Local Installation

Handl is a server-side solution. If you want to test it, use the web demo. We do not use long-term data storage, all incoming files are transferred to the server via a secure protocol, processed in RAM and deleted immediately after the results are returned to the user.

Use the instructions below if you want to deploy a Handl solution in your own closed IT loop.

Hardware Requirements

Configuration scenarios:

1. Minimum - no more than 1 document per minute

• Processor: 6 cores, 4.1 GHz, AVX 2 expansion; benchmark: Intel Core i5-10600KF
• RAM: 32 Gb

• Mass storage: 500 Gb SSD

2. Standard - up to 60 documents per minute

• Graphics Card: Nvidia Tesla T4

• Processor: 8 cores, 3.9 GHz, AVX 2 expansion; Benchmark: Intel Xeon W-2245
• RAM: 64GB

• Disk Drive: 1 Tb NVMe

3. Corporate - up to 600 documents per minute

• Graphics Card

  Nvidia Tesla T4 x2

• Processor: 8 cores, 3.9 GHz, AVX 2 expansion; benchmark: Intel Xeon W-2245 x2
• RAM: 128GB

• Storage: 1 Tb NVMe x2

The above are the hardware requirements for productive operation. Handl also runs on weak configurations, such as a Core i5-8250U 1.6 GHz / 8 Gb RAM / 250 Gb SSD laptop. However, performance on such weak configurations is not guaranteed

Ubuntu

Environment requirements:

1. Ubuntu operating system version 18.04

2. Docker container management system

3. docker-compose

4. Nvidia drivers with the latest version available:

   1. nvidia-docker

   2. Nvidia drivers with the latest version available

   3. CUDA version 11.1 or later

5. Internet access for license conformation:

   • https://license.ml.handl.ai/check/v2

   • Dynamic IP

   • Port: 443

   • Protocol: TCP

   • Request: POST

Configuration files

I. docker-compose.yml

Create a docker-compose.yml file depending on the specification of your server.

Copy the following settings:

version: "3"

services:

  queue:
    image: rabbitmq:3.7-management-alpine
    restart: always
    logging: &short_logging
      driver: "json-file"
      options:
        max-file: "10"
        max-size: "100m"

  redis:
    image: redis:6-alpine
    restart: always
    logging: *short_logging

  mongo:
    image: mongo:3.6-stretch
    restart: always
    logging:
      driver: none

  worker: &service
    image: registry.handl.ai/public/worker:v3.6.12
    restart: always
    env_file: &env .env
    command: ""
    depends_on:
      - queue
      - mongo
      - redis
    logging: *short_logging
    volumes:
      - ./errlogs:/logs:rw

  front:
    image: registry.handl.ai/public/docr-demo-nginx:v3.6.12
    volumes:
      - ./nginx.conf:/etc/nginx/conf.d/nginx.conf
    env_file: *env
    ports:
      - ${API_PORT:-8080}:80
    depends_on:
      - api

  api:
    <<: *service
    image: registry.handl.ai/public/api:v3.6.12

  classifier:
    <<: *service
    image: registry.handl.ai/public/classifier:v3.6.12

  multidocnet:
    <<: *service
    image: registry.handl.ai/public/multidocnet:v3.6.12

  heuristics:
    <<: *service
    image: registry.handl.ai/public/heuristics:v3.6.12

  wordnet:
    <<: *service
    image: registry.handl.ai/public/wordnet:v3.6.12

  ocr:
    <<: *service
    image: registry.handl.ai/public/ocr:v3.6.12

  fieldnet:
    <<: *service
    image: registry.handl.ai/public/fieldnet:v3.6.12

  checkbox_segm:
    <<: *service
    image: registry.handl.ai/public/checkbox-segm:v3.6.12

  table_handler:
    <<: *service
    image: registry.handl.ai/public/table-handler:v3.6.12

  face:
    <<: *service
    image: registry.handl.ai/public/face:v3.6.12

II. env и nginx.conf

Create file .env:

Copy the following settings:

VERSION=v3.6.10

# front
TRY_ENDPOINT=/try

DEFAULT_NORMALIZATION_FIAS=False

# Legacy - не работает, но необходимо для запуска
DADATA_TOKEN=
DADATA_SECRET=

AUTOCODE_REPORT=
AUTOCODE_TOKEN=

Download the nginx.conf file:

725B

nginx.conf

nginx.conf

Put the three files docker-compose.yml, .env, and nginx.conf in the same directory.

