcessingNode):
node_id = "AdjustContrast"
search_aliases=["contrast"]
display_name = "Adjust Contrast"
category="image/adjustments"
description = "Adjust the contrast of an image."
description = "Adjust contrast of all images."
extra_inputs = [
io.Float.Input(
"factor",
@ -835,10 +800,8 @@ class AdjustContrastNode(ImageProcessingNode):
class ShuffleDatasetNode(ImageProcessingNode):
node_id = "ShuffleDataset"
search_aliases=["shuffle", "randomize", "mix"]
display_name = "Shuffle Images List"
category = "image/batch"
description = "Randomly shuffle the order of images in a list."
display_name = "Shuffle Image Dataset"
description = "Randomly shuffle the order of images in the dataset."
is_group_process = True # Requires full list to shuffle
extra_inputs = [
io.Int.Input(
@ -860,15 +823,13 @@ class ShuffleImageTextDatasetNode(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="ShuffleImageTextDataset",
search_aliases=["shuffle", "randomize", "mix"],
display_name = "Shuffle Pairs of Image-Text",
category = "image/batch",
description = "Randomly shuffle the order of pairs of image-text in a list.",
display_name="Shuffle Image-Text Dataset",
category="dataset/image",
is_experimental=True,
is_input_list=True,
inputs=[
io.Image.Input("images", tooltip="List of images to shuffle."),
io.String.Input("texts", tooltip="List of texts to shuffle.", force_input=True),
io.String.Input("texts", tooltip="List of texts to shuffle."),
io.Int.Input(
"seed",
default=0,
@ -904,11 +865,8 @@ class ShuffleImageTextDatasetNode(io.ComfyNode):
class TextToLowercaseNode(TextProcessingNode):
node_id = "TextToLowercase"
search_aliases=["lowercase"]
display_name = "Convert Text to Lowercase (DEPRECATED)"
category = "text"
description = "Convert text to lowercase."
is_deprecated = True # This node is superseded by the Convert Text Case node
display_name = "Text to Lowercase"
description = "Convert all texts to lowercase."
@classmethod
def _process(cls, text):
@ -917,11 +875,8 @@ class TextToLowercaseNode(TextProcessingNode):
class TextToUppercaseNode(TextProcessingNode):
node_id = "TextToUppercase"
search_aliases=["uppercase"]
display_name = "Convert Text to Uppercase (DEPRECATED)"
category = "text"
description = "Convert text to uppercase."
is_deprecated = True # This node is superseded by the Convert Text Case node
display_name = "Text to Uppercase"
description = "Convert all texts to uppercase."
@classmethod
def _process(cls, text):
@ -930,10 +885,8 @@ class TextToUppercaseNode(TextProcessingNode):
class TruncateTextNode(TextProcessingNode):
node_id = "TruncateText"
search_aliases=["truncate", "cut", "shorten"]
display_name = "Truncate Text"
category = "text"
description = "Truncate text to a maximum length."
description = "Truncate all texts to a maximum length."
extra_inputs = [
io.Int.Input(
"max_length", default=77, min=1, max=10000, tooltip="Maximum text length."
@ -947,10 +900,8 @@ class TruncateTextNode(TextProcessingNode):
class AddTextPrefixNode(TextProcessingNode):
node_id = "AddTextPrefix"
display_name = "Add Text Prefix (DEPRECATED)"
category = "text"
display_name = "Add Text Prefix"
description = "Add a prefix to all texts."
is_deprecated = True # This node is superseded by the Concatenate Text node
extra_inputs = [
io.String.Input("prefix", default="", tooltip="Prefix to add."),
]
@ -962,10 +913,8 @@ class AddTextPrefixNode(TextProcessingNode):
class AddTextSuffixNode(TextProcessingNode):
node_id = "AddTextSuffix"
display_name = "Add Text Suffix (DEPRECATED)"
category = "text"
display_name = "Add Text Suffix"
description = "Add a suffix to all texts."
is_deprecated = True # This node is superseded by the Concatenate Text node
extra_inputs = [
io.String.Input("suffix", default="", tooltip="Suffix to add."),
]
@ -977,10 +926,8 @@ class AddTextSuffixNode(TextProcessingNode):
class ReplaceTextNode(TextProcessingNode):
node_id = "ReplaceText"
display_name = "Replace Text (DEPRECATED)"
category = "text"
display_name = "Replace Text"
description = "Replace text in all texts."
is_deprecated = True # This node is superseded by the other Replace Text node
extra_inputs = [
io.String.Input("find", default="", tooltip="Text to find."),
io.String.Input("replace", default="", tooltip="Text to replace with."),
@ -993,10 +940,8 @@ class ReplaceTextNode(TextProcessingNode):
class StripWhitespaceNode(TextProcessingNode):
