ComfyUI as API instance setup

Update article. Update article 2

this is the update but still published,

this is the update 2, and saving as published

## Aditya Research: ComfyUI and Cloud Deployment - [21 Sept 24 - Deployment requirement notes)](https://docs.google.com/document) > I believe it could be more efficient to employ GPUs with custom specifications tailored to ComfyUI’s requirements rather than utilizing the standard n1-standard-1 machine type. Based on community discussions, ComfyUI’s performance is often influenced by the specific hardware specifications, particularly the GPU, aligned with the models being used. Although it is technically feasible to deploy ComfyUI on GKE, I anticipate that familiarizing myself with GKE deployments and related technologies like Helm would require a significant time investment. Therefore, I believe it would be more expedient to provide ComfyUI via GCE initially, allowing for a quicker delivery to Bagas and the React Native developers so they can begin testing and experimentation.

- [21 Sept 24 - Deploying to GCP (using VM instance instead of GKE)](https://github.com/karaposu/comfyui-on-cloud/tree/main) Perhaps we could replicate the setup code and establish a new Google Kubernetes Engine (GKE) cluster with a single node that adheres to our desired custom specifications. We create a GCE instance instead of GKE, because we considering the points that - Given that our application is a single-service, user interface-focused deployment, we opted for a GCE instance instead of GKE. Given its relatively simple architecture and the fact that it doesn't require the complexities of a microservices-based architecture - For a single-service application like Comfy-UI, GCE can be more cost-effective as it eliminates the overhead associated with managing a Kubernetes cluster. - If the backend of our AIFiltre application requires the implementation of multiple services, transitioning to GKE might be a viable option to accommodate a microservices-based architecture. However, I have determined that the backend is currently utilizing Cloud Run.

## Setup Process (on GCE) Created a new GCE Instance with name comfyvm, with the following specs details: - Zone: us-central1-f - GPUs: 1 x NVIDIA T4 > Due to our reliance on artificial intelligence, we are utilizing GPUs to accelerate rendering and machine learning processes. - Storage size & Memory: 130GB & 16GB > A storage capacity of 130GB is necessary to accommodate the memory-intensive nature of our models.

## How to run - To access the Comfy-UI instance, please visit the following link. Ensure that you have been granted permission to join the GCP project associated with it. [GCE Instance - comfyvm](https://console.cloud.google.com/compute/instancesDetail/zones/us-central1-f/instances/comfyvm) - Then locate the "Start/Resume" button positioned at the uppermost section of the webpage. Proceed by clicking this button to initiate the Comfy-UI instance, to stop the instance click "Stop" button. - Once start button clicked, the console will displayed on http://34.42.155.162:8188, but it may takes a secs for starting up initialization. ### Below is video details how to run the workflow https://github.com/user-attachments/assets/2a3bb7b6-3d81-410c-9cfb-4fb2efce67c8