Java Library: Overview

This guide will help you get the most of the Mindee Java client library to easily extract data from your documents.



The library is written and compiled with Java 8. This guide assumes you have a compatible JVM language installed



Installation using Apache Maven is recommended.

The library is tested on Java versions 8 and 11.

Other installation methods and/or Java versions may work, but are not officially supported.


The easiest way to use the Mindee client library for your project is by adding
the maven dependency in your project's POM:


For the latest version of the Library please refer to our maven central repository

Development Installation

If you'll be modifying the source code, you'll need to follow these steps to get started.

  1. First clone the repo.
git clone [email protected]:mindee/mindee-api-java.git
  1. Navigate to the cloned directory and install all required libraries.
mvn clean install 

Updating the Library

When starting out it is recommended that you use the latest release version of the library from
our maven central repository.

Future updates can be made using the maven versions plugin.

mvn versions:use-next-releases -Dincludes=com.mindee.sdk:mindee-api-java


Using Mindee's APIs can be broken down into the following steps:

  1. Initialize a Client
  2. Load a file
  3. Send the file to Mindee's API
  4. Retrieve the response
  5. Process the response in some way

Let's take a deep dive into how this works.

Initializing the Client

The com.mindee.MindeeClient class centralizes document configurations into a single class.

The MindeeClient requires your API key.

You can either pass these directly to the constructor or through environment variables.

Pass the API key directly

import com.mindee;

// Init a new client and passing the key directly
MindeeClient client = new MindeeClient("<your mindee api key>");

Set the API key in the environment

API keys can be set as environment variables.

The following environment variable will set the global API key:


Then in your code:

// Init a new client without an API key
MindeeClient client = new MindeeClient();

HttpClient Customizations

Mindee's API lives on the internet and many internal applications on corporate networks may therefore need to configure an HTTP proxy to access it.
This is possible by using a MindeeClient configured to use a user provided instance of the com.mindee.http.MindeeApi interface.

There are a few layers to this:

  • The default implementation of com.mindee.http.MindeeApi interface is com.mindee.http.MindeeHttpApi
  • MindeeHttpApi can be initialized with an Apache HttpComponents HttpClientBuilder.
  • HttpClientBuilder can be configured for use cases like proxying requests, custom authentication schemes, setting SSL Context etc.

To Configure a MindeeClient to use a proxy, the following code can be referenced.

import com.mindee.MindeeClient;
import com.mindee.input.LocalInputSource;
import com.mindee.parsing.common.PredictResponse;
import com.mindee.product.invoice.InvoiceV4;

public class SimpleMindeeClient {
    public static void main(String[] args) throws IOException {

      // You can also configure things like caching, custom HTTPS certs,
      // timeouts and connection pool sizes here.
      // See:
      String proxyHost = "myproxy.local";
      int proxyPort = 8181;
      HttpHost proxy = new HttpHost(proxyHost, proxyPort);

      DefaultProxyRoutePlanner routePlanner = new DefaultProxyRoutePlanner(proxy);
      HttpClientBuilder httpclientBuilder = HttpClients.custom().setRoutePlanner(routePlanner);

      // Build MindeeHttpAPI using the HtppClientBuilder
      MindeeHttpApi mindeeHttpApi = MindeeHttpApi.builder()
          .mindeeSettings(new MindeeSettings("my-api-key"))

      // Inject the MindeeHttpAPI instance directly into the MindeeClient
      MindeeClient mindeeClient = new MindeeClient(mindeeHttpApi);
      // Parse a file as usual
      PredictResponse<InvoiceV4> invoiceDocument = mindeeClient.parse(
          LocalInputSource(new File("/path/to/the/file.ext"))

Loading a Source File

Before being able to send a file to the API, it must first be loaded.

You don't need to worry about different MIME types, the library will take care of handling
all supported types automatically.

Once a file is loaded, interacting with it is done in exactly the same way, regardless
of how it was loaded.

Loading a file allows performing PDF operations on it.

There are a few different ways of loading a document file, depending on your use case:

You can also send distant files.
However, in this case nothing is done or can be done locally.

File Object

Load a object.
When using this option you do not need to pass in a file name - the API uses the file.getName() method to get the file name.

LocalInputSource localInputSource = new LocalInputSource(
    new File("path/to/document/document.pdf")


Load file contents from a base64-encoded string.

Note: The filename of the encoded file is required when calling the method.

LocalInputSource localInputSource = new LocalInputSource(b64String, "document.pdf");


Load file contents from a byte array.

Note: The original filename of the encoded file is required when calling the method.

// Get Byte Array from a File, Multipart File, Input Stream, or as a method parameter
byte[] fileAsBytes = ....;
LocalInputSource localInputSource = new LocalInputSource(fileAsBytes, "document.pdf");
PredictResponse<InvoiceV4> response = mindeeClient.parse(InvoiceV4.class, LocalInputSource);


Alternatively, an HTTPS URL can be loaded:

URL documentUrl = new URL("https://path/to/document");
PredictResponse<InvoiceV4> response = mindeeClient.parse(InvoiceV4.class, documentUrl);

Parsing a Document

The MindeeClient has multiple overloaded parse methods available for parsing the documents
and you will get LocalInputSource.

This can be done by implicitly by calling the parse(Class<T> type) method with the expected response type from
the parse method (InvoiceResponse, ReceiptResponse, PassportResponse, or even you custom class).

Each document type available in the library has its corresponding Response class.
This is detailed in each document-specific guide.

Off-the-Shelf Documents

Simply setting the correct class is enough:

// After the document has been loaded
PredictResponse<ReceiptV4> receiptV4Inference = mindeeClient.parse(ReceiptV4.class, localInputSource);

For more fine-grained control over parsing the documents you can have a look on the parse override method.

Custom Documents

In this case, you will have two ways to handle them.

The first one enables the possibility to use a class object which represents a kind of dictionary where,
keys will be the name of each field define in your Custom API model (on the Mindee platform).

It also requires that you instantiate a new Endpoint object to define the information of your custom API built.

CustomEndpoint endpoint = new CustomEndpoint(
    "1.0" // optional

PredictResponse<CustomV1> customDocument = mindeeClient.parse(localInputSource, endpoint);

The second one is using your own class.
See more information here

Processing the Response

Regardless of the model, it will be encapsulated in a Document object and therefore will have the following attributes:


Regroup the prediction on all the pages of the document and the prediction for all the document.

Document Level Prediction

The Response object for each document type has an attribute that represents data extracted from the entire document.
It's possible to have the same field in various pages, but at the document level only the highest confidence field data
will be shown (this is all done automatically at the API level).

PredictResponse<InvoiceV4> response = documentClient.parse(

// print the complete response

// print all prediction data

// print the document-level prediction

// print the page-level predictions
  page -> System.out.println(page.toString())

Each inference specific class will have its own specific attributes, these correspond to the various fields extracted from the document.

Page Level Prediction

The pagesPrediction attribute is a list of prediction objects, often of the same class as the documentPrediction attribute.
But sometimes, it could have a different class which would extend the documentPrediction class.

Each page element contains the data extracted for a particular page of the document.


Join our Slack