node_id = "StripWhitespace"
display_name = "Strip Whitespace (DEPRECATED)"
category = "text"
display_name = "Strip Whitespace"
description = "Strip leading and trailing whitespace from all texts."
is_deprecated = True # This node is superseded by the Trim Text node
@classmethod
def _process(cls, text):
@ -1007,13 +952,11 @@ class StripWhitespaceNode(TextProcessingNode):
class ImageDeduplicationNode(ImageProcessingNode):
"""Remove duplicate or very similar images from a list using perceptual hashing."""
"""Remove duplicate or very similar images from the dataset using perceptual hashing."""
node_id = "ImageDeduplication"
search_aliases=["deduplicate", "remove duplicates", "similarity filter"]
display_name = "Deduplicate Images"
category = "image/batch"
description = "Remove duplicate or very similar images from a list."
display_name = "Image Deduplication"
description = "Remove duplicate or very similar images from the dataset."
is_group_process = True # Requires full list to compare images
extra_inputs = [
io.Float.Input(
@ -1083,9 +1026,7 @@ class ImageGridNode(ImageProcessingNode):
"""Combine multiple images into a single grid/collage."""
node_id = "ImageGrid"
search_aliases=["grid", "collage", "combine"]
display_name = "Make Image Grid"
category="image/batch"
display_name = "Image Grid"
description = "Arrange multiple images into a grid layout."
is_group_process = True # Requires full list to create grid
is_output_list = False # Outputs single grid image
@ -1161,12 +1102,9 @@ class MergeImageListsNode(ImageProcessingNode):
"""Merge multiple image lists into a single list."""
node_id = "MergeImageLists"
search_aliases=["list", "merge list", "make list"]
display_name = "Merge Image Lists (DEPRECATED)"
category = "image/batch"
display_name = "Merge Image Lists"
description = "Concatenate multiple image lists into one."
is_group_process = True # Receives images as list
is_deprecated = True # This node is superseded by the Create List node
@classmethod
def _group_process(cls, images):
@ -1181,11 +1119,9 @@ class MergeTextListsNode(TextProcessingNode):
"""Merge multiple text lists into a single list."""
node_id = "MergeTextLists"
display_name = "Merge Text Lists (DEPRECATED)"
category = "text"
display_name = "Merge Text Lists"
description = "Concatenate multiple text lists into one."
is_group_process = True # Receives texts as list
is_deprecated = True # This node is superseded by the Create List node
@classmethod
def _group_process(cls, texts):
@ -1206,10 +1142,8 @@ class ResolutionBucket(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="ResolutionBucket",
search_aliases=["bucket by resolution", "group by resolution", "batch by resolution"],
display_name="Resolution Bucket",
category="training",
description="Group latents and conditionings into buckets",
category="dataset",
is_experimental=True,
is_input_list=True,
inputs=[
@ -1302,8 +1236,7 @@ class MakeTrainingDataset(io.ComfyNode):
node_id="MakeTrainingDataset",
search_aliases=["encode dataset"],
display_name="Make Training Dataset",
category="training",
description="Encode images with VAE and texts with CLIP to create a training dataset of latents and conditionings.",
category="dataset",
is_experimental=True,
is_input_list=True, # images and texts as lists
inputs=[
@ -1318,7 +1251,6 @@ class MakeTrainingDataset(io.ComfyNode):
"texts",
optional=True,
tooltip="List of text captions. Can be length n (matching images), 1 (repeated for all), or omitted (uses empty string).",
force_input=True
),
],
outputs=[
@ -1388,10 +1320,9 @@ class SaveTrainingDataset(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="SaveTrainingDataset",
search_aliases=["export dataset", "save dataset"],
search_aliases=["export training data"],
display_name="Save Training Dataset",
category="training",
description="Save encoded training dataset (latents + conditioning) to disk for efficient loading during training.",
category="dataset",
is_experimental=True,
is_output_node=True,
is_input_list=True, # Receive lists
@ -1493,8 +1424,7 @@ class LoadTrainingDataset(io.ComfyNode):
node_id="LoadTrainingDataset",
search_aliases=["import dataset", "training data"],
display_name="Load Training Dataset",
category="training",
description="Load encoded training dataset (latents + conditioning) from disk for use in training.",
category="dataset",
is_experimental=True,
inputs=[
io.String.Input(

