Sensible

https://docs.sensible.so
Admin
Sensible is a document extraction platform that automatically extracts information from documents...
Tokens:399,844
Snippets:2,353
Trust Score:8.4
Update:10 months ago
Show doc for...
Tokens:
Raw
### Extract Document Data using cURL

Source: https://docs.sensible.so/docs/quickstart

This snippet shows how to download an example PDF document and send it to the Sensible.so API for data extraction using cURL. It requires a Sensible.so API key and specifies the document type for extraction. The output is a JSON object containing the extracted data.

```shell
curl -L https://raw.githubusercontent.com/sensible-hq/sensible-docs/v0/assets/pdfs/1_extract_your_first_data.pdf --output 1_extract_your_first_data.pdf && curl --request POST --url "https://api.sensible.so/v0/extract/layout_basics" --header "Authorization: Bearer YOUR_API_KEY" --header "Content-Type: application/pdf" --data-binary "@1_extract_your_first_data.pdf"
```

--------------------------------

### Example API Response for Document Extraction

Source: https://docs.sensible.so/docs/quickstart

This JSON object represents a successful API response from Sensible.so after extracting data from a document. It showcases the structure of the extracted information within the 'parsed_document' field, including a sample 'your_first_extracted_field'.

```json
{
   "parsed_document":{
      "your_first_extracted_field":{
         "type":"string",
         "value":"Welcome to your first document"
      }
   }
}
```

--------------------------------

### Anchor Object Configuration Example in JSON

Source: https://docs.sensible.so/docs/anchor

This JSON snippet demonstrates the structure of an Anchor object, illustrating the use of 'start', 'end', 'includeEnd', and 'match' parameters for defining search boundaries and criteria within a document. It's useful for configuring how Sensible SO extracts specific data.

```json
{
  "fields": [
    {
      "id": "simple_label",
      "anchor": {
        "start": "My section heading. Start matching at the start of this line",
        "end": "My footer text. Stop matching before it",
        "includeEnd": true,
        "match": 
          [
            {
              "type": "includes",
              "text": "Only finds anchor if you match this string in a line that is between the start and end lines",
            },
          ]      
      },
      "method": {
        "id": "label",
        "position": "below"
      }
    }
  ]
}
```

--------------------------------

### List Configurations Response Example

Source: https://docs.sensible.so/reference/configuration

Example JSON response when listing configurations for a document type. Each configuration object includes its name, creation timestamp, and version details.

```json
[
  {
    "name": "anyco_auto_insurance_quote",
    "created": "2025-12-09T10:03:50.203Z",
    "versions": [
      {
        "version_id": "fdE0LO4d2GftiGvPyeYbgpstRWyLrEdm",
        "datetime": "2025-12-09T10:03:50.203Z",
        "environments": [
          "production",
          "development"
        ],
        "draft": false
      }
    ]
  }
]
```

--------------------------------

### Classification Summary Example - JSON

Source: https://docs.sensible.so/reference/retrieving-results

Provides example data for the 'ClassificationSummary'. This shows how different configurations are evaluated based on fingerprints, fingerprints present, and their associated scores.

```json
[
  {
    "configuration": "config_for_x_company",
    "fingerprints": 2,
    "fingerprints_present": 2,
    "score": {
      "value": 3,
      "fields_present": 4,
      "penalities": 0.5
    }
  },
  {
    "configuration": "acme_co",
    "fingerprints": 2,
    "fingerprints_present": 2,
    "score": {
      "value": 0,
      "fields_present": 2,
      "penalities": 1.5
    }
  }
]
```

--------------------------------

### Anchor Configuration: Match Array vs. Start and Match

Source: https://docs.sensible.so/docs/match-arrays

Demonstrates two ways to configure anchors in JSON: using a 'match' array for sequential pattern matching and using separate 'start' and 'match' parameters for anchored searching. The 'match' array finds instances where patterns occur in sequence anywhere in the document. The 'start' and 'match' parameters find a 'match' pattern only after the 'start' pattern has been encountered, discarding earlier document content.

```json
{
  "fields": [
    {
      "id": "match_array",
      "match": "all",
      "anchor": {
        "match": [
          {
            "type": "startsWith",
            "text": "here is an A line"
          },
          {
            "type": "startsWith",
            "text": "here is a B line"
          }
        ]
      },
      "method": {
        "id": "passthrough"
      }
    },
    {
      "id": "start_and_match",
      "match": "all",
      "anchor": {
        "start": {
          "type": "startsWith",
          "text": "here is an A line"
        },
        "match": {
          "type": "startsWith",
          "text": "here is a B line"
        }
      },
      "method": {
        "id": "passthrough"
      }
    }
  ]
}
```

