The Stitch Import API - or IAPI for short - is a REST API that allows you to push arbitrary data into your data warehouse. Once the data enters the Import API, it’ll be processed and sent through Stitch like data from any other integration.
Before you get started, consider where the data that you want to push currently lives: is it already in a database or SaaS integration we support? A few of our SaaS integrations allow you to sync data at the tabular level; all database integrations (except for Mongo) can be synced to the columnar level. In addition, you can define Replication Methods for tables that come from a database integration.
In short, unless you want complete control over data replication, it’s usually easier to use the native integration instead of using the Import API.
To use the Import API, you’ll need:
- A little bit of technical know-how. If you’re comfortable writing and maintaining a small Ruby or PHP script, you’ll be more than qualified.
- A Stitch account. Free and paid accounts both have access to the Import API. If you don’t have an account, you can sign up for a free trial on our website.
Authentication with the Import API is done with a single API access token placed in the
Authorization header of your request and your client ID, which will be used in the request body.
Generating an API Access Token
You can generate an API access token by logging into your Stitch account and adding an Import API integration:
- On the Stitch Dashboard page, click the Add an Integration button.
Click the Import API icon.
Enter a name for the integration. This is the name that will display on the for the integration; it’ll also be used to create the schema in your data warehouse.
For example, the name “Stitch Import API” would create a schema called
stitch_import-apiin the data warehouse. This schema is where all the tables for this integration will be stored.
- Click the Save and Generate Token button.
Your API access token has write acccess to your Stitch account, so keep it secret, keep it safe. If at any time your API access token is lost or compromised, you can revoke it and generate a new one.
Example Request Header
The API access token is used to authenticate to the Import API via bearer auth, as demonstrated in the example below:
Locating Your Client ID
Your client ID is used in the request body in the
client_id field to perform the second authentication step for the Import API.
Log into your Stitch account and look at the URL while you’re on the Stitch Dashboard page to find your client ID. It’s the four digit number between
The Import API uses standard HTTP return codes to indicate the status of a request. Your app should handle each of the following return statuses gracefully:
|200 OK||Request was successful.|
|201 Created||New data was added as a result of the request.|
|202 Accepted||The request was accepted and will be processed later; new data will be added once the request is processed.|
|400 Bad Request||The request includes malformed JSON/Transit, did not provide an array of records or contains more than 20,000 records.|
|401 Unauthorized||The request did not have a valid API access token. You should generate a new token and re-try your request.|
|403 Forbidden||The request had a valid API access token, but is not permitted. You should check that the
|405 Method Not Allowed||The request has an HTTP method that is not supported by the endpoint.
|413 Request Entity Too Large||The size of the request body was over 4MB.|
|415 Unsupported Media Type||The content header of the request did not specify JSON or Transit.|
|422 Unprocessable Entity||Your request was missing a required parameter. Check that the array of records in your request contain all of the required request body fields.|
|504 Gateway Timeout||The Import API could not process the request in time. Try again later.|
|503 Service Unavailable||The Import API is experiencing problems. Try again later.|
Generally speaking, codes in:
2xxrange indicate success
4xxrange indicate a bad request - for example: invalid or omitted credentials, a required field is missing, and so on
5xxrange indicate an error on our end. We recommend checking our status page for reported outages if you receive a code in this range. If nothing has been reported and these errors persist, please reach out to our support team.
The Import API supports three methods:
- Status - The
Statusendpoint can be used to determine if the Import API is operating correctly. This endpoint doesn’t require any authentication.
- Validate - The
Validateendpoint can be used to validate requests but will not persist them to Stitch. We recommend using this endpoint for development and testing.
- Upsert - The
Upsertendpoint is used to push data into your data warehouse.
Status endpoint can be used to determine if the Import API is operating correctly. This endpoint doesn’t require any authentication.
Validate endpoint can be used to validate requests but will not persist them to Stitch. We recommend using this endpoint for development and testing. The behavior of this endpoint mirrors that of the
Upsert endpoint, with two exceptions:
- If the request is valid, a
200 OKresponse will be returned.
- Regardless of whether Stitch is functional, a
503 Service Unavailableresponse will never be returned.
Upsert endpoint is used to push data into your data warehouse. In addition to the required request body fields this endpoint will only accept requests that contain the following properties:
- The data being sent must be valid JSON or Transit AND specified in the request header:
- If a Primary Key is a
string, it must be 256 characters or less.
- An array of data cannot contain more than 10,000 individual data points.
- The request body cannot be larger than 4MB.
Example Request Header
Required Request Body Fields
For the Import API to successfully accept and process your data, your request body must contain all of the request body fields listed in the table below.
|client_id||This field should contain your four digit client ID.||
|table_name||This field should contain the name of the destination table, ex:
|sequence||This is a property that tells the Import API the order in which the data points in the request body should be considered. We'll go into more detail on this field in the Defining the Sequence.|
|action||This field contains the action for the endpoint. For the
|key_names||This field defines the Primary Key and will contain an array of field names that unqiuely identify the row that the record belongs to.|
|data||This field contains the data to be upserted into your data warehouse.|
If the request was successful, the response will have an HTTP status code of
201 Created and an empty body.
When you push data to an arbitrary table name using the Import API, the table will be generated dynamically in your data warehouse. When creating requests, keep the following in mind:
- The Import API doesn’t enforce any limitation on the hierarchy of your tables.
- There aren’t any commands to create or destroy a table in the Import API.
- You should create one table for each type of data point that you’ll push. For example: if you have customer and product data, you should create
- Each data point pushed to a single table should have the same data structure. For example: if a
- You can push more than one table using the same API access token. Think of it this way: one schema for each API access token. All tables pushed using the same API access token will be housed in the same schema in your data warehouse.
Defining the Sequence
Sequence properties communicate the order in which data points should be considered – newer data points can replace older ones, but not vice versa. Every data point pushed to the Import API must have a
A simple solution is just to use the current timestamp, but before doing so, consider the following:
Are the rows being considered frequently updated? Rows that are updated every few milliseconds can result in failure if records with identical key values are pushed simultaneously. This means that records with the same key values cannot be sent during the same clock resolution.
For example: if the resolution is measured in milliseconds, records with identical key values cannot be sent during the same millisecond.
Are the records coming from multiple sources? If records from multiple sources will be sent to the Import API, the time clocks of these sources must be synchronized. This is especially important if different sources are pushing rows to the same table.
Take a look at the following example. The first three requests are all for one record (
id 10), while the fourth is for a different record (
If the requests were received in this order:
- Requests 1 and 2 would continue to Stitch, but not Request 3. This is because Request 3 has a lower sequence value than Request 2.
- Request 4 would continue to Stitch.
Did this article help? If you have questions or feedback, please reach out to us.