View File

@ -419,17 +419,15 @@ class VoxelToMeshBasic(IO.ComfyNode):
def define_schema(cls):
return IO.Schema(
node_id="VoxelToMeshBasic",
display_name="Voxel to Mesh (Basic) (DEPRECATED)",
display_name="Voxel to Mesh (Basic)",
category="3d",
description="Converts a voxel grid to a mesh.",
is_deprecated=True, # This node is superseded by the Voxel To Mesh node
inputs=[
IO.Voxel.Input("voxel"),
IO.Float.Input("threshold", default=0.6, min=-1.0, max=1.0, step=0.01),
],
outputs=[
IO.Mesh.Output(),
],
]
)
@classmethod
@ -455,10 +453,9 @@ class VoxelToMesh(IO.ComfyNode):
node_id="VoxelToMesh",
display_name="Voxel to Mesh",
category="3d",
description="Converts a voxel grid to a mesh.",
inputs=[
IO.Voxel.Input("voxel"),
IO.Combo.Input("algorithm", options=["surface net", "basic"]),
IO.Combo.Input("algorithm", options=["surface net", "basic"], advanced=True),
IO.Float.Input("threshold", default=0.6, min=-1.0, max=1.0, step=0.01),
],
outputs=[

View File

@ -55,10 +55,9 @@ class ImageCropV2(IO.ComfyNode):
def define_schema(cls):
return IO.Schema(
node_id="ImageCropV2",
search_aliases=["crop", "cut", "trim"],
search_aliases=["trim"],
display_name="Crop Image",
category="image/transform",
description = "Crop an image to the specified dimensions.",
essentials_category="Image Tools",
has_intermediate_output=True,
inputs=[

View File

@ -8,82 +8,6 @@ from comfy_api.latest import _io
MISSING = object()
class NotNode(io.ComfyNode):
@classmethod
def define_schema(cls):
return io.Schema(
node_id="ComfyNotNode",
display_name="Not",
category="utils/logic",
description="Logical NOT operation. Returns true if the value is falsy. Uses Python's rules for truthiness.",
search_aliases=["invert", "toggle", "negate", "flip boolean"],
inputs=[
io.AnyType.Input("value"),
],
outputs=[
io.Boolean.Output(),
],
)
@classmethod
def execute(cls, value) -> io.NodeOutput:
return io.NodeOutput(not value)
class AndNode(io.ComfyNode):
@classmethod
def define_schema(cls):
template = io.Autogrow.TemplatePrefix(
input=io.AnyType.Input("value"),
prefix="value",
min=1,
)
return io.Schema(
node_id="ComfyAndNode",
display_name="And",
category="utils/logic",
description="Logical AND operation. Returns true if all of the values are truthy. Uses Python's rules for truthiness.",
search_aliases=["all", "every"],
inputs=[
io.Autogrow.Input("values", template=template),
],
outputs=[
io.Boolean.Output(),
],
)
@classmethod
def execute(cls, values: io.Autogrow.Type) -> io.NodeOutput:
return io.NodeOutput(all(values.values()))
class OrNode(io.ComfyNode):
@classmethod
def define_schema(cls):
template = io.Autogrow.TemplatePrefix(
input=io.AnyType.Input("value"),
prefix="value",
min=1,
)
return io.Schema(
node_id="ComfyOrNode",
display_name="Or",
category="utils/logic",
description="Logical OR operation. Returns true if any of the values are truthy. Uses Python's rules for truthiness.",
search_aliases=["any", "some"],
inputs=[
io.Autogrow.Input("values", template=template),
],
outputs=[
io.Boolean.Output(),
],
)
@classmethod
def execute(cls, values: io.Autogrow.Type) -> io.NodeOutput:
return io.NodeOutput(any(values.values()))
class SwitchNode(io.ComfyNode):
@classmethod
def define_schema(cls):
@ -337,9 +261,6 @@ class LogicExtension(ComfyExtension):
return [
SwitchNode,
CustomComboNode,
NotNode,
AndNode,
OrNode,
# SoftSwitchNode,
# ConvertStringToComboNode,
# DCTestNode,

