mindee 3.16.0 → 3.18.0

data/docs/payslip_fra_v3.md

@@ -0,0 +1,319 @@
+---
+title: FR Payslip OCR Ruby
+category: 622b805aaec68102ea7fcbc2
+slug: ruby-fr-payslip-ocr
+parentDoc: 6294d97ee723f1008d2ab28e
+---
+The Ruby OCR SDK supports the [Payslip API](https://platform.mindee.com/mindee/payslip_fra).
+
+Using the [sample below](https://github.com/mindee/client-lib-test-data/blob/main/products/payslip_fra/default_sample.jpg), we are going to illustrate how to extract the data that we want using the OCR SDK.
+![Payslip sample](https://github.com/mindee/client-lib-test-data/blob/main/products/payslip_fra/default_sample.jpg?raw=true)
+
+# Quick-Start
+```rb
+require 'mindee'
+
+# Init a new client
+mindee_client = Mindee::Client.new(api_key: 'my-api-key')
+
+# Load a file from disk
+input_source = mindee_client.source_from_path('/path/to/the/file.ext')
+
+# Parse the file
+result = mindee_client.enqueue_and_parse(
+  input_source,
+  Mindee::Product::FR::Payslip::PayslipV3
+)
+
+# Print a full summary of the parsed data in RST format
+puts result.document
+
+# Print the document-level parsed data
+# puts result.document.inference.prediction
+
+```
+
+**Output (RST):**
+```rst
+########
+Document
+########
+:Mindee ID: a479e3e7-6838-4e82-9a7d-99289f34ec7f
+:Filename: default_sample.jpg
+
+Inference
+#########
+:Product: mindee/payslip_fra v3.0
+:Rotation applied: Yes
+
+Prediction
+==========
+:Pay Period:
+  :End Date: 2023-03-31
+  :Month: 03
+  :Payment Date: 2023-03-29
+  :Start Date: 2023-03-01
+  :Year: 2023
+:Employee:
+  :Address: 52 RUE DES FLEURS 33500 LIBOURNE FRANCE
+  :Date of Birth:
+  :First Name: Jean Luc
+  :Last Name: Picard
+  :Phone Number:
+  :Registration Number:
+  :Social Security Number: 123456789012345
+:Employer:
+  :Address: 1 RUE DU TONNOT 25210 DOUBS
+  :Company ID: 12345678901234
+  :Company Site:
+  :NAF Code: 1234A
+  :Name: DEMO COMPANY
+  :Phone Number:
+  :URSSAF Number:
+:Bank Account Details:
+  :Bank Name:
+  :IBAN:
+  :SWIFT:
+:Employment:
+  :Category: Cadre
+  :Coefficient: 600,000
+  :Collective Agreement: Construction -- Promotion
+  :Job Title: Directeur Régional du Développement
+  :Position Level: Niveau 5 Echelon 3
+  :Seniority Date:
+  :Start Date: 2022-05-01
+:Salary Details:
+  +--------------+-----------+--------------------------------------+--------+-----------+
+  | Amount       | Base      | Description                          | Number | Rate      |
+  +==============+===========+======================================+========+===========+
+  | 6666.67      |           | Salaire de base                      |        |           |
+  +--------------+-----------+--------------------------------------+--------+-----------+
+  | 9.30         |           | Part patronale Mutuelle NR           |        |           |
+  +--------------+-----------+--------------------------------------+--------+-----------+
+  | 508.30       |           | Avantages en nature voiture          |        |           |
+  +--------------+-----------+--------------------------------------+--------+-----------+
+:Pay Detail:
+  :Gross Salary: 7184.27
+  :Gross Salary YTD: 18074.81
+  :Income Tax Rate: 17.60
+  :Income Tax Withheld: 1030.99
+  :Net Paid: 3868.32
+  :Net Paid Before Tax: 4899.31
+  :Net Taxable: 5857.90
+  :Net Taxable YTD: 14752.73
+  :Total Cost Employer: 10486.94
+  :Total Taxes and Deductions: 1650.36
+:Paid Time Off:
+  +-----------+--------+-------------+-----------+-----------+
+  | Accrued   | Period | Type        | Remaining | Used      |
+  +===========+========+=============+===========+===========+
+  |           | N-1    | VACATION    |           |           |
+  +-----------+--------+-------------+-----------+-----------+
+  | 6.17      | N      | VACATION    | 6.17      |           |
+  +-----------+--------+-------------+-----------+-----------+
+  | 2.01      | N      | RTT         | 2.01      |           |
+  +-----------+--------+-------------+-----------+-----------+
+```
+
+# Field Types
+## Standard Fields
+These fields are generic and used in several products.
+
+### Basic Field
+Each prediction object contains a set of fields that inherit from the generic `Field` class.
+A typical `Field` object will have the following attributes:
+
+* **value** (`String`, `Float`, `Integer`, `Boolean`): corresponds to the field value. Can be `nil` if no value was extracted.
+* **confidence** (Float, nil): the confidence score of the field prediction.
+* **bounding_box** (`Mindee::Geometry::Quadrilateral`, `nil`): contains exactly 4 relative vertices (points) coordinates of a right rectangle containing the field in the document.
+* **polygon** (`Mindee::Geometry::Polygon`, `nil`): contains the relative vertices coordinates (`Point`) of a polygon containing the field in the image.
+* **page_id** (`Integer`, `nil`): the ID of the page, always `nil` when at document-level.
+* **reconstructed** (`Boolean`): indicates whether an object was reconstructed (not extracted as the API gave it).
+
+
+Aside from the previous attributes, all basic fields have access to a `to_s` method that can be used to print their value as a string.
+
+## Specific Fields
+Fields which are specific to this product; they are not used in any other product.
+
+### Bank Account Details Field
+Information about the employee's bank account.
+
+A `PayslipV3BankAccountDetail` implements the following attributes:
+
+* `bank_name` (String): The name of the bank.
+* `iban` (String): The IBAN of the bank account.
+* `swift` (String): The SWIFT code of the bank.
+Fields which are specific to this product; they are not used in any other product.
+
+### Employee Field
+Information about the employee.
+
+A `PayslipV3Employee` implements the following attributes:
+
+* `address` (String): The address of the employee.
+* `date_of_birth` (String): The date of birth of the employee.
+* `first_name` (String): The first name of the employee.
