Making predictions

Before you can make predictions with a model it needs to be trained and active. If you want to make predictions through the CLI, cURL or with one of the SDKs, make sure that you have the modelId for the model you would like to test. The modelId can be found under Models -> Your model -> Overview.

App

Go to Model testing and select the model you want to use. You can either use a document from one of your datasets or upload a document from your computer.

CLI

las models list # Pick an active model for testing
las documents create receipt.pdf
las predictions create <documentId> <modelId> # use the documentId from the previous step

cURL

curl -X POST 'https://api.lucidtech.ai/v1/documents' \
--header 'Content-Type: application/json' \
--header 'Authorization: Bearer eyJra...' \
--data-raw '{
    "content": "JVBERi0xLjQ...",
    "contentType": "application/pdf",
}'

curl -X POST 'https://api.lucidtech.ai/v1/predictions' \
--header 'Content-Type: application/json' \
--header 'Authorization: Bearer eyJra...' \
--data-raw '{
    "documentId": "<documentId>",
    "modelId": "<modelId>"
}'

Python

from pathlib import Path

client.list_models() # Pick an active model for testing
document = client.create_document(Path('document.pdf').read_bytes(), 'application/pdf')
client.create_prediction(document_id=document['documentId'], model_id='<modelId>')

The output should have a structure as shown below

{
  "predictionId": "<predictionId>",
  "modelId": "<modelId>",
  "documentId": "<documentId>",
  "predictions": [
    {
      "label": "total_amount",
      "value": "5154.06",
      "confidence": 0.9758162361908527
    },
    {
      "label": "purchase_date",
      "value": "2019-12-23",
      "confidence": 0.96797316745869735
    },
  ],
  "timestamp": 1629188787,
  "inferenceTime": 2.7935566902160645
}

info

See the confidence section to read more about confidence levels and how you can use them to feel confident in your model's predictions.

If you want to learn more about the basics, check out models and predictions in the Concepts section.