--------------------------------

### Example Output Structure (JSON)

Source: https://docs.sensible.so/docs/pick-values

This JSON structure represents an example output after processing. It categorizes business classification and lists income-related answers, demonstrating how the configuration might be applied to structure data.

```json
{
  "business_classification": {
    "value": "partnership",
    "type": "string"
  },
  "income_yes_answers": [
    {
      "value": "income_over_100k",
      "type": "string"
    },
    {
      "value": "professional_services",
      "type": "string"
    }
  ]
}
```

--------------------------------

### Mapper Method Example

Source: https://docs.sensible.so/docs/mapper

An example demonstrating how to use the Mapper method to standardize a time duration from a raw input field to a numerical string representation.

```APIDOC
## Mapper Method Example: Standardizing Premium Period

### Description

This example configures a computed field `premium_period_months` that uses the Mapper method to convert various representations of a premium period (e.g., "Six months", "ANNUAL") into a standardized numerical string (e.g., "6", "12").

### Configuration

```json
{
  "fields": [
    {
      "id": "_premium_period_raw",
      "anchor": "duration",
      "method": {
        "id": "row"
      }
    }
  ],
  "computed_fields": [
    {
      "id": "premium_period_months",
      "method": {
        "id": "mapper",
        "source_id": "_premium_period_raw",
        "mappings": {
          "ANNUAL": "12",
          "BIANNUAL": "6",
          "MONTHLY": "1",
          "Six months": "6",
          "six months": "6",
          "Twelve months": "12",
          "twelve months": "12"
        },
        "default": "mapping doesn't exist"
      }
    }
  ]
}
```

### Example Document Input

Assume a document contains the value "Six months" for the "duration" field.

### Output Example

For the given document input and configuration, the output would be:

```json
{
  "_premium_period_raw": {
    "type": "string",
    "value": "Six months"
  },
  "premium_period_months": {
    "value": "6",
    "type": "string"
  }
}
```
```

--------------------------------

### Example Output with Annotations in JSON

Source: https://docs.sensible.so/docs/types

This JSON example demonstrates the output when 'annotateSuperscriptAndSubscript' is enabled. Footnote symbols are formatted as '[^1]' and '[^2]', and text continues appropriately.

```json
{
  "lease_duration": {
    "type": "string",
    "value": "12/31/2023 [^1] . Thereafter, it shall be month-to-month on the same terms and conditions as stated herein plus $80.00 month to month charge, save any changes made pursuant to law, until terminated by notice of at least 30 days. [^2]\n[^1] Landlord shall send notice of new terms 30 days before current terms end. [^2] Proper 30 day notice (in writing or email) must be received by the Landlord."
  }
}
```

--------------------------------

### Generate Excel Extraction (Node.js)

Source: https://docs.sensible.so/reference/get-excel-from-documents

This example demonstrates how to generate an Excel file from Sensible extractions using Node.js. It utilizes the `axios` library to make a GET request to the API. The extracted data is converted into an Excel spreadsheet, and the response contains a URL to download the file.

```javascript
const axios = require('axios');

async function generateExcel(ids) {
  try {
    const response = await axios.get(`https://api.sensible.so/v0/generate_excel/${ids}`, {
      headers: {
        'accept': 'application/json'
      }
    });
    return response.data;
  } catch (error) {
    console.error('Error generating Excel:', error);
    throw error;
  }
}

// Example usage:
// const extractionIds = 'uuid1,uuid2';
// generateExcel(extractionIds).then(data => console.log(data.url)).catch(err => console.error(err));
```

--------------------------------

### Successful Response for Auth Token

Source: https://docs.sensible.so/reference/review-extractions

Example JSON response when successfully obtaining an authorization token for reviewer magic links. It includes the token, creation/expiration dates, and granted permissions.

```json
{
  "token": "09b33af4-6811-4d8e-8c56-2f197b8648d6:4174564c-2211-4e42-bcc4-66e025b29728",
  "created": "2025-01-01T18:43:45.532Z",
  "expires": "2025-01-15T22:14:35.720Z",
  "grants": [
    {
      "route": "/documents/{id}",
      "method": "GET",
      "path": {
        "id": "b84bd1c8-113e-4e1e-8462-379f0dde2abf"
      }
    },
    {
      "route": "/documents/{id}",
      "method": "PUT",
      "path": {
        "id": "b84bd1c8-113e-4e1e-8462-379f0dde2abf"
      }
    }
  ],
  "usage": []
}
```

