Anki-Connect Spanish cards

It is a web-service allowing to generate language learning cards (Spanish to Russian, can be changed pretty easily) automatically:

1. A multi-choice QA card: single russian word, multiple spanish choices
2. A basic card with an audio on the front and spanish word or sentence on the back
3. A basic card and its reversed version with an image and spanish word on the front, and russian work on the back

Which gives in total 4 anki cards generated.
It calls Anki via Anki Connect to create new cards and to sync new card to AnkiWeb,
so after generating cards you can use it via any Anki client.

All audio files are stored in a personal Yandex Disk, and served from there, giving the opportunity to create lots of quality flashcards.

Demo

TODO

• Ask for a deck to put generated cards to
• Use external LLM to generate questions for QA multi-choice card
• Use https://github.com/pndurette/gTTS library and provide option to generate diffeent voices locally or use Google Translate TTS feature
• Store media files in Anki, not on Yandex Disk
• Add another card type: cloze deletion
• Add support for DeepSeek
• (fix) improved startup time by avoiding unnecessary dependencies
• Add support for other source and target languages
• Add image resize option to optimize storage
• Add option to avoid saving sound recordings from gTTS on YandexDisk
• Document obtaining tokens for different services required to run the system
• Document environment variables
• Avoid duplicates (use storage service)

System requirements

1. AnkiConnect plugin installed
2. Docker and compose
3. No GPU required for docker-compose.gigachat.gtts.yaml

Running

1. Run Anki with Anki Connect add-on enabled, make sure it uses port 8765
2. Run the rest of the system with Docker Compose:

# giga-chat for card generation (the fastest option)
docker compose -f docker-compose.gigachat.gtts.yaml up

or

# deepseek for card generation
docker compose -f docker-compose.deepseek.gtts.yaml up

or

# everything local
docker compose -f docker-compose.local_llm.yaml up

3. Wait for all service to become healthy (first run might take long if you choose to use locally hosted LLM and TTS model)
4. Open https://localhost:8443/
5. Default user is admin, default password is admin

Changing password

Edit .env file changing variables NGINX_USER_NAME, and NGINX_PASSWORD.

Running Anki as a background service

I use Ubuntu, and so can offer the following solution for Ubuntu users (might work for users of other Linux distributions:

1. Install X Virtual Framebuffer:

sudo apt-get install -y xvfb

2. Register it as a Systemd service:

• use a template
• register a new service:

  sudo cp anki-autoconnect.service /etc/systemd/system/
  # register a service
  sudo systemctl enable anki-autoconnect.service
  # start the service
  sudo systemctl start anki-autoconnect.service
  # check whether it is started or not
  sudo systemctl status anki-autoconnect.service

Running AnkiLangCards as a service

1. Register it as a Systemd service:

• use a template
• register a new service:

  sudo cp anki-lang-cards.gigachat.gtts.service /etc/systemd/system/anki-lang-cards.service
#  or
#  sudo cp docker-compose.deepseek.gtts.yaml /etc/systemd/system/anki-lang-cards.service
#  or
#  sudo cp anki-lang-cards.local_llm.gtts.service /etc/systemd/system/anki-lang-cards.service
#  or
#  sudo cp anki-lang-cards.gigachat.service /etc/systemd/system/anki-lang-cards.service
#  or
#  sudo cp anki-lang-cards.local_llm.service /etc/systemd/system/anki-lang-cards.service
  # register a service
  sudo systemctl enable anki-lang-cards.service
  # start the service
  sudo systemctl start anki-lang-cards.service
  # check whether it is started or not
  sudo systemctl status anki-lang-cards.service

Component diagram

%%{init: {'theme': 'base', 'themeVariables': { 'primaryBorderColor': '#ff9999', 'primaryBackgroundColor': '#ffeeee'}}}%%

flowchart TD
    subgraph Reverse_Proxy["Reverse Proxy"]
        nginx(Nginx)
    end

    subgraph Application_Services["Application Services"]
        storage(Storage Service)
        translate(Translate Service)
        yandex(Yandex Disk Service)
        anki(Anki Card Generator)
        image(Image Search Service)
        audio_orchestrator(Audio Service Orchestrator)
    end

    subgraph Machine_Learning_Services["Machine Learning Services"]
        audio(Audio Service)
        llm(LLM Service)
    end

    subgraph Data_Persistence["Data Persistence"]
        postgres(PostgreSQL)
    end

    Browser --> nginx

    nginx --> storage
    nginx --> translate
    nginx --> yandex
    nginx --> anki
    nginx --> image
    nginx --> audio_orchestrator
    nginx --> llm

    storage --> postgres

    audio_orchestrator --> audio
    audio_orchestrator --> yandex