cli.md

Using the CLI

Installation

Install the CLI via the Python package manager pip

>> $ pip install lucidtech-las-cli

Make a prediction on a document

List models that are available for predictions

>> $ las models list
{
  "models": [
    {
      "modelId": "las:model:<hex>",
      ...
    }
  ],
  "nextToken": null
}

Upload a document

>> $ las documents create invoice.pdf
{
  "documentId": "las:document:<hex>",
  "contentType": "application/pdf"
}

Run inference on the document using a model

>> $ las predictions create las:document:<hex> las:model:<hex>
{
  "documentId": "las:document:<hex>",
  "predictions": [
    ...
  ]
}

Set ground truth of document

When uploading data that will be used for training and evaluation, we need to provide a ground truth. We can then use the optional parameters --ground-truth-path or --ground-truth-fields.

>> $ las documents create invoice.pdf --ground-truth-path ground_truth.json
{
  "documentId": "las:document:<hex>",
  "contentType": "application/pdf"
  "groundTruth": [
    ...
  ]
}

In this case the ground_truth.json should be on the following format

{
    "total_amount": "299.00",
    "due_date": "2020-03-20"
}

Update an existing document

If for instance a prediction reveals incorrect values in the ground truth of a document, we can update the existing document with new ground truth values.

>> $ las documents update las:document:<hex> --ground-truth-fields total_amount=300.00 due_date=2020-02-28
{
  "documentId": "las:document:<hex>",
  "groundTruth": [
    ...
  ]
}

Create a document with consent id

{% hint style="info" %} Consent ID is an identifier you can assign to documents to keep track of document ownership for your customers. {% endhint %}

>> $ las documents create invoice.pdf --consent-id las:consent:<hex>
{
  "documentId": "las:document:<hex>",
  "contentType": "application/pdf",
  "consentId": "las:consent:<hex>"
}

Get document and download document content

>> $ las documents create invoice.pdf --consent-id las:consent:<hex>
{
  "documentId": "las:document:<hex>",
  "contentType": "application/pdf",
  "consentId": "las:consent:<hex>"
}
>> $ las documents get las:document:<hex> --download-content invoice2.pdf
{
  "documentId": "las:document:<hex>",
  "contentType": "application/pdf",
  "consentId": "las:consent:<hex>",
  "content": "XXXXXXXXX... [TRUNCATED]"
}

Revoking consent and deleting documents

Suppose we wish to delete all documents associated with a customer in our ERP database or other systems. We need to provide a consent_id to the prediction method that uniquely identifies the customer and use that consent_id to delete documents.

>> $ las consents delete las:consent:<hex>
{
  "consentId": "las:consent:<hex>",
  "documentIds": [
    ...
  ]
}

Create a batch and associate a few documents with it

Creating a batch is a way to group documents. This is useful for specifying batches of documents to use in improving the model later.

>> $ las batches create
{
  "batchId": "las:batch:<hex>",
  ...
}