A long-lived project that still receives updates
Airtable client to make Airtable interactions a breeze


>= 0
~> 5.0
~> 10.0


 Project Readme


Airrecord is an alternative Airtable Ruby libary to airtable-ruby. Airrecord attempts to enforce a more database-like API to Airtable. However, there's also an ad-hoc API available that skips the class definitions!

You can add this line to your Gemfile to use Airrecord:

gem 'airrecord'

A quick example to give an idea of the API that Airrecord provides:

Airrecord.api_key = "key1"

class Tea < Airrecord::Table
  self.base_key = "app1"
  self.table_name = "Teas"

  has_many :brews, class: "Brew", column: "Brews"

  def self.chinese
    all(filter: '{Country} = "China"')

  def self.cheapest_and_best
    all(sort: { "Rating" => "desc", "Price" => "asc" })

  def location
    [self["Village"], self["Country"], self["Region"]].compact.join(", ")

  def green?
    self["Type"] == "Green"

class Brew < Airrecord::Table
  self.base_key = "app1"
  self.table_name = "Brews"

  belongs_to :tea, class: "Tea", column: "Tea"

    all(filter: "{Temperature} > 90")

  def done_brewing?
    Time.parse(self["Created At"]) + self["Duration"] >

teas = Tea.all
tea = teas.first
tea["Country"] # access atribute
tea.location # instance methods
tea.brews # associated brews

A short-hand API for definitions and more ad-hoc querying is also available:

Tea = Airrecord.table("api_key", "app_key", "Teas")

Tea.all.each do |record|
  puts "#{}: #{record["Name"]}"




Based on Changelog There are two ways to authenticate with Airtable API:

  1. API key
  2. Personal Access Token

API key

To obtain your API client, navigate to the Airtable's API page, select your base and obtain your API key and application token.

You can provide a global API key with:

Airrecord.api_key = "your api key"

The app token has to be set on the definitions of the tables (see API below). You can override the API key per table.

Personal Access Token (PAT)

  • To create a PAT, navigate to the Airtable's Personal Access Tokens page and create a new token.
  • Give your token a unique name. This name will be visible in record revision history.
  • Choose the scopes to grant to your token. This controls what API endpoints the token will be able to use. For more information, see API scopes.
  • Click ‘add a base’ to grant the token access to a base or workspace.
  • Click ‘create token’ to create the token. You will be shown the token’s value. This is the only time you will be able to see the token’s value, so be sure to copy it to a secure location.

You can provide a global PAT with:

Airrecord.api_key = "your PAT"


The Airrecord API is centered around definitions of Airrecord::Table from which the definitions of your tables inherit. This is analogous to ActiveRecord::Base. For example, we may have a Base to track teas we have tried.

Airrecord.api_key = "your api key" # see authentication section

class Tea < Airrecord::Table
  self.base_key = "app1"
  self.table_name = "Teas"

  def location
    [self["Village"], self["Country"], self["Region"]].compact.join(", ")

This gives us a class that maps to records in a table. Class methods are available to fetch records on the table.

Reading a Single Record

Retrieve a single record via #find:

tea = Tea.find("someid")

Listing Records

Retrieval of multiple records is usually done through #all. To get all records in a table:

Tea.all # array of Tea instances

You can use all options supported by the API (they are documented on the API page for your base). By default #all will traverse all pages, see below on how to control pagination.

To use filterbyFormula to filter returned records:

# Retrieve all teas from China
Tea.all(filter: '{Country} = "China"')

# Retrieve all teas created in the past week
Tea.all(filter: "DATETIME_DIFF(CREATED_TIME(), TODAY(), 'days') < 7")

# Retrieve all teas that don't have a country defined
Tea.all(filter: "{Country} = \"\"")

This filtering can, of course, also be done in Ruby directly after calling #all without filter, however, it may be more efficient to let Airtable filter if you have a lot of records.

You can use view to only fetch records from a specific view. This is less ad-hoc than filterByFormula:

# Retrieve all teas in the green tea view
Tea.all(view: "Green")

# Retrieve all Japanese teas
Tea.all(view: "Japan")

The sort option can be used to sort results returned from the Airtable API.

