ViewComfy API Integration Guidelines

You have a solid grasp of Python and FastAPI integrations, especially when it comes to building web applications. Your focus is on integrating the ViewComfy API into these apps using Python.

Overview of the ViewComfy API

The ViewComfy API is a serverless API designed with the FastAPI framework. It executes custom ComfyUI workflows, and you'll make requests using the httpx library.

Important Things to Keep in Mind

• Cold Start: The first time you call the API, you might notice a delay. This is due to a cold start, which can lead to longer response times.
• Variable Generation Times: Execution times can vary quite a bit. Some workflows finish in under 2 seconds, while others might take several minutes.

API Request Guidelines

• Parameters Object: Remember, the params object shouldn’t be empty. If you don’t provide specific parameters, just adjust the seed value.

Response Format

When you make a request, the API responds with data in this JSON structure:

{
  "prompt_id": "string", // Unique identifier for the prompt
  "status": "string", // Current execution status
  "completed": bool, // Indicates if execution is complete
  "execution_time_seconds": float, // Time taken to execute
  "prompt": dict, // Original prompt configuration
  "outputs": [ // List of output files (optional)
    {
      "filename": "string", // Name of the output file
      "content_type": "string", // MIME type of the file
      "data": "string", // Base64 encoded file content
      "size": int // File size in bytes
    }
  ]
}

Steps to Integrate the ViewComfy API

1. Deploy Your Workflow

Start by deploying your ComfyUI workflow on the ViewComfy dashboard using the workflow_api.json file.

2. Calling the Workflow with the API

You can access the ViewComfy API through a standard POST request. It also supports streaming responses with Server-Sent Events, which lets you track ComfyUI logs in real-time.

3. Obtain Your API Keys

To use the API endpoint, generate your API keys from the ViewComfy dashboard.

4. Extract Workflow Parameters

Next, find the parameters in your workflow using the ViewComfy_API/Python/workflow_parameters_maker.py script. This script will help flatten your workflow_api.json.

Your flattened JSON should look something like this:

{
  "_3-node-class_type-info": "KSampler",
  "3-inputs-cfg": 6,
  "_6-node-class_type-info": "CLIP Text Encode (Positive Prompt)",
  "6-inputs-clip": [ "38", 0 ],
  "6-inputs-text": "A woman raising her head with hair blowing in the wind",
  "_52-node-class_type-info": "Load Image",
  "52-inputs-image": "<path_to_my_image>"
}

In this example, the first key-value pair shows that node 3 is the KSampler, and the "3-inputs-cfg" key sets its configuration value.

5. Update Your Script with Parameters

Now, grab the ViewComfy endpoint from your dashboard and assign it to view_comfy_api_url. You’ll also need your Client ID and Client Secret set up like this:

view_comfy_api_url = "<Your_ViewComfy_endpoint>"
client_id = "<Your_ViewComfy_client_id>"
client_secret = "<Your_ViewComfy_client_secret>"

Next, use the keys from the flattened JSON file to set your parameters. For example, if you want to change the prompt and input image, do the following:

params = {}
params["6-inputs-text"] = "A flamingo dancing on top of a server in a pink universe, masterpiece, best quality, very aesthetic"
params["52-inputs-image"] = open("/home/gbieler/GitHub/API_tests/input_img.png", "rb")

6. Calling the API

After adding your parameters to ViewComfy_API/Python/main.py, run the API call by executing:

python main.py

This command sends your parameters to ViewComfy_API/Python/api.py, where the functions for API calls and output handling live.

• The script defaults to the infer_with_logs function, which streams generation logs from ComfyUI. If you want a standard POST request instead, you can use the infer function.

The returned result object will include the workflow outputs and generation details, with outputs saved in your working directory.

API Functionality Overview

The API file (api.py) includes functions for interacting with the API and managing responses. Generally, you’ll only need to modify the main file (main.py) for your specific workflow.

API Endpoints

• infer: This endpoint waits for the request to complete before returning results.
• infer_with_logs: This endpoint gives real-time updates with ComfyUI logs, like a progress bar. To use it, pass a function to handle incoming log messages.

Extracting API Parameters

If you want to pull all parameters from your workflow_api.json, call the workflow_api_parameter_creator function. This will create a dictionary containing all the parameters from your workflow.