Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.
JavaScript 3d-generationai-agentsanimationavatardigital-humangenerative-aiglbgltfimage-to-3dllmmcpmcp-servermodel-context-protocolsolanatext-to-3dthreejsweb-componentswebglwebxrx402
fix(hunyuan3d): deploy the RTX worker without a rolling update that the one-GPU quota cannot fit
nirholasa74cee933e44
diff --git a/workers/model-hunyuan3d/cloudbuild.hunyuan21rtx.yaml b/workers/model-hunyuan3d/cloudbuild.hunyuan21rtx.yamlindex bd8a33517..cd82bf4da 100644--- a/workers/model-hunyuan3d/cloudbuild.hunyuan21rtx.yaml+++ b/workers/model-hunyuan3d/cloudbuild.hunyuan21rtx.yaml@@ -55,32 +55,51 @@ steps:- --all-tags- ${_REGION}-docker.pkg.dev/$PROJECT_ID/${_REPO}/server+ # The deploy cannot be a plain rolling update. Cloud Run enforces the RTX PRO+ # 6000 quota at ONE GPU here, and the serving revision holds it (min 1), so a+ # new revision's instance can never start: it logs only "Starting new+ # instance" once a minute until the 4 minute startup probe expires, and the+ # revision is left not Ready (2026-09-18, revision 00002, stuck three weeks).+ # So: release the GPU with service-level manual scaling at 0 (no revision is+ # created), deploy, then restore autoscaling with the service-level cap of 1.+ # The restore runs even when the deploy fails, which hands the GPU back to the+ # previous revision. The lane is down for one model load (~10 min) while the+ # new instance boots; forge fails over meanwhile. Shell variables are written+ # $$name so Cloud Build passes them through instead of substituting them.- id: deployname: gcr.io/google.com/cloudsdktool/cloud-sdk:slim- entrypoint: gcloud+ entrypoint: bashargs:- - run- - deploy- - ${_SERVICE}- - --image=${_REGION}-docker.pkg.dev/$PROJECT_ID/${_REPO}/server:rtx-$SHORT_SHA- - --region=${_REGION}- - --service-account=${_RUN_SA}- - --platform=managed- - --port=8080- - --gpu=1- - --gpu-type=nvidia-rtx-pro-6000- - --no-gpu-zonal-redundancy- - --no-cpu-throttling- - --cpu=20- - --memory=80Gi- - --min-instances=${_MIN_INSTANCES}- - --max-instances=${_MAX_INSTANCES}- - --timeout=900- - --allow-unauthenticated- - --add-volume=name=weights,type=cloud-storage,bucket=${_WEIGHTS_BUCKET}- - --add-volume-mount=volume=weights,mount-path=/weights- - --set-env-vars=^;^GCS_BUCKET=${_GCS_BUCKET};HUNYUAN3D_MODEL=2.1;WEIGHTS_DIR=/weights/hunyuan3d-2.1;WEIGHTS_GCS_URI=gs://${_WEIGHTS_BUCKET}/hunyuan3d-2.1;WEIGHTS_LOCAL_DIR=/tmp/hunyuan3d-2.1;DINO_DIR=/weights/dinov2-giant;DINO_GCS_URI=gs://${_WEIGHTS_BUCKET}/dinov2-giant;DINO_LOCAL_DIR=/tmp/dinov2-giant;U2NET_HOME=/weights/u2net;MAX_CONCURRENT=1- - --set-secrets=API_KEY=avatar-reconstruction-key:latest+ - -c+ - |+ set -u+ gcloud run services update ${_SERVICE} --region=${_REGION} --scaling=0 --quiet || true+ gcloud run deploy ${_SERVICE} \+ --image=${_REGION}-docker.pkg.dev/$PROJECT_ID/${_REPO}/server:rtx-$SHORT_SHA \+ --region=${_REGION} \+ --service-account=${_RUN_SA} \+ --platform=managed \+ --port=8080 \+ --gpu=1 \+ --gpu-type=nvidia-rtx-pro-6000 \+ --no-gpu-zonal-redundancy \+ --no-cpu-throttling \+ --cpu=20 \+ --memory=80Gi \+ --min-instances=${_MIN_INSTANCES} \+ --max-instances=${_MAX_INSTANCES} \+ --timeout=900 \+ --allow-unauthenticated \+ --add-volume=name=weights,type=cloud-storage,bucket=${_WEIGHTS_BUCKET} \+ --add-volume-mount=volume=weights,mount-path=/weights \+ '--set-env-vars=^;^GCS_BUCKET=${_GCS_BUCKET};HUNYUAN3D_MODEL=2.1;WEIGHTS_DIR=/weights/hunyuan3d-2.1;WEIGHTS_GCS_URI=gs://${_WEIGHTS_BUCKET}/hunyuan3d-2.1;WEIGHTS_LOCAL_DIR=/tmp/hunyuan3d-2.1;DINO_DIR=/weights/dinov2-giant;DINO_GCS_URI=gs://${_WEIGHTS_BUCKET}/dinov2-giant;DINO_LOCAL_DIR=/tmp/dinov2-giant;U2NET_HOME=/weights/u2net;MAX_CONCURRENT=1' \+ --set-secrets=API_KEY=avatar-reconstruction-key:latest \+ --quiet+ deployed=$$?+ gcloud run services update ${_SERVICE} --region=${_REGION} --scaling=auto --max=${_MAX_INSTANCES} --quiet+ restored=$$?+ if [ "$$deployed" -ne 0 ]; then exit "$$deployed"; fi+ exit "$$restored"images:- ${_REGION}-docker.pkg.dev/$PROJECT_ID/${_REPO}/server:rtx-$SHORT_SHA