sys.ai

Describes ai utilities in the Javascript API.

sys.ai

This package encompasses methods for making us of ai features in your application.

embeddingsBatchProcessing(inputs,callback)

This will trigger a background job to embed a list of inputs. It will trigger a batch embedding generation on Vertex AI API and then it will execute a callback on the outputs.

See Google’s batch embeddings inputs and outputs. Context can be passed in inputs and used when executing the callback.

Parameters
NameTypeRequiredDescription
inputsobjectsyesAn array of JSON objects that will be used to generate embeddings. If id is passed the result of the callback will be mapped to this id.
callbackfunctionyesThe function to evaluate if an element needs to be removed. It will receive the element wrapper as parameter and should return true if the element has to be removed or false otherwise.
Returns

string - The ID of the triggered job.

Samples 
let record = sys.data.findOne('library',{});

// List of inputs.
let array = [{"content":"horror book","recordId":record.id(), "id":0}]

// trigger background job to process embeddings in a batch. 
let jobId = sys.jobs.embeddingsBatchProcessing(array, function(prediction) {
  // get library from context
  let library = sys.data.findOne('library',{id: prediction.instance.recordId})

  // find related books with the output embedding
  let records = sys.data.find('books', {
    _indexedFilter: 'semantic',
    _indexedFilterEmbedding: prediction.predictions[0].embeddings.values,
    _vectorSearchLimit: 3,
    _numCandidates: 200,
    _queryScoring:0.72
  });
  let ids = [];
  while (records.hasNext()) {
    let r = records.next();
    ids.push(r.id());
  
  }
  record.field('books').val(ids)
  sys.data.save(record)
  return "ok"
})

Edit this page on GitHub
← Namespaces
sys.app →