namespace :generate_documentation do desc "Generate documentation for hard lettings validations" task describe_lettings_validations: :environment do client = OpenAI::Client.new(access_token: ENV["OPENAI_API_KEY"]) en_yml = File.read("./config/locales/en.yml") include Validations::SetupValidations include Validations::HouseholdValidations include Validations::PropertyValidations include Validations::FinancialValidations include Validations::TenancyValidations include Validations::DateValidations include Validations::LocalAuthorityValidations form = FormHandler.instance.forms["current_lettings"] validation_methods = public_methods.select { |method| method.starts_with?("validate_") } all_methods = [Validations::SetupValidations, Validations::HouseholdValidations, Validations::PropertyValidations, Validations::FinancialValidations, Validations::TenancyValidations, Validations::DateValidations, Validations::LocalAuthorityValidations].map { |x| x.instance_methods + x.private_instance_methods }.flatten all_helper_methods = all_methods - validation_methods validation_methods.each do |meth| if Validation.where(validation_name: meth.to_s, bulk_upload_specific: false).exists? Rails.logger.info("Validation #{meth} already exists") next end validation_source = method(meth).source helper_methods_source = all_helper_methods.map { |helper_method| if validation_source.include?(helper_method.to_s) method(helper_method).source end }.compact.join("\n") begin response = client.chat( parameters: { model: "gpt-3.5-turbo", messages: [ { role: "system", content: "You write amazing documentation, as a senior technical writer. Your audience is non-technical team members. You have been asked to document the validations in a Rails application. The application collects social housing data for different collection years. There are validations on different fields, sometimes the validations depend on several fields. Describe what given validation does, be very explicit about all the different validation cases (be specific about the years for which these validations would run, which values would be invalid or which values are required, look at private helper methods to understand what is being checked in more detail). Quote the error messages that would be added in each case exactly. Here is the translation file for validation messages: #{en_yml}. You should create `create_documentation_for_given_validation` method. Call it once, and include the documentation for given validation.", }, { role: "user", content: "Describe #{meth} validation in detail. Here is the content of the validation: #{validation_source} Look at these helper methods where needed to understand what is being checked in more detail: #{helper_methods_source}", }, ], tools: [ { type: "function", function: { name: "create_documentation_for_given_validation", description: "Use this function to save the complete documentation, covering given validation in the provided code.", parameters: { type: :object, properties: { description: { type: :string, description: "A human-readbale description of the validation", }, cases: { type: :array, description: "A list of cases that this validation triggers on, each with specific details", items: { type: :object, description: "Information about a single case that triggers this validation", properties: { case_description: { type: :string, description: "A human-readable description of the case in which this validation triggers", }, errors: { type: :array, description: "The error messages that would be added if this case triggers the validation", items: { type: :object, description: "Information about a single error message for a specific field", properties: { error_message: { type: :string, description: "A single error message", }, field: { type: :string, description: "The field that the error message would be added to.", }, }, required: %w[error_message field], }, }, from: { type: :number, description: "the year from which the validation starts. If this validation runs for logs with a startdate after a certain year, specify that year here, only if it is not specified in the validation method, leave this field blank", }, to: { type: :number, description: "the year in which the validation ends. If this validation runs for logs with a startdate before a certain year, specify that year here, only if it is not specified in the validation method, leave this field blank", }, validation_type: { type: :string, enum: %w[presence format minimum maximum range inclusion length other], description: "The type of validation that is being performed. This should be one of the following: presence (validates that the question is answered), format (validates that the answer format is valid), minimum (validates that entered value is more than minimum allowed value), maximum (validates that entered value is less than maximum allowed value), range (values must be between two values), inclusion (validates that the values that are not allowed arent selected), length (validates the length of the answer), other", }, other_validated_models: { type: :string, description: "Comma separated list of any other models (other than log) that were used in this validation. These are possible models (only add a value to this field if other validation models are one of these models): User, Organisation, Scheme, Location, Organisation_relationship, LaRentRange. Only leave this blank if no other models were used in this validation.", }, }, required: %w[case_description errors validation_type other_validated_models], }, }, }, required: %w[description cases], }, }, }, ], tool_choice: { type: "function", function: { name: "create_documentation_for_given_validation" } }, }, ) rescue StandardError => e Rails.logger.error("Failed to describe #{meth}") Rails.logger.error("Error #{e.message}") sleep(5) next end begin result = JSON.parse(response.dig("choices", 0, "message", "tool_calls", 0, "function", "arguments")) result["cases"].each do |case_info| case_info["errors"].each do |error| Validation.create!(log_type: "lettings", validation_name: meth.to_s, description: result["description"], field: error["field"], error_message: error["error_message"], case: case_info["case_description"], section: form.get_question(error["field"], nil)&.subsection&.id, from: case_info["from"] || "", to: case_info["to"] || "", validation_type: case_info["validation_type"], hard_soft: "hard", other_validated_models: case_info["other_validated_models"]) end end Rails.logger.info("******** described #{meth} ********") rescue StandardError => e Rails.logger.error("Failed to save #{meth}") Rails.logger.error("Error #{e.message}") end end end desc "Generate documentation for soft lettings validations" task describe_soft_lettings_validations: :environment do include Validations::SoftValidations client = OpenAI::Client.new(access_token: ENV["OPENAI_API_KEY"]) current_form = FormHandler.instance.forms["current_lettings"] previous_form = FormHandler.instance.forms["previous_lettings"] all_helper_methods = Validations::SoftValidations.private_instance_methods # describe all soft validations validation_descriptions = {} Validations::SoftValidations.instance_methods.each do |meth| validation_source = method(meth).source helper_methods_source = all_helper_methods.map { |helper_method| if validation_source.include?(helper_method.to_s) method(helper_method).source end }.compact.join("\n") begin response = client.chat( parameters: { model: "gpt-3.5-turbo", messages: [ { role: "system", content: "You write amazing documentation, as a senior technical writer. Your audience is non-technical team members. You have been asked to document the validations in a Rails application. The application collects social housing data for different collection years. There are validations on different fields, sometimes the validations depend on several fields. You are given a method that contains a validation. Describe what given method does, be very explicit about all the different validation cases (be specific about the years for which these validations would run, which values would be invalid or which values are required, look at private helper methods to understand what is being checked in more detail). You should create `create_documentation_for_given_validation` method. Call it once, and include the documentation for given validation.", }, { role: "user", content: "Describe #{meth} validation in detail. Here is the content of the validation: #{validation_source} Look at these helper methods where needed to understand what is being checked in more detail: #{helper_methods_source}", }, ], tools: [ { type: "function", function: { name: "create_documentation_for_given_validation", description: "Use this function to save the complete documentation, covering given validation in the provided code.", parameters: { type: :object, properties: { description: { type: :string, description: "A human-readbale description of the validation", }, validation_type: { type: :string, enum: %w[presence format minimum maximum range inclusion length other], description: "The type of validation that is being performed. This should be one of the following: presence (validates that the question is answered), format (validates that the answer format is valid), minimum (validates that entered value is more than minimum allowed value), maximum (validates that entered value is less than maximum allowed value), range (values must be between two values), inclusion (validates that the values that are not allowed arent selected), length (validates the length of the answer), other", }, other_validated_models: { type: :string, description: "Comma separated list of any other models (other than log) that were used in this validation. These are