View File

@ -11,8 +11,8 @@ class LTXVAudioVAELoader(io.ComfyNode):
def define_schema(cls) -> io.Schema:
return io.Schema(
node_id="LTXVAudioVAELoader",
display_name="Load LTXV Audio VAE",
category="loaders",
display_name="LTXV Audio VAE Loader",
category="audio",
inputs=[
io.Combo.Input(
"ckpt_name",
@ -40,7 +40,7 @@ class LTXVAudioVAEEncode(VAEEncodeAudio):
return io.Schema(
node_id="LTXVAudioVAEEncode",
display_name="LTXV Audio VAE Encode",
category="latent/audio",
category="audio",
inputs=[
io.Audio.Input("audio", tooltip="The audio to be encoded."),
io.Vae.Input(
@ -63,7 +63,7 @@ class LTXVAudioVAEDecode(io.ComfyNode):
return io.Schema(
node_id="LTXVAudioVAEDecode",
display_name="LTXV Audio VAE Decode",
category="latent/audio",
category="audio",
inputs=[
io.Latent.Input("samples", tooltip="The latent to be decoded."),
io.Vae.Input(

View File

@ -28,7 +28,7 @@ from comfy_extras.mediapipe.face_landmarker import FaceLandmarker
from comfy_extras.mediapipe.face_geometry import transformation_matrix_from_detection
FaceDetectionType = io.Custom("FACE_DETECTION_MODEL")
FaceLandmarkerType = io.Custom("FACE_LANDMARKER")
FaceLandmarksType = io.Custom("FACE_LANDMARKS")
_CANONICAL_KEYS = ("canonical_vertices", "procrustes_indices", "procrustes_weights")
@ -204,19 +204,18 @@ class LoadMediaPipeFaceLandmarker(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="LoadMediaPipeFaceLandmarker",
search_aliases=["face", "facial", "mediapipe", "face landmark", "face mesh", "blazeface", "face detection"],
display_name="Load Face Detection Model (MediaPipe)",
display_name="Load MediaPipe Face Landmarker",
category="loaders",
inputs=[
io.Combo.Input("model_name", options=folder_paths.get_filename_list("detection"),
tooltip="Face detection model from models/detection/."),
io.Combo.Input("model_name", options=folder_paths.get_filename_list("mediapipe"),
tooltip="Face Landmarker safetensors from models/mediapipe/."),
],
outputs=[FaceDetectionType.Output()],
outputs=[FaceLandmarkerType.Output()],
)
@classmethod
def execute(cls, model_name) -> io.NodeOutput:
sd = comfy.utils.load_torch_file(folder_paths.get_full_path_or_raise("detection", model_name), safe_load=True)
sd = comfy.utils.load_torch_file(folder_paths.get_full_path_or_raise("mediapipe", model_name), safe_load=True)
wrapper = FaceLandmarkerModel(sd)
return io.NodeOutput(wrapper)
@ -235,12 +234,10 @@ class MediaPipeFaceLandmarker(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="MediaPipeFaceLandmarker",
search_aliases=["face", "facial", "mediapipe", "face landmark", "face mesh", "blazeface", "face detection"],
display_name="Detect Face Landmarks (MediaPipe)",
display_name="MediaPipe Face Landmarker",
category="image/detection",
description="Detects facial landmarks using MediaPipe model.",
inputs=[
FaceDetectionType.Input("face_detection_model"),
FaceLandmarkerType.Input("face_landmarker"),
io.Image.Input("image"),
io.Combo.Input("detector_variant", options=["short", "full", "both"], default="short",
tooltip="Face detector range. 'short' is tuned for close-up faces "
@ -264,9 +261,9 @@ class MediaPipeFaceLandmarker(io.ComfyNode):
)
@classmethod
def execute(cls, face_detection_model, image, detector_variant, num_faces, min_confidence,
def execute(cls, face_landmarker, image, detector_variant, num_faces, min_confidence,
missing_frame_fallback) -> io.NodeOutput:
canonical = face_detection_model.canonical_data
canonical = face_landmarker.canonical_data
img_np = _image_to_uint8(image)
B, H, W = img_np.shape[:3]
chunk = 16
@ -279,7 +276,7 @@ class MediaPipeFaceLandmarker(io.ComfyNode):
with tqdm(total=B, desc=f"MediaPipe Face Landmarker ({variant})") as tq:
