- 🧳 Launched May 2022
- 🚪 Release v3.0.0
- 📦Code size 95 KB
Add your own learned words to the app to reinforce your knowledge with a daily quiz.
Lessons are formed by priority words. The more often the user uses the bot, the more productive it works.
Checks such as card-word, word-card - with a regular approach, they allow you to very effectively consolidate the learned words, especially when the user chooses his own examples with words.
Track your progress. Train every day. Edit words. Build your code on top of the API.
You can check how the user part of the telegram bot (@teleword_learning_bot) works.
A quick overview of the main functionality
First of all, you enter the bot using the /start command
And /help command to get to the main menu
The user is verified; the data is entered into the table
Now it's time to add new data, this is how it usually looks for me
I open the series with subtitles, a dictionary (for me, it's just a browser) and a Teleword
I watch a series until I stumble upon an unfamiliar word
I write down a remark with this word and look it up in the dictionary
I copy the description and wait for a successful addition
Great, I click continue and continue to enjoy the series only a level smarter than I was 30 seconds ago
Under the hood:
- the data is checked for correctness
- the part of speech of the word is pulled up for more effective testing
- rating is being prepared
You have added more than 15 initial words (I recommend at least 30-40) - now you can start(/lesson) filling words
with tests
This is what we remember! Let's read the example again to strengthen our knowledge.
One of the hardest parts of the backend is hidden behind the lesson formations.
It is after the lesson that all user statistics are collected - his mistakes, successes, attendance, the rating of words changes.
The occupation itself is formed by repeated enumeration with a bit of randomness.
You can see Python code in function
services.lessons.build_lesson()The logic for checking the correctness of the answer is enclosed in two coroutines:
- one of them checks if the lesson is finished, takes the next task from the lesson state in redis and displays it to the user
bot.handlers.lessons.next_task- the other one catches the user's response (a button or a typed letter), checks it and updates the lesson state
bot.handlers.lessons.check_answerAt the end, ratings, daily statistics, points and the streak are saved in one transaction
services.lessons.finish_lesson()
Who doesn't like to look at statistics? Make a smart face and click /statistic
We climb into the database, count the number of words
services.vocabulary.count_words(), and the user table gives the total number of points and the streak - days in a row with at least one lessonservices.streak.current_streak().We transfer data by day from the statistics table to plotting through matplotlib, in a worker thread, so the bot keeps answering other users meanwhile
services.stats.recent_stats()->bot.charts.render_stats_chart()->bot.handlers.statistic.statistic_cmd()
We do not hide anything from users, if they need words with the right data type - hold on command /data
If you are a regular user - my advice to you - enjoy a beautiful Excel file 😎
The logic is simple, choose what and with what you want to do and do it
By revalidating the input, checking permissions, and running a cascading deletion of words when trying to delete an example.
Or maybe we want to develop something of our own on the Teleword api? Then let's kindly distract our good old friend admin from swimming in the Mediterranean Sea and drinking cocktails
As soon as the admin approves the request, the key arrives right in the chat, together with detailed instructions for use
All methods of the restful interface are implemented with FastAPI in
teleword/api/app.py, interactive documentation is at/docs. The key is sent in theX-API-Keyheader and is stored only as a hash.The administrator still looks at every request personally, but now it is one tap: the key is generated and sent to the user automatically.
Let's make sure the linux machine is up-to-date.
sudo apt update && sudo apt upgrade -y
Installing useful packages
sudo apt install git openssh
Let's generate ssh keys using the s4lv ed25519 method
ssh-keygen -t ed25519
Setting up access to ssh keys
cd /root/.ssh/
touch config
nano config
Host <SSH NAME>
HostName github.com
IdentityFile ~/.ssh/<KEY>
IdentitiesOnly yes
Copy the public key and transfer it to GitHub
cat <KEY>.pub
cd #
git clone git@<SSH NAME>:vinter-man/teleword.git
yes
Create the config from the example and fill in the bot token and your Telegram id
cd teleword
cp .env.example .env
nano .env
Do not forget to change the config file for yourself.
Install Docker with the compose plugin
docker compose build --quiet
docker compose up -d
docker compose ps
The stack is self-contained, so it can live on a server next to other projects: its own network and volumes, read-only containers with memory and CPU limits, rotated logs. The bot needs no open ports; the API listens on 127.0.0.1:8790 only.
View logs in real time
docker compose logs -f bot
Install uv, it brings the right Python version itself
uv sync
uv run python -m teleword.db
uv run python -m teleword.bot
uv run uvicorn --factory teleword.api.app:create_app
Without REDIS_URL in .env the bot keeps conversation state in memory. Run the tests with uv run pytest
- Telegram - https://t.me/vindemure
- Mail - [email protected]