possible models (only add a value to this field if other validation models are one of these models): User, Organisation, Scheme, Location, Organisation_relationship, LaRentRange. Only leave this blank if no other models were used in this validation.", }, }, required: %w[description validation_type other_validated_models], }, }, }, ], tool_choice: { type: "function", function: { name: "create_documentation_for_given_validation" } }, }, ) rescue StandardError => e Rails.logger.info("Failed to describe #{meth}") Rails.logger.info("Error #{e.message}") sleep(5) next end result = JSON.parse(response.dig("choices", 0, "message", "tool_calls", 0, "function", "arguments")) validation_descriptions[meth.to_s] = result end # add a validation for each interruption screen page for both forms [current_form, previous_form].each do |form| interruption_screen_pages = form.pages.select { |page| page.questions.first.type == "interruption_screen" } interruption_screen_pages_grouped_by_question = interruption_screen_pages.group_by { |page| page.questions.first.id } interruption_screen_pages_grouped_by_question.each do |_question_id, pages| pages.map do |page| subsection_pages = form.subsection_for_page(page).pages page_index = subsection_pages.index(page) page_the_validation_applied_to = subsection_pages[page_index - 1] loop do break unless page_the_validation_applied_to.questions.first.type == "interruption_screen" page_index -= 1 page_the_validation_applied_to = subsection_pages[page_index - 1] end validation_depends_on_hash = page.depends_on.each_with_object({}) do |depends_on, result| depends_on.each do |key, value| if validation_descriptions.include?(key) result[key] = value end end end if validation_depends_on_hash.empty? Rails.logger.info("No validation description found for #{page.questions.first.id}") next end if Validation.where(validation_name: validation_depends_on_hash.keys.first, field: page_the_validation_applied_to.questions.first.id, from: form.start_date).exists? Rails.logger.info("Validation #{validation_depends_on_hash.keys.first} already exists for #{page_the_validation_applied_to.questions.first.id} for start year #{form.start_date.year}") next end result = validation_descriptions[validation_depends_on_hash.keys.first] informative_text = page.informative_text if informative_text.present? && !(informative_text.is_a? String) informative_text = I18n.t(page.informative_text["translation"]) end title_text = page.title_text if title_text.present? && !(title_text.is_a? String) title_text = I18n.t(page.title_text["translation"]) end error_message = [title_text, informative_text, page.questions.first.hint_text].compact.join("\n") case_info = page.depends_on.first.values.first ? "Provided values fulfill the description" : "Provided values do not fulfill the description" Validation.create!(log_type: "lettings", validation_name: validation_depends_on_hash.keys.first.to_s, description: result["description"], field: page_the_validation_applied_to.questions.first.id, error_message:, case: case_info, section: form.get_question(page_the_validation_applied_to.questions.first.id, nil)&.subsection&.id, from: form.start_date, to: form.start_date + 1.year, validation_type: result["validation_type"], hard_soft: "soft", other_validated_models: result["other_validated_models"]) Rails.logger.info("******** described #{validation_depends_on_hash.keys.first} for #{page_the_validation_applied_to.questions.first.id} ********") end end end end desc "Generate documentation for hard bu lettings validations" task describe_bu_lettings_validations: :environment do client = OpenAI::Client.new(access_token: ENV["OPENAI_API_KEY"]) en_yml = File.read("./config/locales/en.yml") [[FormHandler.instance.forms[FormHandler.instance.form_name_from_start_year(2023, "lettings")], BulkUpload::Lettings::Year2023::RowParser], [FormHandler.instance.forms[FormHandler.instance.form_name_from_start_year(2024, "lettings")], BulkUpload::Lettings::Year2024::RowParser]].each do |form, row_parser_class| validation_methods = row_parser_class.private_instance_methods.select { |method| method.starts_with?("validate_") } all_helper_methods = row_parser_class.private_instance_methods(false) + row_parser_class.instance_methods(false) - validation_methods field_mapping_for_errors = row_parser_class.new.send("field_mapping_for_errors") validation_methods.each do |meth| if Validation.where(validation_name: meth.to_s, bulk_upload_specific: true, from: form.start_date).exists? Rails.logger.info("Validation #{meth} already exists for #{form.start_date.year}") next end validation_source = row_parser_class.instance_method(meth).source helper_methods_source = all_helper_methods.map { |helper_method| if validation_source.include?