for i in range(0, B, chunk):
end = min(i + chunk, B)
res.extend(face_detection_model.detect_batch(
res.extend(face_landmarker.detect_batch(
[img_np[bi] for bi in range(i, end)],
num_faces=int(num_faces),
score_thresh=float(min_confidence),
@ -309,7 +306,7 @@ class MediaPipeFaceLandmarker(io.ComfyNode):
per_bb.append({"x": x1, "y": y1, "width": x2 - x1, "height": y2 - y1, "label": "face", "score": float(f["score"])})
bboxes.append(per_bb)
return io.NodeOutput({"frames": frames, "image_size": (H, W),
"connection_sets": face_detection_model.connection_sets}, bboxes)
"connection_sets": face_landmarker.connection_sets}, bboxes)
# Topology keys unioned by the 'all' connections preset (contour parts + irises + nose).
@ -335,10 +332,8 @@ class MediaPipeFaceMeshVisualize(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="MediaPipeFaceMeshVisualize",
search_aliases=["face", "facial", "mediapipe", "face landmark", "face mesh", "blazeface", "face detection", "visualize"],
display_name="Visualize Face Landmarks (MediaPipe)",
display_name="MediaPipe Face Mesh Visualize",
category="image/detection",
description="Draws face landmarks mesh on the input image.",
inputs=[
FaceLandmarksType.Input("face_landmarks"),
io.Image.Input("image", optional=True, tooltip="If not connected, a black canvas will be used."),
@ -448,10 +443,8 @@ class MediaPipeFaceMask(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="MediaPipeFaceMask",
search_aliases=["face", "facial", "mediapipe", "face mask", "blazeface", "face detection", "visualize"],
display_name="Draw Face Mask (MediaPipe)",
display_name="MediaPipe Face Mask",
category="image/detection",
description="Draws a mask from face landmarks.",
inputs=[
FaceLandmarksType.Input("face_landmarks"),
io.DynamicCombo.Input(

View File

@ -103,10 +103,8 @@ class MoGePanoramaInference(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="MoGePanoramaInference",
search_aliases=["moge", "panorama", "depth", "geometry", "depth estimation", "geometry estimation"],
display_name="Run MoGe Panorama Inference",
display_name="MoGe Panorama Inference",
category="image/geometry_estimation",
description="Run MoGe on an equirectangular panorama by splitting it into 12 perspective views, running inference on each, and merging the results into a single depth map.",
inputs=[
MoGeModelType.Input("moge_model"),
io.Image.Input("image", tooltip="Equirectangular panorama (any aspect)."),
@ -224,9 +222,7 @@ class MoGeInference(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="MoGeInference",
search_aliases=["moge", "depth", "geometry", "depth estimation", "geometry estimation"],
display_name="Run MoGe Inference",
description="Run MoGe on a single image to estimate depth and geometry.",
display_name="MoGe Inference",
category="image/geometry_estimation",
inputs=[
MoGeModelType.Input("moge_model"),
@ -281,9 +277,7 @@ class MoGeRender(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="MoGeRender",
search_aliases=["moge", "render", "geometry", "depth", "normal"],
display_name="Render MoGe Geometry",
description="Render a depth map or normal map from geometry data",
display_name="MoGe Render",
category="image/geometry_estimation",
inputs=[
MoGeGeometry.Input("moge_geometry"),
@ -348,9 +342,7 @@ class MoGePointMapToMesh(io.ComfyNode):
def define_schema(cls):
return io.Schema(
node_id="MoGePointMapToMesh",
search_aliases=["moge", "mesh", "geometry", "point map"],
display_name="Convert MoGe Point Map to Mesh",
description="Convert a MoGe point map into a 3D mesh.",
display_name="MoGe Point Map to Mesh",
category="image/geometry_estimation",
inputs=[
MoGeGeometry.Input("moge_geometry"),

