README.md

Visit API Contract

Semantic version 0.6.0

The below establishes a contract for a general purpose API supporting "interview"-like workflows where the user is presented with a series of prompts.

GET /api/v1/visits/:id/interaction

Current state of the visit may always be retrieved by GET /api/v1/visits/:id/interaction

Response body will always be a JSON object. Example:

{
    "state_name": "new_user_welcome",
    "title": "Welcome, Stranger!",
    "content": [
        {
            "content_type": "display_only",
            "content_name": "intro_paragraph",
            "content_label": "<p>Please tell us a little about yourself to get started.</p>"
        },
        {
            "content_type": "free_text_input",
            "content_name": "first_name",
            "content_label": "First Name",
            "required": true,
            "max_length": 60
        },
        {
            "content_type": "select_input",
            "content_name": "age_category",
            "content_label": "Age",
            "required": true,
            "options": [
                {
                    "option_label": "I am under age 18 and am completing this with my guardian.",
                    "option_value": "under_18",
                },
                {
                    "option_label": "I am age 18 or older.",
                    "option_value": "over_18",
                }
            ]
        }
    ],
    "actions": {
        "continue": {
            "action_label": "Continue",
        },
        "go_back": {
            "action_label": "Go Back",
        },
        "cancel_visit": {
            "action_label": "Cancel visit"
        }
    }
}

If the action is successful, it must return these keys (all keys are always present unless noted otherwise):

• state_name: String, may be used as a screen identifier such as "welcome"

• title: String, human-readable slide title/heading, usage optional, will be translated into the current locale

• content: Array of objects. Each object if any will have these keys:

  • content_type: String

    • Must be one of ["display_only", "boolean_input", "select_input", "free_text_input"].
    • If the content type is "display_only", the client should display the label to the user. The label may contain HTML.
    • If the content type is "boolean_input", it means the user should provide a true or false answer to the question. This could look like a checkbox, or two radio buttons, or if it stands alone it could be two action buttons.
    • If the content type is "select_input", it means the user is required to select one of the options provided. This could look like a radio button or a select menu.
    • If the content type is "free_text_input", it means the user is required to enter text, and any characters are allowed, possibly up to a max_length. The size of the input UI element is up to the client to decide.
    • If the content type is "numeric_input", it means the user is required enter a number, possibly restricted to be integer_only.
    • If the content type is "date_input", it means the user is required to enter a date. The API will expect yyyy-mm-dd format, though user interfaces should be localized (e.g. mm/dd/yyyy format in U.S.)
    • If more types need to be added, API developers should notify clients, treat it as a breaking change, and not release until all clients are ready to handle the new types.

  • content_name: String

    • Identifies the content item in the current view.
    • Included also for "display_only" content, as it helps with front-end frameworks that require a unique identifier for repeat data.
    • The value must be used as the key for providing key-value pairs in the responses portion of the next request.

  • content_label: String

    • Human-readable label, translated into the current locale.
    • Often plain text but in general may have HTML value. Security sanitization is strongly recommended by both the server and client.
    • May serve as the <legend> for a radio button <fieldset>.

  • required: Boolean

    • Indicates whether a user response to this item is required in order to take the "continue" action.
    • Only present for input content items.
    • Inputs are never required for other actions.

  • select_none: Boolean. Special field that may be present for "boolean_input" content. This signals that the current item, if true, must be the only true item among all other boolean inputs in the current view, and that if any other boolean input is true, this item must not be true. Useful for a "None of the above" checkbox.

  • max_length: Number. Special field that is returned for "free_text_input" content. This signals that the input must be limited to this number of characters.

  • integer_only: Boolean. Special field that is returned for "numeric_input" content. This signals that the input must be an integer, not a floating point / decimal number.

  • options: Array

    • Always present for "select_input"

    • Possibly present for "free_text_input" to allow the client to provide the options as suggestions (TBD, future work)

    • Never present for "display_only" content

    • If present, always an array of one or more objects. Each object will have these keys:

      • option_label: String, human-readable label for this specific option, translated into the current locale
      • option_value: [Varies, String, Boolean, or Numeric]: value representing what must be returned in the next request if this option is selected by the user. Within a content item, no two options will ever have the same value.

• actions: Object, with zero or more of the following keys, depending on current user state:

  • continue: Object.

    • This is the primary action for going forward in the interview.
    • It may be missing for end states where there is no option to continue.
    • For this action, it is implied that all input fields listed in "content" must be included in the request.

  • go_back: Object.

    • This allows the user to go back one step, effectively un-doing their previous action.
    • It may be missing when going back is not possible or not permitted.

  • cancel_visit: Object.

    • This fully ends and permanently cancels the current visit.
    • It may be missing if there is no option to cancel, for example after a data request has been submitted to others.

  • see_other_options: Object.

    • This shows the user their other options available at this moment other than continuing the visit.
    • It may be missing if there are no other options, or if we do not wish to highlight any other options in the current state.

  • If an action is listed, then it means that the action is currently supported and the server must accept it.

  • Each action must have these keys:

    • action_label: String.
      • Human-readable label for the action, translated into the current locale.
      • Note that, for example, the Continue action may not be labeled "Continue" in cases where another description may be more sensible.

POST /api/v1/visits/:id/interaction

• Actions may be taken via POST /api/v1/visits/:id/interaction. When POSTing, request body JSON data is required. Example:

  {
      "action_name": "continue",
      "responses": {
          "first_name": "Magdalena",
          "age_category": "over_18"
      }
  }

  Request JSON must include these keys:

  • action_name: String, required. Must be one of the items listed in the "actions" object from the GET or previous POST response.

  • responses: Object, required. Must include a key-value pair for each of the required inputs listed in the previous content response. Each key must be the name of a content item, and each value must be represent the user's input for that item.

• If the request is successful, the state of the visit should change, and the POST response data should be different than the previous response data but must be identical in structure to the GET response, allowing each response to be processed by the same client code again and again.

Error handling

• 401: You are not authenticated.

• 404: The visit cannot be found or you do not have access to it.

• 422: The action cannot be completed and requires user correction.

• 500: The action cannot be completed due to a user-uncorrectable server error.

• If the response is an error 4xx or 5xx, the response will be an object that contains an array of objects:

  { errors: [{ reason: "reason_identifier", message: "Human readable text error reason" }] }