--------------------------------

### Website Sensible.so Parameters

Source: https://docs.sensible.so/docs/deprecated-summarizer

Documentation for the 'instructions' and 'samples' parameters used in the /websites/sensible_so project.

```APIDOC
## Website Sensible.so API

### Description
This section details the parameters available for the /websites/sensible_so project, focusing on how to guide and refine information extraction using instructions and examples.

### Method
N/A (This describes parameters, not a specific endpoint)

### Endpoint
/websites/sensible_so

### Parameters
#### Query Parameters
- **instructions** (string) - Required - A prompt for the LLM, describing how to transform information from the text in the Source ID parameter. For more information about how to write prompts, see [Query Group](https://docs.sensible.so/docs/query-group) tips. For an example of using this parameter, see the Examples section.
- **samples** (object array) - Optional - Short snippets of text similar to the text in the Source ID parameter, with examples of the information to extract. Use in addition to the Instructions parameter to increase the LLM's accuracy. Contains these parameters:
  - **prompt** (string) - Required - An example of the sort of free text that you want to transform.
  - **values** (string array) - Required - The target information to transform from this prompt. This array is a parallel array to the Fields parameter's array. Parallel arrays are the same length and same sequence. If the LLM can't find the target information in the Source ID parameter, it can generate an arbitrary value. To override this behavior, specify a Sample parameter whose Prompt parameter has a text snippet that's missing the target data, and whose Values array indicates the data is missing (for example, "N/A" or "not found"). For an example of using this parameter, see the Examples section.

### Request Example
```json
{
  "instructions": "Extract the company name and its founding year from the provided text.",
  "samples": [
    {
      "prompt": "Acme Corporation was founded in 1985 and has been a leader in innovation ever since.",
      "values": ["Acme Corporation", "1985"]
    },
    {
      "prompt": "Globex Inc., established in 1992, specializes in cutting-edge technology.",
      "values": ["Globex Inc.", "1992"]
    }
  ]
}
```

### Response
#### Success Response (200)
- **extracted_data** (object) - An object containing the extracted information based on the instructions and samples provided.

#### Response Example
```json
{
  "extracted_data": {
    "company_name": "Acme Corporation",
    "founding_year": "1985"
  }
}
```
```

--------------------------------

### Concise Repeat Match for nth Occurrence

Source: https://docs.sensible.so/changelog/august-2025

The 'repeat' match object provides a concise alternative to match arrays for finding the nth occurrence of a string or regular expression. This example shows how to match the fifth line starting with 'customer account'.

```json
{
  "match": [
    {
      "type": "repeat",
      "times": 5,
      "match": { "type": "startsWith", "text": "customer account" }
    }
  ]
}
```

--------------------------------

### Configuration Example JSON String

Source: https://docs.sensible.so/reference/publish-configuration-by-version

This snippet shows an example of a JSON string used to define configuration parameters. It includes settings for fingerprints, preprocessors, fields, and computed fields. This structure is used to customize how data is processed and extracted.

```json
{
	"fingerprint": {
		"tests": [
			"anyco",
			"quoted coverage changes"
		]
	},
	"preprocessors": [{
		"type": "pageRange",
		"startPage": 0,
		"endPage": 2
	}],
	"fields": [{
		"id": "_driver_name_raw",
		"anchor": "name of driver",
		"method": {
			"id": "label",
			"position": "below"
		}
	}],
	"computed_fields": [{
		"id": "driver_name_last",
		"method": {
			"id": "split",
			"source_id": "_driver_name_raw",
			"separator": ", ",
			"index": 1
		}
	}]
}
```