Running Handl

Navigate to the directory containing the three configuration files. Type the following command at the command line:

docker-compose up -d --force-recreate

You are all set! You can now use the local version of the web demo by opening the http://localhost:8080 link in your browser. You can also access the server using the API.

Debian

Instructions for using Debian 10 (buster) + Nvidia Tesla T4 (example)

1. Add non-free contrib repositories to /etc/apt/sources.list:

   1. deb http://deb.debian.org/debian buster-backports main non-free contrib
   2. deb-src http://deb.debian.org/debian buster-backports main non-free contrib
2. apt update

3. apt install linux-headers-$(uname -r) make gcc acpid dkms nvidia-tesla-418-driver nvidia-tesla-418-smi nvidia-cuda-toolkit

4. systemctl reboot

After that, follow the instructions in Ubuntu related section.

Windows

You can install Handl on a Windows 10 computer. To do this, you will need to perform the following steps.

1. Install Docker

1. Download the Docker installer from the official website. We recommend using the version from the Stable channel.
2. Run the installer and follow the instructions on the screen. At the end of the installation, you need to reboot your computer.

3. After restarting, check the notification area in the bottom right corner. If the whale icon isn't there, your account probably doesn't have permissions to run Docker. If this is the case, type the word Docker into the search box. Right-click on the Docker Desktop application and select "Run as administrator"

2. Install PowerShell

1. Download the PowerShell installer from the official repository. We recommend using a stable version from the Windows (x64) line.
2. Run the installer and follow the instructions on the screen.

3. Type the word PowerShell into the search box. Right-click on the PowerShell application and select "Run as administrator".

3. Verify that the environment is prepared

1. Run the docker version command in PowerShell.

2. Run the docker run hello-world command.

3. If you don't get any errors in these two steps, then everything is working fine

4. Prepare the .env configuration file

Create a folder on your computer with any name, for example Handl. Create an untitled file with an .env extension. Open that file with a text editor, for example Notepad, and copy there the contents of the window below:

# common
VERSION=v3.5.12

# cuda
CUDA_VISIBLE_DEVICES=
NVIDIA_VISIBLE_DEVICES=all
NVIDIA_DRIVER_CAPABILITIES=all

# license
LICENSE=your_license_key

Replace the "your_key_license" text with your license key. Save the resulting file in the Handl folder.

5. Prepare the docker-compose.yml and nginx.conf configuration files

Download the docker-compose.yml file and nginx.conf:

2KB

docker-compose.yml

docker-compose.yml

725B

nginx.conf

nginx.conf

Put these files in the Handl folder.

6. Give Docker access to the folder with the configuration files

1. Right-click the Docker icon in the bottom right corner of the screen

2. Select Settings from the drop down menu

3. Select the Resources section, File Sharing sub-item.

4. Click on the blue plus icon in the circle and specify the path to the Handl directory, for example C:\Handl

5. Click Apply & Restart button

7. Running Handl

1. Go back to the PowerShell that you started in step 2.

2. Navigate to the Handl folder using the command cd C:\Handl

3. Run the docker-compose up -d --force-recreate command and wait for all containers to finish their start up

You are all set!

You can now use the local version of the web demo by opening the http://localhost:8080 link in your browser.

You can also access the local version of Handl using the API. To do this:

1. Put an image of a passport or other type document in the Handl folder.

2. In PowerShell, run curl -X POST "http://localhost:12345/recognize?doc_type=passport_main&quality=75&dpi=300&auto_pdf_raw_images=true&pdf_raw_images=true&token=ваш_ключ_лицензии" -H "accept: application/json" -H "Content-Type: multipart/form-data" -F "image=@C:\Handl\Pasport_RF.jpg;type=image/jpeg"