求助：如何在H2O Steam中将R代码发布为REST API

Got it, let's walk through exactly how to turn your R code into a deployable REST API using H2O Steam. I’ve gone through this process a few times, so here’s a practical, no-fluff breakdown:

1. Prep Your R Code for API Use #

First, you need to wrap your R logic in a way that can handle HTTP requests and return structured responses. For flexibility, I recommend using the plumber package—it’s designed specifically for building APIs in R and plays nicely with Steam.

Here’s a simple example:

library(plumber)
library(jsonlite)

# Define your API endpoint and logic
#* @post /compute-average
#* @param values: Numeric array to calculate average from
function(values) {
  # Calculate the average (your custom logic goes here)
  avg <- mean(as.numeric(values))
  # Return a JSON response
  return(toJSON(list(average = avg), auto_unbox = TRUE))
}

# For local testing (Steam will handle this part in deployment)
# pr() %>% pr_run(port = 8080)

If you don’t want to use plumber, you can write a basic function that parses raw JSON input and outputs JSON—but plumber makes routing and request handling way cleaner.

2. Package Your Dependencies #

Steam needs to know which R packages your code relies on. Create a requirements.txt file (or use renv to generate a lockfile for reproducibility) listing all required packages:

plumber
jsonlite
dplyr # Add any other packages your code uses

3. Deploy via H2O Steam Web UI #

Once your code and dependencies are ready, head to the Steam dashboard:

• Log in and navigate to the Apps section, then click New App.
• Under Runtime, select R (make sure your Steam cluster has R installed—if not, ask your admin to add it via the Steam Admin panel first).
• Upload your R script (e.g., api.R) and your dependencies file.
• Set the Entry Point to your script filename (Steam will execute this when starting the app).
• Configure resource limits (CPU, memory) based on your code’s needs, then hit Deploy.

4. Get Your API Endpoint #

After deployment finishes, go to your app’s details page in Steam. You’ll see a base URL for your API (e.g., http://<steam-cluster-ip>/apps/<your-app-id>).

If you used plumber, your endpoint will be the path you defined—so in the example above, it would be http://<steam-cluster-ip>/apps/<your-app-id>/compute-average.

5. Test the API #

Use curl or Postman to verify it works:

curl -X POST -H "Content-Type: application/json" -d '{"values": [2,4,6,8,10]}' http://<your-api-endpoint>

You should get a response like:

{"average":6}

6. Integrate with Production Apps #

Now that your API is live, you can call it from any production application just like any other REST API—send POST requests with JSON payloads and parse the JSON responses.

内容的提问来源于stack exchange，提问作者Nguyễn Thanh Tú