View File

@ -60,7 +60,7 @@ folder_names_and_paths["geometry_estimation"] = ([os.path.join(models_dir, "geom
folder_names_and_paths["optical_flow"] = ([os.path.join(models_dir, "optical_flow")], supported_pt_extensions)
folder_names_and_paths["detection"] = ([os.path.join(models_dir, "detection")], supported_pt_extensions)
folder_names_and_paths["mediapipe"] = ([os.path.join(models_dir, "mediapipe")], supported_pt_extensions)
output_directory = os.path.join(base_path, "output")
temp_directory = os.path.join(base_path, "temp")

View File

@ -1556,6 +1556,12 @@ paths:
type: string
enum: [asc, desc]
description: Sort direction
- name: job_ids
in: query
schema:
type: string
x-runtime: [cloud]
description: "[cloud-only] Comma-separated UUIDs to filter assets by associated job."
- name: include_public
in: query
schema:
@ -2508,25 +2514,37 @@ paths:
/api/assets/import:
post:
operationId: importPublishedAssets
operationId: importAssets
tags: [assets]
summary: "[cloud-only] Import published assets into the caller's library"
description: |
[cloud-only] Imports the specified published assets into the caller's asset library. New DB records reference the same storage objects; no file copying occurs. Assets the caller already owns (by hash) are deduplicated. The `id` field on each returned `AssetInfo` is the caller's newly-created private asset ID, not the published asset ID supplied in the request.
summary: Import assets from external URLs
description: "[cloud-only] Imports one or more assets from external URLs into the cloud asset store."
x-runtime: [cloud]
requestBody:
required: true
content:
application/json:
schema:
$ref: "#/components/schemas/ImportPublishedAssetsRequest"
type: object
required:
- imports
properties:
imports:
type: array
items:
$ref: "#/components/schemas/AssetImportRequest"
description: Assets to import
responses:
"200":
description: Successfully imported assets
description: Import initiated
content:
application/json:
schema:
$ref: "#/components/schemas/ImportPublishedAssetsResponse"
type: object
properties:
assets:
type: array
items:
$ref: "#/components/schemas/Asset"
"400":
description: Bad request
content:
@ -7361,35 +7379,24 @@ components:
type: string
description: Target path on the runtime filesystem
ImportPublishedAssetsRequest:
AssetImportRequest:
type: object
x-runtime: [cloud]
description: "[cloud-only] Request body for importing published assets into the caller's library."
description: "[cloud-only] A single asset to import from an external URL."
required:
- published_asset_ids
- url
properties:
published_asset_ids:
url:
type: string
format: uri
description: URL of the asset to import
name:
type: string
description: Display name for the imported asset
tags:
type: array
description: IDs of published assets (inputs and models) to import.
items:
type: string
share_id:
type: string
nullable: true
description: |
Optional. Share ID of the published workflow these assets belong to. When provided (non-null, non-empty): all `published_asset_ids` must belong to this share's workflow version; returns 400 if the share is not found or any asset does not belong to it. When omitted, null, or empty string: no share-scoped validation is performed and the assets are validated only against global rules (preserved for clients that have not yet adopted `share_id`).
ImportPublishedAssetsResponse:
type: object
x-runtime: [cloud]
description: "[cloud-only] Response after importing published assets. Each returned `AssetInfo.id` is the caller's newly-created private asset ID, not the published asset ID supplied in the request."
required:
- assets
properties:
assets:
type: array
items:
$ref: "#/components/schemas/AssetInfo"
RemoteAssetMetadata:
type: object

View File

@ -1,5 +1,5 @@
comfyui-frontend-package==1.43.18
comfyui-workflow-templates==0.9.82
comfyui-workflow-templates==0.9.79
comfyui-embedded-docs==0.5.0
torch
torchsde