--------------------------------

### GET /websites/sensible_so/configurations

Source: https://docs.sensible.so/reference/create-configuration

Retrieves a list of configurations for a given website. Includes details about versions and environments.

```APIDOC
## GET /websites/sensible_so/configurations

### Description
Retrieves a list of all configurations associated with the website.
This endpoint returns detailed information about each configuration version, including its ID, creation timestamp, and associated environments.

### Method
GET

### Endpoint
`/websites/sensible_so/configurations`

### Parameters
#### Query Parameters
None

### Request Example
None

### Response
#### Success Response (200)
- **name** (string) - The name of the configuration.
- **created** (string) - The date and time the configuration was created.
- **configuration** (object) - The configuration details, including fingerprint, preprocessors, fields, and computed fields. The actual configuration is returned as a stringified JSON.
  - **fingerprint** (object) - Defines tests for the configuration.
    - **tests** (array of strings) - A list of tests to be applied.
  - **preprocessors** (array of objects) - Rules for processing pages.
    - **type** (string) - The type of preprocessor.
    - **startPage** (integer) - The starting page for the preprocessor.
    - **endPage** (integer) - The ending page for the preprocessor.
  - **fields** (array of objects) - Defines the fields to be extracted.
    - **id** (string) - The unique identifier for the field.
    - **anchor** (string) - A reference point for extracting the field.
    - **method** (object) - The method used for extraction.
      - **id** (string) - The identifier of the extraction method.
      - **position** (string) - The position relative to the anchor.
  - **computed_fields** (array of objects) - Defines fields that are computed based on other fields.
    - **id** (string) - The unique identifier for the computed field.
    - **method** (object) - The method used for computation.
      - **id** (string) - The identifier of the computation method.
      - **source_id** (string) - The ID of the source field.
      - **separator** (string) - The separator used in the splitting method.
      - **index** (integer) - The index of the split part to use.
- **version_id** (string) - The current version ID of the configuration.
- **versions** (array of objects) - A list of available versions for the configuration.
  - **version_id** (string) - The unique identifier for a specific version.
  - **datetime** (string) - The date and time this version was created.
  - **environments** (array of strings) - The environments where this version is deployed.
  - **draft** (boolean) - Indicates if this version is a draft.

#### Response Example
```json
{
  "name": "Example Configuration",
  "created": "2023-10-27T10:00:00Z",
  "configuration": {
    "fingerprint": {
      "tests": [
        "anyco",
        "quoted coverage changes"
      ]
    },
    "preprocessors": [
      {
        "type": "pageRange",
        "startPage": 0,
        "endPage": 2
      }
    ],
    "fields": [
      {
        "id": "_driver_name_raw",
        "anchor": "name of driver",
        "method": {
          "id": "label",
          "position": "below"
        }
      }
    ],
    "computed_fields": [
      {
        "id": "driver_name_last",
        "method": {
          "id": "split",
          "source_id": "_driver_name_raw",
          "separator": ", ",
          "index": 1
        }
      }
    ]
  },
  "version_id": "v1.0.0",
  "versions": [
    {
      "version_id": "fdE0LO4d2GftiGvPyeYbgpstRWyLrEdm",
      "datetime": "2023-10-27T10:00:00Z",
      "environments": [
        "production",
        "development"
      ],
      "draft": false
    }
  ]
}
```

#### Error Response (400)
- **description**: Bad Request
- **content**: text/plain
  - **schema**: string (Bad Request)
```

--------------------------------

### Retrieve Extraction by ID using JavaScript (Node.js)

Source: https://docs.sensible.so/reference/retrieve-extractions

This Node.js snippet shows how to fetch extraction results using the Sensible API. It makes a GET request to the documents endpoint with the extraction ID and handles the JSON response. Ensure you have the 'node-fetch' or a similar library installed.

```javascript
const fetch = require('node-fetch');

async function getExtraction(id) {
  const response = await fetch(`https://api.sensible.so/v0/documents/${id}`, {
    method: 'GET',
    headers: {
      'accept': 'application/json'
    }
  });
  return await response.json();
}

// Example usage:
// getExtraction('your-document-id').then(data => console.log(data));
```

--------------------------------

### Summarizer Method Example - Free Text Question

Source: https://docs.sensible.so/changelog/december-2022

Demonstrates how to use the Summarizer method with the new Instructions parameter to ask a free-text question and extract specific data points from a document. It shows the expected JSON output for rent details.

```json
{
  "rent_computed": [
    {
      "rent_in_dollars": "$895.00",
      "payment_time_period": "month",
      "payment_due": "on or before the 1st day of each month"
    }
  ]
}
```

