How to use Google AutoML on web app? - firebase

Question

I have a Google Cloud AutoML NL model ready to use. I wish to link this with my web app with Firebase backend. Following is my code to call. There is authorization issue. I want to understand how to authorize the app to help the client apps get access to AutoML model.
async add(data){
var headers = new Headers();
headers.append('Content-Type:application/json')
var options = {
method: 'POST',
headers,
body: JSON.stringify(data)
}
var request = new Request('https://automl.googleapis.com/v1beta1/projects/project1/locations/us-central1/models/TCN5678:predict', options )
var response = await fetch(request)
var status = await response.status
console.log(status)}

Answer 1

After struggling for hours, finally I could resolve this. I am not sure how it can work for, other than Firebase (and NL AutoML). I used Firebase Cloud Function to work around and used a hidden doc which gives access to AutoML npm. The given code require some changes. Firebase CF is able to authenticate without explicitly authorizing. Following is a suggested code and I am able to get the classification of prediction with AutoML. Hope it helps others too. Lastly, it seems Google docs is a way of testing searching skills and patience, not sure how it helps them:
const automl = require('#google-cloud/automl');
exports.sendToAML = functions.database.ref('/path/to/text').onWrite((snapshot, context) =>{
var client = new automl.PredictionServiceClient({
// optional auth parameters.
});
var formattedName = client.modelPath('bucketId', 'us-central1', 'TCN****3567595');
var payload = {
"textSnippet": {
"content": snapshot.after._data.text,
"mime_type": "text/plain"
},
};
var request = {
name: formattedName,
payload: payload,
};
client.predict(request)
.then(responses => {
var response = responses[0];
return console.log(response.payload[0].classification.score)
})
.catch(err => {
console.error(err);
});
});
`

Answer 2

You should either use service accounts or OAuth 2.0 as the authentication method. Try to avoid the use of API keys as your credentials could be exposed and misused, generating undesired charges.