Introduction

Sensible extracts structured data from documents, for example PDFs of business forms, using SenseML, a JSON-formatted query language. Write SenseML to extract from your custom documents, or leverage our open-source SenseML configuration library for common business document types.

In this quickstart, use an example SenseML configuration and example PDF to get a quick "hello world" API response.

• If you instead want a guided tour of SenseML concepts, see Getting started.
• If you instead want to explore SenseML without much explanation, then sign up for a free account and check out our interactive in-app tutorials: extract_your_first_data, tables and rows, checkboxes, paragraphs, and regions, and a blank-slate challenge.

Extract example document data

To run an API call and return extracted, structured data from a downloaded example document:

1. Get an account at sensible.so.

   NOTE In the Sensible app, don't rename of the default doc type (senseml_basics) or delete the 1_extract_your_first_data config, or this example fails.

2. Copy your API key from your account page.

3. Copy the following code example into a plain-text application and replace <YOUR_API_TOKEN> with your API key:

curl -L https://github.com/sensible-hq/sensible-docs/raw/main/readme-sync/assets/v0/pdfs/1_extract_your_first_data.pdf \
  --output 1_extract_your_first_data.pdf && \
curl --request POST \
  --url "https://api.sensible.so/v0/extract/senseml_basics" \
  --header "Authorization: Bearer <YOUR_API_TOKEN>" \
  --header "Content-Type: application/pdf" \
  --data-binary "@1_extract_your_first_data.pdf"

curl -L https://github.com/sensible-hq/sensible-docs/raw/main/readme-sync/assets/v0/pdfs/1_extract_your_first_data.pdf ^
  --output 1_extract_your_first_data.pdf && ^
curl --request POST ^
  --url "https://api.sensible.so/v0/extract/senseml_basics" ^
  --header "Authorization: Bearer <YOUR_API_TOKEN>" ^
  --header "Content-Type: application/pdf" ^
  --data-binary "@1_extract_your_first_data.pdf"

3. Run the code sample in a command prompt. The code downloads an example PDF (1_extract_your_first_data.pdf ) and runs it against an example document type (senseml_basics). The API returns a parsed_document object with the extracted data, as well as metadata about the extraction, in a response like the following:

{
   "id":"153753e0-5673-466f-aa61-5175200c210d",
   "created":"2022-02-02T21:30:01.981Z",
   "status":"COMPLETE",
   "type":"senseml_basics",
   "configuration":"1_extract_your_first_data",
   "parsed_document":{
      "your_first_extracted_field":{
         "type":"string",
         "value":"Welcome to your first document"
      }
   },
   "validations":[
      
   ],
   "validation_summary":{
      "fields":1,
      "fields_present":1,
      "errors":0,
      "warnings":0,
      "skipped":0
   },
   "classification_summary":[
      {
         "configuration":"1_extract_your_first_data",
         "score":{
            "value":1,
            "fields_present":1,
            "penalties":0
         }
      },
      {
         "configuration":"2_tables_and_rows",
         "score":{
            "value":0,
            "fields_present":0,
            "penalties":0
         }
      },
      {
         "configuration":"3_checkboxes_paragraphs_and_regions",
         "score":{
            "value":0,
            "fields_present":0,
            "penalties":0
         }
      },
      {
         "configuration":"4_extract_from_scratch",
         "score":{
            "value":0,
            "fields_present":0,
            "penalties":0
         }
      }
   ],
   "errors":[
      
   ]
}

(Optional) See how it works in the Sensible app

To see this example in the Sensible app:

1. Log into the Sensible app.

2. Navigate to the first tutorial config.

3. Visually examine the example PDF (middle pane), config (left pane), and extracted data (right pane) to better understand the API call you just ran:

Next

• Learn concepts with more detailed examples in the Getting Started Guide
• Check out the SenseML method reference docs to write your own extractions
• See the API reference and example code