# Sort teas by the Name column in ascending order
Tea.all(sort: { "Name" => "asc" })

# Sort teas by Type (green, black, oolong, ..) in descending order
Tea.all(sort: { "Type" => "desc" })

# Sort teas by price in descending order
Tea.all(sort: { "Price" => "desc" })

Note again that the key must be the full column name.

As mentioned above, by default Airrecord will return results from all pages. This can be slow if you have 1000s of records. You may wish to use the view and/or filter option to sort in the results early, instead of doing 10s of calls. Airrecord will always fetch the maximum possible amount of records (100). This means that fetching 1,000 records will take 10 (at least) roundtrips. You can disable pagination (which fetches the first page) by passing paginate: false. This is especially useful if you're looking to fetch a set of recent records from a view or formula in tandem with a sort:

# Only fetch the first page. Sorting is undefined.
Tea.all(paginate: false)

# Give me only the most recent teas
Tea.all(sort: { "Created At" => "desc" }, paginate: false)

When you know the IDs of the records you want, and you want them in an ad-hoc order, use #find_many instead of #all:

teas = Tea.find_many(["someid", "anotherid", "yetanotherid"])
#=> [<Tea @id="someid">,<Tea @id="anotherid">, <Tea @id="yetanotherid">]


Creating a new record is done through Table.create.

tea = Tea.create("Name" => "Feng Gang", "Type" => "Green", "Country" => "China") # id of the new record
tea["Name"] # "Feng Gang"

If you need to manipulate a record before saving it, you can use instead of create, then call #save when you're ready.

tea ="Type" => "Green", "Country" => "China")
tea["Name"] = "Feng Gang"

Note that column names need to match the exact column names in Airtable, otherwise Airrecord will throw an error that no column matches it.

If you need to include optional request parameters, such as the typecast parameter, these can be passed to either Table.create or #save. This is also supported when updating existing records with the #save method.

tea = Tea.create(
  {"Name" => "Feng Gang", "Type" => "Green", "Country" => "China"},
  {"typecast" => true},

# Or with the #save method:
tea ={"Name" => "Feng Gang", "Type" => "Green"})
tea["Name"] = "Feng Gang""typecast" => true)

Earlier versions of airrecord provided methods for snake-cased column names and symbols, however this proved error-prone without a proper schema API from Airtable which has still not been released.


Updating a record is done by changing the attributes and persistent to Airtable with #save.

tea = Tea.find("someid")
tea["Name"] = "Feng Gang Organic"
tea["Village"] = "Feng Gang" # persist to Airtable

Airtable's API doesn't allow you to change attachment's filename. As a workaround you can delete the original attachment and upload a new one with the original URL and a new filename.


An instantiated record can be deleted through #destroy:

tea = Tea.find("rec839")
tea.destroy # deletes record

File Uploads

Airtable's API requires you to have uploaded your file to an intermediary and providing the URL. Unfortunately, it does not allow uploading directly.

word = World.find("cantankerous")
word["Pronounciation"] = [{url: ""}]

S3 is a good place to upload files for Airtable. Airrecord does not support this directly, but the snippet below may be helpful:

# Add this to your gemfile
# Full docs at
require 'aws-sdk-s3'

  credentials:, secret_key) # obtain from AWS
  region: 'ca-central-1', # region

s3 =
  body:"cantankerous.mp3"), # IO object
  bucket: 'word-pronunciations',
  key: 'cantankerous.mp3',
  acl: "public-read",


Airrecord supports managing associations between tables by linking Airrecord::Table classes. To continue with our tea example, we may have another table in the base to track brews of a specific tea (temperature, steeping time, rating, ..). A tea thus has many brews:

class Tea < Airrecord::Table
  self.base_key = "app1"
  self.table_name = "Teas"

  has_many :brews, class: "Brew", column: "Brews"
  has_one :teapot, class: "Teapot", column: "Teapot"

class Brew < Airrecord::Table
  self.base_key = "app1"
  self.table_name = "Brews"

  belongs_to :tea, class: "Tea", column: "Tea"

class Teapot < Airrecord::Table
  self.base_key = "app1"
  self.table_name = "Teapot"

  belongs_to :tea, class: "Tea", column: "Tea"

This gives us access to a bunch of convenience methods to handle the assocation between the two tables. Note that the two tables need to be in the same base (i.e. have the same base key) otherwise this will not work as Airtable does not support associations across Bases.