(helper_method.to_s) row_parser_class.instance_method(helper_method).source end }.compact.join("\n") begin response = client.chat( parameters: { model: "gpt-3.5-turbo", messages: [ { role: "system", content: "You write amazing documentation, as a senior technical writer. Your audience is non-technical team members. You have been asked to document the validations in a Rails application. The application collects social housing data for different collection years. There are validations on different fields, sometimes the validations depend on several fields. Describe what given validation does, be very explicit about all the different validation cases (be specific about the years for which these validations would run, which values would be invalid or which values are required, look at private helper methods to understand what is being checked in more detail). Quote the error messages that would be added in each case exactly. Here is the translation file for validation messages: #{en_yml}. You should create `create_documentation_for_given_validation` method. Call it once, and include the documentation for given validation.", }, { role: "user", content: "Describe #{meth} validation in detail. Here is the content of the validation: #{validation_source} Look at these helper methods where needed to understand what is being checked in more detail: #{helper_methods_source}", }, ], tools: [ { type: "function", function: { name: "create_documentation_for_given_validation", description: "Use this function to save the complete documentation, covering given validation in the provided code.", parameters: { type: :object, properties: { description: { type: :string, description: "A human-readbale description of the validation", }, cases: { type: :array, description: "A list of cases that this validation triggers on, each with specific details", items: { type: :object, description: "Information about a single case that triggers this validation", properties: { case_description: { type: :string, description: "A human-readable description of the case in which this validation triggers", }, errors: { type: :array, description: "The error messages that would be added if this case triggers the validation", items: { type: :object, description: "Information about a single error message for a specific field", properties: { error_message: { type: :string, description: "A single error message", }, field: { type: :string, description: "The field that the error message would be added to.", }, }, required: %w[error_message field], }, }, validation_type: { type: :string, enum: %w[presence format minimum maximum range inclusion length other], description: "The type of validation that is being performed. This should be one of the following: presence (validates that the question is answered), format (validates that the answer format is valid), minimum (validates that entered value is more than minimum allowed value), maximum (validates that entered value is less than maximum allowed value), range (values must be between two values), inclusion (validates that the values that are not allowed arent selected), length (validates the length of the answer), other", }, other_validated_models: { type: :string, description: "Comma separated list of any other models (other than log) that were used in this validation. These are possible models (only add a value to this field if other validation models are one of these models): User, Organisation, Scheme, Location, Organisation_relationship, LaRentRange. Only leave this blank if no other models were used in this validation.", }, }, required: %w[case_description errors validation_type other_validated_models], }, }, }, required: %w[description cases], }, }, }, ], tool_choice: { type: "function", function: { name: "create_documentation_for_given_validation" } }, }, ) rescue StandardError => e Rails.logger.error("Failed to describe #{meth}") Rails.logger.error("Error #{e.message}") sleep(5) next end begin result = JSON.parse(response.dig("choices", 0, "message", "tool_calls", 0, "function", "arguments")) result["cases"].each do |case_info| case_info["errors"].each do |error| error_fields = field_mapping_for_errors.select { |_key, values| values.include?(error["field"].to_sym) }.keys error_fields = [error["field"]] if error_fields.empty? error_fields.each do |error_field| Validation.create!(log_type: "lettings", validation_name: meth.to_s, description: result["description"], field: error_field, error_message: error["error_message"], case: case_info["case_description"], section: form.get_question(error_field, nil)&.subsection&.id, from: form.start_date, to: form.start_date + 1.year, validation_type: case_info["validation_type"], hard_soft: "hard", other_validated_models: case_info["other_validated_models"], bulk_upload_specific: true) end end end Rails.logger.info("******** described #{meth} ********") rescue StandardError => e Rails.logger.error("Failed to save #{meth}") Rails.logger.error("Error #{e.message}") end end end end end