Here are the changes that will take place.
The "url" input parameter
When a URL is passed inside the "url" parameter, Dandelion API fetches the content of the URL for you and uses an AI algorithm to extract the relevant part of the document to work on. This feature will be dropped from January 10th. If you are using this parameter, there is a very simple workaround: you just need to download the content of the URL on your own and pass it inside the "html" parameter, which behaves exactly as the "url" parameter by extract the main content of the web page.
Moreover, until now the "html" parameter required one unit more per request (like "url"). We decided that this won't be the case anymore: from January 10th, "html" calls won't cost +1 unit per request anymore.
There are many very good, stable and widely used libraries to manage HTTP calls for all languages, here’s a few suggestions:
User-defined spots and Text Classification API
These two features allow users to define their own custom data: respectively, spots and classification models. Custom data are very rarely used, but on the other hand the maintenance costs generated by this kind of data (replication, caching, consistency checks among knowledge base updates, etc…) is significant.
We decided to dismiss this feature from the Dandelion API’s standard plans, they will be only available for custom enterprise projects.
The user-defined spots and the Text Classification API will be dismissed from January 10th.
We are aware that these changes will affect some of you, and we are sorry for any inconvenience. However, dismissing these features will let us improve and focus on the value-bringing parts of Dandelion API, offering a much better service for everyone (including users of the free plan).