+* `last_name` (String): The last name of the employee.
+* `phone_number` (String): The phone number of the employee.
+* `registration_number` (String): The registration number of the employee.
+* `social_security_number` (String): The social security number of the employee.
+Fields which are specific to this product; they are not used in any other product.
+
+### Employer Field
+Information about the employer.
+
+A `PayslipV3Employer` implements the following attributes:
+
+* `address` (String): The address of the employer.
+* `company_id` (String): The company ID of the employer.
+* `company_site` (String): The site of the company.
+* `naf_code` (String): The NAF code of the employer.
+* `name` (String): The name of the employer.
+* `phone_number` (String): The phone number of the employer.
+* `urssaf_number` (String): The URSSAF number of the employer.
+Fields which are specific to this product; they are not used in any other product.
+
+### Employment Field
+Information about the employment.
+
+A `PayslipV3Employment` implements the following attributes:
+
+* `category` (String): The category of the employment.
+* `coefficient` (String): The coefficient of the employment.
+* `collective_agreement` (String): The collective agreement of the employment.
+* `job_title` (String): The job title of the employee.
+* `position_level` (String): The position level of the employment.
+* `seniority_date` (String): The seniority date of the employment.
+* `start_date` (String): The start date of the employment.
+Fields which are specific to this product; they are not used in any other product.
+
+### Paid Time Off Field
+Information about paid time off.
+
+A `PayslipV3PaidTimeOff` implements the following attributes:
+
+* `accrued` (Float): The amount of paid time off accrued in the period.
+* `period` (String): The paid time off period.
+
+#### Possible values include:
+ - N
+ - N-1
+ - N-2
+
+* `pto_type` (String): The type of paid time off.
+
+#### Possible values include:
+ - VACATION
+ - RTT
+ - COMPENSATORY
+
+* `remaining` (Float): The remaining amount of paid time off at the end of the period.
+* `used` (Float): The amount of paid time off used in the period.
+Fields which are specific to this product; they are not used in any other product.
+
+### Pay Detail Field
+Detailed information about the pay.
+
+A `PayslipV3PayDetail` implements the following attributes:
+
+* `gross_salary` (Float): The gross salary of the employee.
+* `gross_salary_ytd` (Float): The year-to-date gross salary of the employee.
+* `income_tax_rate` (Float): The income tax rate of the employee.
+* `income_tax_withheld` (Float): The income tax withheld from the employee's pay.
+* `net_paid` (Float): The net paid amount of the employee.
+* `net_paid_before_tax` (Float): The net paid amount before tax of the employee.
+* `net_taxable` (Float): The net taxable amount of the employee.
+* `net_taxable_ytd` (Float): The year-to-date net taxable amount of the employee.
+* `total_cost_employer` (Float): The total cost to the employer.
+* `total_taxes_and_deductions` (Float): The total taxes and deductions of the employee.
+Fields which are specific to this product; they are not used in any other product.
+
+### Pay Period Field
+Information about the pay period.
+
+A `PayslipV3PayPeriod` implements the following attributes:
+
+* `end_date` (String): The end date of the pay period.
+* `month` (String): The month of the pay period.
+* `payment_date` (String): The date of payment for the pay period.
+* `start_date` (String): The start date of the pay period.
+* `year` (String): The year of the pay period.
+Fields which are specific to this product; they are not used in any other product.
+
+### Salary Details Field
+Detailed information about the earnings.
+
+A `PayslipV3SalaryDetail` implements the following attributes:
+
+* `amount` (Float): The amount of the earning.
+* `base` (Float): The base rate value of the earning.
+* `description` (String): The description of the earnings.
+* `number` (Float): The number of units in the earning.
+* `rate` (Float): The rate of the earning.
+
+# Attributes
+The following fields are extracted for Payslip V3:
+
+## Bank Account Details
+**bank_account_details** ([PayslipV3BankAccountDetail](#bank-account-details-field)): Information about the employee's bank account.
+
+```rb
+puts result.document.inference.prediction.bank_account_details.value
+```
+
+## Employee
+**employee** ([PayslipV3Employee](#employee-field)): Information about the employee.
+
+```rb
+puts result.document.inference.prediction.employee.value
+```
+
+## Employer
+**employer** ([PayslipV3Employer](#employer-field)): Information about the employer.
+
+```rb
+puts result.document.inference.prediction.employer.value
+```
+
+## Employment
+**employment** ([PayslipV3Employment](#employment-field)): Information about the employment.
+
+```rb
+puts result.document.inference.prediction.employment.value
+```
+
+## Paid Time Off
+**paid_time_off** (Array<[PayslipV3PaidTimeOff](#paid-time-off-field)>): Information about paid time off.
+
+```rb
+for paid_time_off_elem in result.document.inference.prediction.paid_time_off do
+  puts paid_time_off_elem.value
+end
+```
+
+## Pay Detail
+**pay_detail** ([PayslipV3PayDetail](#pay-detail-field)): Detailed information about the pay.
+
+```rb
+puts result.document.inference.prediction.pay_detail.value
+```
+
+## Pay Period
+**pay_period** ([PayslipV3PayPeriod](#pay-period-field)): Information about the pay period.
+
+```rb
+puts result.document.inference.prediction.pay_period.value
+```
+
+## Salary Details
+**salary_details** (Array<[PayslipV3SalaryDetail](#salary-details-field)>): Detailed information about the earnings.
+
+```rb
+for salary_details_elem in result.document.inference.prediction.salary_details do
+  puts salary_details_elem.value
+end
+```
+
+# Questions?
+[Join our Slack](https://join.slack.com/t/mindee-community/shared_invite/zt-2d0ds7dtz-DPAF81ZqTy20chsYpQBW5g)

