singularitynet/example-service repository overview

⁠example-service

Simple image classification service compatible with the SingularityNET daemon

⁠Getting Started

⁠Prerequisites

• Python 3.6.5⁠

⁠Installing

• Clone the git repository

$ git clone [email protected]:singnet/example-service.git
$ cd example-service

• Install the dependencies

$ pip install -r requirements.txt

⁠Configuration

• The following default configuration can be overridden by populating configuration/config.json in the source tree with the desired values

{
  "SERVER_PORT": 5001,
  "MINIMUM_SCORE": 0.20,
  "LOG_LEVEL": 10
}

• SERVER_PORT: the port on which the example service will listen for incoming JSON-RPC calls over http
• MINIMUM_SCORE: the minimum confidence score (between 0 and 1 inclusive) required to return a given prediction
• LOG_LEVEL: the logging verbosity

⁠Running

⁠Standalone

• Invoke the example service directly

$ python image_classification_service

⁠With SingularityNET Daemon

⁠SingularityNET Daemon Configuration

Create snetd.config.json file containing the following:

{
  "passthrough_enabled": true
}

in order to enable example service work with daemon. See SingularityNet daemon configuration⁠ for detailed configuration description.

⁠Running Service + Daemon on Host

• Invoke the run-snet-service script which launches both snetd and the example service

$ ./scripts/run-snet-service

⁠Running Service + Daemon in Docker Container

• Ensure that PASSTHROUGH_ENDPOINT is configured to be "http://127.0.0.1:5001⁠" in your daemon configuration
• Run the docker image with your daemon configuration (where HOST_PORT is the port you want the daemon bound to on your host)

$ docker run --detach -p HOST_PORT:DAEMON_LISTENING_PORT -v /path/to/config:/snetd.config singularitynet/example-service:latest

⁠Testing

• Invoke the test-call script against a running instance of the example service

$ ./scripts/test-call

⁠Building Docker Image

• Invoke the docker build script

$ ./scripts/build-docker

⁠License

This project is licensed under the MIT License - see the LICENSE⁠ file for details.