The DeepSquare Art Generator is a web application that allows users to generate unique, high-quality images using a deep learning model. The application takes user inputs such as prompt text, seed value, number of steps, and image size, and uses them as input for the Stable Difussion generative model to generate a new image. The application then displays the generated image to the user.
pnpm i
- Start the experss server :
pnpm run dev - Then headout to http://localhost:3000 and enjoy
The code initializes an Express app with a single router that has two routes /get and /draw.
GET /get:
This route takes a job_id query parameter and returns a JSON response with an image_url property. The job_id parameter is used to fetch the logs of a DeepSquare job and parse them to find a link to the generated image. The image_url property in the response is a modified version of the found link with a different domain name.
GET /draw:
This route takes a prompt query parameter and returns a JSON response with a num and a seed property.
The prompt parameter is used to create a DeepSquare job with specific parameters.
The job is then submitted to the DeepSquare server using the DeepSquareClient API.
The num property in the response is the job ID of the submitted job, and the seed property is a random seed used in the job parameters.
The environment variables used in this code are:
PRIVATE_KEY: The private key to use for the DeepSquare client.METASCHEDULER_ADDR: The URL of the Metascheduler contract.CREDIT_ADDR: The address of the Credit contract.ENDPOINT: The endpoint URL of the Ethereum node to use for interacting with the blockchain.ROOT_PATH: The root path for the app. If not specified, the default value of/will be used.
These environment variables are loaded using the dotenv package from a .env file in the project directory. The .env file should contain the key-value pairs for each variable, like so:
PRIVATE_KEY=<PRIVATE_KEY>
METASCHEDULER_ADDR=<METASCHEDULER_ADDRESS>
CREDIT_ADDR=<CREDIT_ADDRESS>
ENDPOINT=<ETHEREUM_NODE_ENDPOINT>
ROOT_PATH=<ROOT_PATH>Note that the values of these variables are specific to the user's configuration and should not be shared with others.
Default values can be found in .env.example.
This is a TypeScript code that exports a function createJob that creates a job configuration object. The job configuration is passed to the job execution system of DeepSquare, that runs the job using the specified resources and environment variables.
The createJob function takes four parameters:
PROMPT: a string that represents the input to a deep learning model for generating images.SEED: a number that represents the seed value for the random number generator used in the deep learning model.STEPS: a number that represents the number of steps to generate an image using the deep learning model.SIZE: a number that represents the size of the image to be generated using the deep learning model.
The job configuration object has several fields:
enableLogging: a boolean that specifies whether to enable logging.resources: an object that specifies the resources required to run the job, including the number of tasks, CPUs per task, memory per CPU, and GPUs per task.env: an array of environment variables that are passed to the job as key-value pairs.output: an object that specifies the output of the job, including the URL of the transfer service used to transfer the output files.continuousOutputSync: a boolean that specifies whether to enable continuous output synchronization.steps: an array of steps to execute the job. Each step contains a name and a run object that specifies the container to run, the resources required, the environment variables to set, the shell to use, and the command to execute.
The createJob function returns the job configuration object as a Job object, which is defined in a separate module imported from @deepsquare/deepsquare-client/src/graphql/client/generated/graphql.