data/lib/mindee/client.rb

@@ -195,6 +195,46 @@ module Mindee
 
     # rubocop:enable Metrics/ParameterLists
 
+    # Sends a document to a workflow.
+    #
+    # @param input_source [Mindee::Input::Source::LocalInputSource, Mindee::Input::Source::UrlInputSource]
+    # @param document_alias [String, nil] Alias to give to the document.
+    # @param priority [Symbol, nil] Priority to give to the document.
+    # @param full_text [Boolean] Whether to include the full OCR text response in compatible APIs.
+    #  This performs a full OCR operation on the server and may increase response time.
+    #
+    # @param public_url [String, nil] A unique, encrypted URL for accessing the document validation interface without
+    # requiring authentication.
+    # @param page_options [Hash, nil] Page cutting/merge options:
+    #
+    #  * `:page_indexes` Zero-based list of page indexes.
+    #  * `:operation` Operation to apply on the document, given the `page_indexes specified:
+    #      * `:KEEP_ONLY` - keep only the specified pages, and remove all others.
+    #      * `:REMOVE` - remove the specified pages, and keep all others.
+    #  * `:on_min_pages` Apply the operation only if document has at least this many pages.
+    #
+    #
+    # @return [Mindee::Parsing::Common::WorkflowResponse]
+    def execute_workflow(
+      input_source,
+      workflow_id,
+      document_alias: nil,
+      priority: nil,
+      full_text: false,
+      public_url: nil,
+      page_options: nil
+    )
+      if input_source.is_a?(Mindee::Input::Source::LocalInputSource) && !page_options.nil? && input_source.pdf?
+        input_source.process_pdf(page_options)
+      end
+
+      workflow_endpoint = Mindee::HTTP::WorkflowEndpoint.new(workflow_id, api_key: @api_key)
+      prediction, raw_http = workflow_endpoint.execute_workflow(input_source, full_text, document_alias, priority,
+                                                                public_url)
+      Mindee::Parsing::Common::WorkflowResponse.new(Product::Generated::GeneratedV1,
+                                                    prediction, raw_http)
+    end
+
     # Load a prediction.
     #
     # @param product_class [Mindee::Inference] class of the product

