Skip to content

Overview

IP Fabric is an API first client, that means that any click in the user interface (UI) has an API call associated with the action of the click.

Python IP Fabric (SDK)

API first principles allow the IP Fabric team to maintain a Software Development Kit (SDK) that can used for automations. While the IP Fabric main GUI is written in TypeScript, we have chosen to write the SDK in Python as it is the most common language used by Network Engineers.

The SDK documentation can be found at https://community-fabric.github.io/python-ipfabric/

While using SDK is more straightforward then calling API directly, we recommend making yourself familiar with the rest of the documentation as many SDK calls map directly to API calls.

Using API Directly

The following documentation is meant to get you started working with API and give you a tutorial on how it works.

Many of the tables within IP Fabric have a ? which provides a description of how to interact with the API, the data required in the payload, and the endpoint to send the request to. This is the easiest way to learn how to use the API and for more information please see Table Description.

Another option is using the network viewer in the web browser’s Developers Tools which will show you the endpoints and data required to make a successful call to the platform.

API Blog Posts

Please take a look at the following blog posts about using the IP Fabric API and Python SDK:

API Versioning

We have adopted a rather fast-moving API versioning schema starting with release 5.0. This allows for tighter control over the changes and provides opportunity for maintaining a backwards compatibility for a manageable time frame.

Design still leverages API version directly in URL over alternative approaches, such as extended Accept header. This allows for unambiguous exchange of state between IPF and customers, nothing more then URL is needed to identify the resource being accessed.

URL Schema a Version in URL

URL schema looks like

https://{hostname}/api/v{major}.{minor}/{resource_path}

we use the following shortened version throughout the documentation:

/api/{api_version}/{resource_path}

where

  • major and minor follows the release version of the platform. So, if the 4.4.3 version of the IP Fabric platform is deployed, its latest API path would be /api/v4.4/.
  • The whole .{minor} is optional and it is treated as 0 if omitted.
  • {api_version} stands for the complete version string, for example v5.1. Please, mind the v prefix.

API version deprecation and allowed changes

IPF commits to keep support for all minor versions within the particular major version (so for example in release 5.4.3 we allow /api/v5.1/ calls). But API may break between major releases.

We call major release a release which introduces some complex functionality. This typically means a large code refactoring, changes of data model etc. took place. Keeping the backwards compatibility is frequently not feasible in such cases (for example v3 brought support for snapshots, v4 brought graphs, v5 RBAC) .

Naturally backwards compatible changes (typically addition of a new attribute) don’t need to be gated and can appear in responses even when older API version is used within the request.

Changes and depreciations are communicated via API documentation and Release Notes. We will mark attributes as deprecated: true in the OpenAPI schema, when we fully migrate to OpenAPI 3.0+ specification.

Responses made with minor < release_minor will contain a custom header warning client about use of potentially old API. We leverage the IETF Draft for Deprecation header . It is included, and set to true (Deprecation: true) in every response to call made with older then current API version.

Requests made with major < release_major (obsolete version) will be refused with HTTP code 410 Gone (also 406 Not Acceptable would be reasonable, but we explicitly mention this in the apidoc in relationship with Accept header, so not to confuse these two the 410 was selected).

Requests made with API version newer than the currently supported version are also refused with 410 Gone (while 404 or 406 would be probably closer semantically, we don’t want to confuse clients, and keep 410 for all version related errors).

Error body, in both cases, is a JSON depicting current platform version as well as the current API version:

{
  "message": "Unsupported API version used.",
  "release_version": "5.4.2+1",
  "api_version": "v5.4"
}

This may look excessive at first, but it is to promote the client code to be changed with every update, as to keep track with all small changes and updates we bring. But in reality, clients would be typically fine, if they update API version when moving to the new major release only. Frequently, the client code is actually not going to change except of the API version. Thus we recommend using a global constant for it.

Technology Table Endpoints

The technology tables use POST requests only for reading information and the payload is used to specify or filter requested data from listed tables. The POST & DELETE request can be used for Intent verification rules at each endpoint. At every technology table, search for the question mark button that exposes the endpoints (can be used with filters as well).

API Endpoint inline description

Payload definition

{
  columns:[],
  filters:{},
  pagination:{},
  snapshot:"snapshotID",
  reports:""
}
  • columns – specifies columns that we request for the endpoint
  • filters – filtering options, for any column or intent verification {"vendor":["like","cisco"],"family":["eq","lap"],"reload":["color","eq","0"]}( optional)
  • pagination - specifies the pagination and response limits {"limit":26,"start":0} (optional)
  • snapshot – defines snapshot ID or we can use: $last, $prev , $lastlocked
  • reports – Intent rules definition (optional)