--------------------------------

### Example Output for Menu Item Extraction (JSON)

Source: https://docs.sensible.so/docs/list-tips

This JSON output demonstrates the result of applying the configurations for extracting dinner, dessert, and wine information from a menu. It shows the structured data for each category, including descriptions and prices, formatted as requested.

```json
{
  "dinners_zipped": [
    {
      "dinner_description": {
        "value": "Grilled salmon with almond rice pilaf and spinach with pine nuts",
        "type": "string"
      },
      "price": {
        "source": "$45",
        "value": 45,
        "unit": "$",
        "type": "currency"
      }
    },
    {
      "dinner_description": {
        "value": "Seared beef with fingerling potatoes and asparagus",
        "type": "string"
      },
      "price": {
        "source": "$56",
        "value": 56,
        "unit": "$",
        "type": "currency"
      }
    }
  ],
  "desserts_zipped": [
    {
      "dessert_description": {
        "value": "Cannoli cake with candied lavender",
        "type": "string"
      },
      "price": {
        "source": "$14",
        "value": 14,
        "unit": "$",
        "type": "currency"
      }
    },
    {
      "dessert_description": {
        "value": "Creme brulee",
        "type": "string"
      },
      "price": {
        "source": "$13",
        "value": 13,
        "unit": "$",
        "type": "currency"
      }
    }
  ],
  "wines_zipped": [
    {
      "wine_name": {
        "value": "House white",
        "type": "string"
      },
      "wine_type": {
        "value": "white:chardonnay",
        "type": "string"
      },
      "wine_description": {
        "value": "Crisp and refreshing with notes of citrus and green apple.",
        "type": "string"
      },
      "smallest_serving_price": {
        "source": "6 oz: $11",
        "value": 11,
        "unit": "$",
        "type": "currency"
      },
      "second_smallest_serving_price": {
        "source": "9 oz: $15",
        "value": 15,
        "unit": "$",
        "type": "currency"
      },
      "bottle_price": {
        "source": "$42",
        "value": 42,
        "unit": "$",
        "type": "currency"
      }
    }
  ]
}
```

--------------------------------

### Retrieve Extraction by ID using Python

Source: https://docs.sensible.so/reference/retrieve-extractions

This Python example illustrates how to retrieve an extraction by its ID using the Sensible API. It utilizes the 'requests' library to send a GET request and process the JSON response. The function returns the parsed document data or raises an exception for non-200 status codes.

```python
import requests

def get_extraction(doc_id):
    url = f"https://api.sensible.so/v0/documents/{doc_id}"
    headers = {
        "accept": "application/json"
    }
    response = requests.get(url, headers=headers)
    response.raise_for_status()  # Raise an exception for bad status codes
    return response.json()

# Example usage:
# try:
#     data = get_extraction('your-document-id')
#     print(data)
# except requests.exceptions.RequestException as e:
#     print(f"Error: {e}")
```

--------------------------------

### Example Output: Anchoring Results

Source: https://docs.sensible.so/docs/match-arrays

Presents the JSON output generated by the Sensible configuration for both 'match_array' and 'start_and_match' anchoring methods. It shows the extracted values for each field ID, illustrating the difference in results based on the anchoring strategy used.

```json
{
  "match_array": [
    {
      "type": "string",
      "value": "Here is a B line."
    },
    {
      "type": "string",
      "value": "Here is a B line that is the 2nd occurrence of \u201CB following A\u201D in the doc"
    }
  ],
  "start_and_match": [
    {
      "type": "string",
      "value": "Here is a B line."
    },
    {
      "type": "string",
      "value": "Here is a B line again."
    },
    {
      "type": "string",
      "value": "Here is a B line for the 4th time."
    },
    {
      "type": "string",
      "value": "Here is a B line that is the 2nd occurrence of \u201CB following A\u201D in the doc"
    }
  ]
}
```

--------------------------------

### Get Document by ID (GET)

Source: https://docs.sensible.so/reference/account-auth-tokens

Retrieves a specific document using its unique identifier. Requires a document ID in the path.

```json
{
  "route": "/documents/{id}",
  "method": "GET",
  "path": {
    "id": "b84bd1c8-113e-4e1e-8462-379f0dde2abf"
  }
}
```