data/lib/mindee/extraction/tax_extractor/tax_extractor.rb

@@ -2,6 +2,8 @@
 
 require_relative 'ocr_extractor'
 
+# rubocop:disable Metrics/ClassLength
+
 module Mindee
   module Extraction
     # Tax extractor class
@@ -72,9 +74,12 @@ module Mindee
         reconstructed_hash['code'] =
           found_hash['code'].nil? ? found_hash['code'] : found_hash['code'].sub(%r{\s*\.*\s*$}, '')
 
-        if found_hash['rate'] && found_hash['rate'] < 1 && (found_hash['rate']).positive?
-          found_hash['rate'] =
-            found_hash['rate'] * 100
+        if found_hash['rate']
+          if found_hash['rate'].abs < 1
+            found_hash['rate'] *= 10
+          elsif found_hash['rate'].abs > 100
+            found_hash['rate'] /= 10
+          end
         end
         found_hash = swap_rates_if_needed(found_hash, min_rate_percentage, max_rate_percentage)
         found_hash = decimate_rates_if_needed(found_hash)
@@ -125,18 +130,28 @@ module Mindee
       # @param found_hash [Hash] Hash of currently retrieved values
       # @return [Hash]
       def self.set_base_and_value(reconstructed_hash, found_hash)
-        if found_hash['base'].nil?
-          reconstructed_hash['base'] = found_hash['base']
-          reconstructed_hash['value'] = found_hash['value']
-        elsif found_hash['value'].nil? && found_hash['base'] < found_hash['value']
-          reconstructed_hash['base'] = found_hash['value']
-          reconstructed_hash['value'] = found_hash['base']
-        else
-          reconstructed_hash['value'] = found_hash['value']
+        base = found_hash['base']
+        value = found_hash['value']
+
+        if base && value
+          reconstructed_hash['base'], reconstructed_hash['value'] = [base, value].minmax
+        elsif base
+          reconstructed_hash['base'] = base
+        elsif value
+          reconstructed_hash['value'] = value
+          calculate_base(reconstructed_hash)
         end
+
         reconstructed_hash
       end
 
+      def self.calculate_base(hash)
+        rate = hash['rate']
+        return unless rate&.positive?
+
+        hash['base'] = hash['value'] / (rate / 100.0)
+      end
+
       # Extracts a single custom type of tax.
       # For the sake of simplicity, this only extracts the first example, unless specifically instructed otherwise.
       # @param ocr_result [Mindee::Parsing::Common::Ocr::Ocr] result of the OCR.
@@ -149,7 +164,6 @@ module Mindee
 