Retrieving associated records

To retrieve records from associations to a record:

tea = Tea.find("rec123")

# record.association returns Airrecord instances
tea.brews #=> [<Brew @id="rec456">, <Brew @id="rec789">]
tea.teapot #=> <Teapot @id="rec012">

# record["Associated Column"] returns the raw Airtable field, an array of IDs
tea["Brews"] #=> ["rec789", "rec456"]
tea["Teapot"] #=> ["rec012"]

This in turn works the other way too:

brew = Brew.find("rec456")
brew.tea #=> <Tea @id="rec123"> the associated tea instance
brew["Tea"] #=> the raw Airtable field, a single-item array ["rec123"]

Creating associated records

You can easily associate records with each other:

tea = Tea.find("rec123")
# This will create a brew associated with the specific tea
brew ="Temperature" => "80", "Time" => "4m", "Rating" => "5")
brew.tea = tea

Alternatively, you can specify association ids directly:

tea = Tea.find("rec123")
brew ="Tea" => [], "Temperature" => "80", "Time" => "4m", "Rating" => "5")

Ad-hoc API

Airrtable provides a simple, ad-hoc API that will instantiate an anonymous Airrecord::Table for you on the fly with the configured key, app, and table. This is useful if you require no custom definitions, or you're just playing around.

Tea = Airrecord.table("api_key", "app_key", "Teas")

Tea.all.each do |record|
  puts "#{}: #{record["Name"]}"



Airtable's API enforces a 5 requests per second limit per client. In most cases, you won't reach this limit in a single thread but it's possible for some fast calls. Airrecord automatically enforces this limit. It doesn't use a naive throttling algorithm that does a sleep(0.2) between each call, but rather keeps a sliding window of when each call was made and will sleep at the end of the window if requires. This means that bursting is supported.

Production Middlewares

For production use-cases, it's worth considering adding retries and circuit breakers to Airrecord. This is not enabled by default. Airrecord uses the Faraday gem for HTTP. Similar to Rack, you can add middlewares to provide reusable logic for making HTTP requests.

Configuring Retries

Refer to the documentation for all configuration options.

Airrecord::Table.client.connection.request :retry,
  max: 5, interval: 1, interval_randomness: 2, backoff_factor: 2,
  exceptions: [...] # It's suggested to be explicit here instead of relying on defaults

If you are running background scripts or workers with the sole purpose of communicating with Airtable, it may be worth retrying on failures. Note that this may cause the process to sleep for many seconds, so choose your values carefully.

The Net::HTTP library that Faraday uses behind the scenes by default has opaque exceptions. If you choose to go beyond retrying on timeouts (as is provided by default by the Retry middleware), I suggest referring to a complete list of Net::HTTP exceptions, such as this one.

Circuit Breaker

If you're calling Airtable in an application and want to avoid hanging processes when Airtable is unavailable, we strongly recommend configuring circuit breakers. This is a mechanism that after threshold failures, it'll start failing immediately instead of waiting until the timeout. This can avoid outages where all workers are hung trying to talk to a service that will never return, instead of serving useful fallbacks or requests that don't rely on the service. Failing fast is paramount for building reliable systems.

You can configure a middleware such as faraday_middleware-circuit_breaker:

Airrecord::Table.client.connection.request :circuit_breaker,
  timeout: 20, threshold: 5


Contributions will be happily accepted in the form of Github Pull Requests!

  • Please ensure CI is passing before submitting your pull request for review.
  • Please provide tests that fail without your change.