--------------------------------

### Extract from Images Example

Source: https://docs.sensible.so/docs/query-group-tips

Example configuration for extracting structured data from images within a document using the Multimodal Engine.

```APIDOC
## Example: Extract from images

### Description
This example demonstrates extracting structured data from real estate photographs embedded in an offering memorandum document using the Multimodal Engine parameter, alongside extracting text data.

### Method
POST (Illustrative for configuration structure)

### Endpoint
/extract (Illustrative)

### Parameters
#### Request Body
- **fields** (array) - Required - An array of field configurations.
  - **method** (object) - Required - Defines the extraction method.
    - **id** (string) - Required - Identifier, e.g., `"queryGroup"`.
    - **searchBySummarization** (string) - Optional - `"page"` for page-level summarization.
    - **multimodalEngine** (object) - Required for image extraction - Configures the multimodal LLM.
      - **region** (string) - Required - `"automatic"` to automatically detect image regions. Include text queries targeting text near the image to help locate context.
    - **confidenceSignals** (boolean) - Optional - `false` as multimodal engine doesn't support confidence signals.
    - **queries** (array) - Required - A list of queries for data extraction.
      - **id** (string) - Required - Unique identifier for the query (e.g., `"trees_present"`).
      - **description** (string) - Required - A natural language description of the data to extract.
      - **type** (string) - Required - The expected data type (e.g., `"string"`).

### Request Example
```json
{
  "fields": [
    {
      "method": {
        "id": "queryGroup",
        "searchBySummarization": "page",
        "multimodalEngine": {
          "region": "automatic"
        },
        "confidenceSignals": false,
        "queries": [
          {
            "id": "trees_present",
            "description": "are there trees on the property? respond true or false",
            "type": "string"
          },
          {
            "id": "multistory",
            "description": "are the buildings multistory? return true or false",
            "type": "string"
          },
          {
            "id": "community_amenities",
            "description": "give one example of a community amenity listed",
            "type": "string"
          },
          {
            "id": "exterior",
            "description": "what is the exterior of the building made of (walls, not roof)?",
            "type": "string"
          }
        ]
      }
    }
  ]
}
```

### Response
#### Success Response (200)
- **field_id** (object) - Contains the extracted value and type.
  - **value** (string) - The extracted data.
  - **type** (string) - The data type of the extracted value.

#### Response Example
```json
{
  "trees_present": {
    "value": "true",
    "type": "string"
  },
  "multistory": {
    "value": "true",
    "type": "string"
  },
  "community_amenities": {
    "value": "Playground",
    "type": "string"
  },
  "exterior": {
    "value": "brick",
    "type": "string"
  }
}
```
```

--------------------------------

### Configuration Parameters

Source: https://docs.sensible.so/docs/query

This section details various parameters that can be configured to control how Sensible interacts with LLMs and processes documents for context.

```APIDOC
## Configuration Parameters

### `default`

- **Type**: boolean
- **Description**: If true, Sensible prompts the LLM to report any uncertainties about the accuracy of its response. If set to `"strict"`, Sensible returns null for a field if its confidence signal is `incorrect_answer`. Not supported if you specify the Multimodal Engine parameter or Source Ids parameter.

### `searchBySummarization`

- **Type**: boolean, `"outline"`, or `"page"` (equivalent to `true`)
- **Default**: `false`
- **Description**: Configures context searching using summaries of document chunks. If set to `page`, each page is a chunk. If set to `outline`, an LLM outlines the document, and each segment of the outline is a chunk. Compatible with documents up to 1,280 pages long. For documents 5 pages or under, the entire document is submitted as context. For documents over 5 pages, `chunkCount` is set to 5 and this parameter's value is ignored.

### `pageRange`

- **Type**: object
- **Description**: Configures the possible page range for finding context in the document. Sensible creates chunks only within the specified page range. Contains `startPage` (zero-based index, inclusive) and `endPage` (zero-based index, exclusive).
- **Note**: Sensible ignores this parameter when searching for a field's anchor. Use the Page Range preprocessor for that.

### `chunkCount`

- **Type**: number
- **Default**: 5
- **Description**: The number of top-scoring document chunks Sensible combines as context for the LLM prompt.
```

--------------------------------

### GET /documents/{id}

Source: https://docs.sensible.so/reference/account-auth-tokens

Retrieves details for a specific document using its ID.

```APIDOC
## GET /documents/{id}

### Description
Retrieves details for a specific document.

### Method
GET

### Endpoint
`/documents/{id}`

### Parameters
#### Path Parameters
- **id** (string) - Required - The unique identifier of the document.

### Response
#### Success Response (200)
- **id** (string) - The document ID.

#### Response Example
{
  "example": {
    "id": "b84bd1c8-113e-4e1e-8462-379f0dde2abf"
  }
}
```