         tax_names.sort!
         found_hash = pick_best(extract_horizontal_tax(ocr_result, tax_names), tax_names)
-        # a tax is considered found horizontally if it has a value, otherwise it is vertical
         if found_hash.nil? || found_hash['value'].nil?
           found_hash = extract_vertical_tax(ocr_result, tax_names,
                                             found_hash)
@@ -240,14 +254,14 @@ module Mindee
         linear_pattern_percent_first = %r{
           ((?:\s*-\s*)?(?:\d*[.,])*\d+[ ]?%?|%?[ ]?(?:\s*-\s*)?(?:\d*[.,])*\d+)?[ .]?
           ([a-zA-ZÀ-ÖØ-öø-ÿ .]*[a-zA-ZÀ-ÖØ-öø-ÿ]?)[ .]?
-          ((?:\s*-\s*)?(?:\d*[.,])+\d{2,})?[ .]*
-          ((?:\s*-\s*)?(\d*[.,])*\d{2,})?
+          ((?:\s*-\s*)?(?:\d*[.,])+\d+)?[ .]*
+          ((?:\s*-\s*)?(\d*[.,])*\d+)?
         }x
         linear_pattern_percent_second = %r{
           ([a-zA-ZÀ-ÖØ-öø-ÿ .]*[a-zA-ZÀ-ÖØ-öø-ÿ]?)[ .]*
           ((?:\s*-\s*)?(?:\d*[.,])*\d+[ ]?%?|%?[ ]?(?:\s*-\s*)?(?:\d*[.,])*\d+)?[ .]?
-          ((?:\s*-\s*)?(?:\d*[.,])+\d{2,})?[ .]*
-          ((?:\s*-\s*)?(\d*[.,])*\d{2,})?
+          ((?:\s*-\s*)?(?:\d*[.,])+\d+)?[ .]*
+          ((?:\s*-\s*)?(\d*[.,])*\d+)?
         }x
         ocr_result.mvision_v1.pages.each.with_index do |page, page_id|
           page.all_lines.each do |line|
@@ -304,7 +318,7 @@ module Mindee
           page.all_words.each do |word|
             next if match_index(word.text, tax_names).nil?
 
-            reconstructed_line = ocr_result.reconstruct_vertically(word.polygon, page_id)
+            reconstructed_line = ocr_result.reconstruct_vertically(word.polygon, page_id, 0.25)
             found_hash['page_id'] = page_id if found_hash['page_id'].nil?
             found_hash['code'] = word.text.strip if found_hash['code'].nil?
             found_hash = extract_vertical_tax_values(reconstructed_line, found_hash)
@@ -316,8 +330,9 @@ module Mindee
       private_class_method :extract_percentage_from_tax, :extract_basis_and_value, :extract_tax_from_horizontal_line,
                            :extract_horizontal_tax, :extract_vertical_tax_values, :extract_vertical_tax,
                            :create_tax_field, :fix_rate, :pick_best, :calculate_score, :curate_values,
-                           :decimate_rates_if_needed, :extract_basis_and_value, :set_base_and_value, :valid_candidate?,
-                           :swap_rates_if_needed
+                           :decimate_rates_if_needed, :set_base_and_value, :valid_candidate?,
+                           :swap_rates_if_needed, :calculate_base
     end
   end
 end
+# rubocop:enable Metrics/ClassLength

data/lib/mindee/http/workflow_endpoint.rb

@@ -0,0 +1,90 @@
+# frozen_string_literal: true
+
+require 'json'
+require 'net/http'
+require_relative 'error'
+
+module Mindee
+  module HTTP
+    # Handles the routing for workflow calls.
+    class WorkflowEndpoint
+      # @return [String]
+      attr_reader :api_key
+      # @return [Integer]
+      attr_reader :request_timeout
+      # @return [String]
+      attr_reader :url
+
+      def initialize(workflow_id, api_key: '')
+        @request_timeout = ENV.fetch(REQUEST_TIMEOUT_ENV_NAME, TIMEOUT_DEFAULT).to_i
+        @api_key = api_key.nil? || api_key.empty? ? ENV.fetch(API_KEY_ENV_NAME, API_KEY_DEFAULT) : api_key
+        base_url = ENV.fetch(BASE_URL_ENV_NAME, BASE_URL_DEFAULT)
+        @url = "#{base_url.chomp('/')}/workflows/#{workflow_id}/executions"
+      end
+
+      # Sends a document to the workflow.
+      # @param input_source [Mindee::Input::Source::LocalInputSource, Mindee::Input::Source::UrlInputSource]
+      # @param document_alias [String, nil] Alias to give to the document.
+      # @param priority [Symbol, nil] Priority to give to the document.
+      # @param full_text [Boolean] Whether to include the full OCR text response in compatible APIs.
+      # @param public_url [String, nil] A unique, encrypted URL for accessing the document validation interface without
+      # requiring authentication.
+      # @return [Array]
+      def execute_workflow(input_source, full_text, document_alias, priority, public_url)
+        check_api_key
+        response = workflow_execution_req_post(input_source, document_alias, priority, full_text, public_url)
+        hashed_response = JSON.parse(response.body, object_class: Hash)
+        return [hashed_response, response.body] if ResponseValidation.valid_async_response?(response)
+
+        ResponseValidation.clean_request!(response)
+        error = Error.handle_error(@url_name, response)
+        raise error
+      end
+
+      # @param input_source [Mindee::Input::Source::LocalInputSource, Mindee::Input::Source::UrlInputSource]
+      # @param document_alias [String, nil] Alias to give to the document.
+      # @param priority [Symbol, nil] Priority to give to the document.
+      # @param full_text [Boolean] Whether to include the full OCR text response in compatible APIs.
+      # @param public_url [String, nil] A unique, encrypted URL for accessing the document validation interface without
+      # requiring authentication.
+      # @return [Net::HTTPResponse, nil]
+      def workflow_execution_req_post(input_source, document_alias, priority, full_text, public_url)
+        uri = URI(@url)
+        params = {}
+        params[:full_text_ocr] = 'true' if full_text
+        uri.query = URI.encode_www_form(params)
+
+        headers = {
+          'Authorization' => "Token #{@api_key}",
+          'User-Agent' => USER_AGENT,
+        }
+        req = Net::HTTP::Post.new(uri, headers)
+        form_data = if input_source.is_a?(Mindee::Input::Source::UrlInputSource)
+                      [['document', input_source.url]]
+                    else
+                      [input_source.read_document]
+                    end
+        form_data.push ['alias', document_alias] if document_alias
+        form_data.push ['public_url', public_url] if public_url
+        form_data.push ['priority', priority.to_s] if priority
+
+        req.set_form(form_data, 'multipart/form-data')
+
+        response = nil
+        Net::HTTP.start(uri.hostname, uri.port, use_ssl: true, read_timeout: @request_timeout) do |http|
+          response = http.request(req)
+        end
+        response
+      end
+
+      # Checks API key
+      def check_api_key
+        return unless @api_key.nil? || @api_key.empty?
+
+        raise "Missing API key. Check your Client Configuration.\n" \
+              'You can set this using the ' \
+              "'#{HTTP::API_KEY_ENV_NAME}' environment variable."
+      end
+    end
+  end
+end

data/lib/mindee/http.rb

@@ -2,3 +2,4 @@
 
 require_relative 'http/endpoint'
 require_relative 'http/error'
+require_relative 'http/workflow_endpoint'

data/lib/mindee/input/sources/base64_input_source.rb

@@ -0,0 +1,31 @@
+# frozen_string_literal: true
+
+require 'stringio'
+
+module Mindee
+  module Input
+    module Source
+      # Load a document from a base64 string.
+      class Base64InputSource < LocalInputSource
+        # @param base64_string [String]
+        # @param filename [String]
+        # @param fix_pdf [Boolean]
+        def initialize(base64_string, filename, fix_pdf: false)
+          io_stream = StringIO.new(base64_string.unpack1('m*'))
+          io_stream.set_encoding Encoding::BINARY
+          super(io_stream, filename, fix_pdf: fix_pdf)
+        end
+
+        # Overload of the same function to prevent a base64 from being re-encoded.
+        # @param close [Boolean]
+        # @return [Array<String, [String, aBinaryString ], [Hash, nil] >]
+        def read_document(close: true)
+          @io_stream.seek(0)
+          data = @io_stream.read
+          @io_stream.close if close
+          ['document', [data].pack('m'), { filename: Source.convert_to_unicode_escape(@filename) }]
+        end
+      end
+    end
+  end
+end

data/lib/mindee/input/sources/bytes_input_source.rb

@@ -0,0 +1,21 @@
+# frozen_string_literal: true
+
+require 'stringio'
+
+module Mindee
+  module Input
+    module Source
+      # Load a document from raw bytes.
+      class BytesInputSource < LocalInputSource
+        # @param raw_bytes [String]
+        # @param filename [String]
+        # @param fix_pdf [Boolean]
+        def initialize(raw_bytes, filename, fix_pdf: false)
+          io_stream = StringIO.new(raw_bytes)
+          io_stream.set_encoding Encoding::BINARY
+          super(io_stream, filename, fix_pdf: fix_pdf)
+        end
+      end
+    end
+  end
+end

data/lib/mindee/input/sources/file_input_source.rb

@@ -0,0 +1,20 @@
+# frozen_string_literal: true
+
+require 'stringio'
+
+module Mindee
+  module Input
+    module Source
+      # Load a document from a file handle.
+      class FileInputSource < LocalInputSource
+        # @param input_file [File]
+        # @param filename [String]
+        # @param fix_pdf [Boolean]
+        def initialize(input_file, filename, fix_pdf: false)
+          io_stream = input_file
+          super(io_stream, filename, fix_pdf: fix_pdf)
+        end
+      end
+    end
+  end
+end