--------------------------------

### Extract from Images Example

Source: https://docs.sensible.so/docs/query

This example demonstrates extracting structured data from both images and text within a document using the Multimodal Engine.

```APIDOC
## POST /api/extract

### Description
Extracts structured data from documents, including information from images and text, leveraging multimodal LLM capabilities.

### Method
POST

### Endpoint
`/api/extract`

### Request Body

- **fields** (array) - Required - An array of field configurations for extraction.
  - **method** (object) - Required - Defines the extraction method.
    - **id** (string) - Required - Identifier for the query group, e.g., `"queryGroup"`.
    - **searchBySummarization** (string) - Optional - Method for searching context (`"page"`, `"outline"`, or `true`).
    - **multimodalEngine** (object) - Optional - Configuration for using a multimodal LLM.
      - **region** (string) - Required if `multimodalEngine` is used - Specifies the region for image context, e.g., `"automatic"`.
    - **confidenceSignals** (boolean) - Optional - If `false`, confidence signals are disabled (not supported with `multimodalEngine`).
    - **queries** (array) - Required - A list of queries for data extraction.
      - **id** (string) - Required - Unique identifier for the query.
      - **description** (string) - Required - A natural language description of what to extract.
      - **type** (string) - Required - The expected data type of the extracted value (e.g., `"string"`).

### Request Example
```json
{
  "fields": [
    {
      "method": {
        "id": "queryGroup",
        "searchBySummarization": "page",
        "multimodalEngine": {
          "region": "automatic"
        },
        "confidenceSignals": false,
        "queries": [
          {
            "id": "trees_present",
            "description": "are there trees on the property? respond true or false",
            "type": "string"
          },
          {
            "id": "multistory",
            "description": "are the buildings multistory? return true or false",
            "type": "string"
          },
          {
            "id": "community_amenities",
            "description": "give one example of a community amenity listed",
            "type": "string"
          },
          {
            "id": "exterior",
            "description": "what is the exterior of the building made of (walls, not roof)?",
            "type": "string"
          }
        ]
      }
    }
  ]
}
```

### Response

#### Success Response (200)
- **field_id** (object) - Contains the extracted value and type.
  - **value** (string) - The extracted data.
  - **type** (string) - The data type of the extracted value.

#### Response Example
```json
{
  "trees_present": {
    "value": "true",
    "type": "string"
  },
  "multistory": {
    "value": "true",
    "type": "string"
  },
  "community_amenities": {
    "value": "Playground",
    "type": "string"
  },
  "exterior": {
    "value": "brick",
    "type": "string"
  }
}
```
```

--------------------------------

### Configuration Response Schema Example

Source: https://docs.sensible.so/reference/update-configuration

This snippet illustrates the structure of a successful configuration response. It includes the configuration name, creation date, the configuration itself as a stringified JSON object, a version ID, and an array of available versions with their respective details.

```json
{
  "name": "anyco_auto_insurance_quote",
  "created": "2023-10-26T20:05:33.000Z",
  "configuration": "{\r\n\t\"fingerprint\": {\r\n\t\t\"tests\": [\r\n\t\t\t\"anyco\",\r\n\t\t\t\"quoted coverage changes\"\r\n\t\t]\r\n\t},\r\n\t\"preprocessors\": [{\r\n\t\t\"type\": \"pageRange\",\r\n\t\t\"startPage\": 0,\r\n\t\t\"endPage\": 2\r\n\t}],\r\n\t\"fields\": [{\r\n\t\t\"id\": \"_driver_name_raw\",\r\n\t\t\"anchor\": \"name of driver\",\r\n\t\t\"method\": {\r\n\t\t\t\"id\": \"label\",\r\n\t\t\t\"position\": \"below\"\r\n\t\t}\r\n\t}],\r\n\t\"computed_fields\": [{\r\n\t\t\t\"id\": \"driver_name_last\",\r\n\t\t\t\"method\": {\r\n\t\t\t\t\"id\": \"split\",\r\n\t\t\t\t\"source_id\": \"_driver_name_raw\",\r\n\t\t\t\t\"separator\": \", \",\r\n\t\t\t\t\"index\": 1\r\n\t\t\t}\r\n\t\t}\r\n\r\n\t]\r\n}",
  "version_id": "exampleVersionId123",
  "versions": [
    {
      "version_id": "fdE0LO4d2GftiGvPyeYbgpstRWyLrEdm",
      "datetime": "2023-10-26T20:05:33.000Z",
      "environments": [
        "production",
        "development"
      ],
      "draft": false
    }
  ]
}
```

--------------------------------

### Extracted Data from Labeled Rows Table Example

Source: https://docs.sensible.so/docs/sections-example-labeled-rows

This JSON output represents the data extracted by the provided Sensible.so configuration from an example document. It shows the structured data for 'fruit_name', 'fiber' (as a number), and '_everything_in_this_section' for multiple fruit entries.

```json
{
  "nutrition_table": [
    {
      "fruit_name": {
        "value": "Apple",
        "type": "string"
      },
      "fiber": {
        "source": "2.4",
        "value": 2.4,
        "type": "number"
      },
      "_everything_in_this_section": {
        "type": "string",
        "value": "Nutrition Apple Notes Calories 50 Per 100 g Fiber (g) 2.4 Skin on, if applicable Protein (g) 0.5"
      }
    },
    {
      "fruit_name": {
        "value": "Banana",
        "type": "string"
      },
      "fiber": {
        "source": "3.1",
        "value": 3.1,
        "type": "number"
      },
      "_everything_in_this_section": {
        "type": "string",
        "value": "Nutrition Banana Notes Calories 90 Per 100 g Fiber (g) 3.1 Skin on, if applicable Protein (g) 1.3"
      }
    },
    {
      "fruit_name": {
        "value": "Cantelope",
        "type": "string"
      },
      "fiber": {
        "source": "0.9",
        "value": 0.9,
        "type": "number"
      },
      "_everything_in_this_section": {
        "type": "string",
        "value": "Nutrition Cantelope Notes Calories 34 Per 100 g Fiber (g) 0.9 Skin on, if applicable Protein (g) 0.8"
      }
    }
  ]
}
```

--------------------------------

### Example Output for Image Resolution Troubleshooting (JSON)

Source: https://docs.sensible.so/docs/query-group-tips

This JSON output shows the results of the 'Troubleshoot Image Resolution' configuration. It contrasts the extracted kwh values when 'onlyImages' is true (higher accuracy) versus false (lower accuracy). It also shows the 'daily_cost' query returning null due to the 'onlyImages: true' setting in the configuration.

```JSON
{
  "bar_graph_original_resolution": {
    "value": "452,378,250,2091,1581,778,672,463,632,677,571,474,461",
    "type": "string"
  },
  "bar_graph_sensible_resolution": {
    "value": "2091,1581,778,672,463,632,677,571,474,461",
    "type": "string"
  },
  "daily_cost": null
}
```

--------------------------------

### Stringified JSON Configuration Example (JavaScript)

Source: https://docs.sensible.so/reference/create-configuration

This snippet demonstrates a configuration returned as a stringified JSON object. It includes settings for fingerprinting, preprocessing, field extraction, and computed fields. This format is often used for configuration data passed as a single string.

```javascript
const configurationString = "{\r\n\t\"fingerprint\": {\r\n\t\t\"tests\": [\r\n\t\t\t\"anyco\",\r\n\t\t\t\"quoted coverage changes\"\r\n\t\t]\r\n\t},\r\n\t\"preprocessors\": [{\r\n\t\t\"type\": \"pageRange\",\r\n\t\t\"startPage\": 0,\r\n\t\t\"endPage\": 2\r\n\t}],\r\n\t\"fields\": [{\r\n\t\t\"id\": \"_driver_name_raw\",\r\n\t\t\"anchor\": \"name of driver\",\r\n\t\t\"method\": {\r\n\t\t\t\"id\": \"label\",\r\n\t\t\t\"position\": \"below\"\r\n\t\t}\r\n\t}],\r\n\t\"computed_fields\": [{\r\n\t\t\t\"id\": \"driver_name_last\",\r\n\t\t\t\"method\": {\r\n\t\t\t\t\"id\": \"split\",\r\n\t\t\t\t\"source_id\": \"_driver_name_raw\",\r\n\t\t\t\t\"separator\": \", \",\r\n\t\t\t\t\"index\": 1\r\n\t\t\t}\r\n\t\t}\r\n\r\n\t]\r\n}";
console.log(JSON.parse(configurationString));
```

--------------------------------

### Sensible Instruct Methods Overview

Source: https://docs.sensible.so/changelog/april-2023

Describes the methods available within Sensible Instruct, a visual authoring tool powered by LLMs. It highlights Query, List, and Table methods as natural-language subsets of SenseML.

```markdown
* [Query](https://docs.sensible.so/docs/query-tips)
* [List](https://docs.sensible.so/docs/list-tips)
* [Table](https://docs.sensible.so/docs/table-tips)
```
Tokens:
Raw