https://github.com/man-c/pycoingecko
trending
/search/trending (Get trending search coins (Top-7) on CoinGecko in the last 24 hours)
cg.get_search_trending()

!portfolio <exchange> <key> <secret>
Doing this command will add a new row to the database, the exchange options are KuCoin and Binance.
After this has happened we open a user socket so we can keep track of the user's buys and sells.
We also make calls every hour to check the prices of the assets on their account (if there are any)

• Make a command for adding exchange information
• Format the exchange (always lower) and check if we support this exchange, if not don't add to db
• Save this information to a database (pandas dataframe, which will be pickled everytime something changes)
• Get the account balance if !portfolio has been used and save the (user, symbol, quantity, price) in db
• Start user socket (and publish new trades in dedicated channel)
• Check assets on account and make a new channel dedicated to this user (for showing their current assets)
• Send the user a message if their SL or TP has been hit

Possible using the API, however we need to check how quick the token expires

Then this post will get highlighted in another channel

Follows a new user and restarts streams and resets self.following_ids

Use emoji's in title of tweets to make it clear what the tweet is

If something about options is mentioned drop this is in the channel.
How are we sure if a tweet is about options?

Use https://scanner.tradingview.com/{keyword}/scan
keyword can be ["america", "forex", "crypto", "indonesia", "india", "cfd", "uk", "brazil", "vietnam", "rsa", "ksa", "australia", "russia", "thailand"]

Currently 1 hour less compared to discord timestamp

Some coins can only be bought on certain exchanges, use coingecko api to determine the exchanges and use their custom emoji to represent them

Now only shows the quote, but not the content itself

Since the news also has impact on the financial markets, maybe this channel contains tweets from news sources (such as @deltaone, @FirstSquawk, @EPSGUID) or scans a news site constantly (CNBC).

show original url instead of t.co

• Command /alert <ticker> <above/below> <price>
• Save this ticker in a .db file
• Use CCXT binance's mini ticker socket
• If the user has been alerted remove this row from the db

For instance !alert TSLA will tag the user at every tweet containing the $TSLA ticker
We need to have a db for this, maybe alerts.db (ticker, user, id)

Divide tweets into multiple text channels:
Charts category:

• crypto
• stocks
  Text category, Images, Other
  images has pictures, but no $
  other has text, but no $

Maybe have a command like !following that shows all accounts and if someone uses !follow it also posts in the #following channel

See same issue on Axie_Manager_Bot

https://finance.yahoo.com/quote/ALTS
Shows after hours 0$, because it is not available for after hours

Check other project that relate to this:

• https://fintwit.ai/
• https://twitter.com/chatterquant

Maybe try using this repo: https://github.com/atreadw1492/yahoo_fin

Sometimes the bot thinks a hashtag is also a coin, for instance #nft

Store the tweet text + reaction sentiment in csv to create more data for model

StockTwits is basically same a Twitter but then specifically for tweets about stocks / crypto.

https://github.com/khmurakami/pystocktwits

https://twitter.com/gabro27/status/1173547934132178944?s=20

Both more and less
Eg $ETH from 100k tweets -> 120k tweets = 20% increase

• add tweet timestamp
• add more fields to embed
• add footer
• add header with bot avatar

Now displays default RTs better, but quote tweets are a bit strange

Because multiple image embeds is not a thing

To keep track of meme stocks
Use: https://www.storybench.org/how-to-scrape-reddit-with-python/ for inspiration

Send the top 10 trading ideas from https://www.tradingview.com/ideas/ of today in a dedicated channel.

Use https://pypi.org/project/tradingview-scraper/ to get the ideas, and filter based on time they were posted. Only use the ideas posted in the last 24h, and sort them based on likes.

Change title and maybe add the retweet emoij

Shows that there is a discord user interested in this tweet

To prevent multiple times the same ticker in 1 tweet

This will only post their tweets in there, and their tweets will not get posted in the other channels (to prevent clutter)

This is a project on its own, but could be implemented here to remove useless images from the chart channels

• Create a dataset of all images and charts that get posted (use scraping script)
• Try fitting unsupervised clustering on it
• Try labeling part of the data and applying pre-trained model (maybe fine-tuned)

• Add pretrained model
• Implement logic

Possible models

We need to consider efficiency, since it should be able to run on the raspberry pi and finetuning should be possible on a home PC.

• mobilenet
• EfficientNet

For instance if $TSLA was mentioned, also add (+2,03%) or so

So if the user makes a charts or copies a charts from a tweet and the alert goes off it will get posted in a specific channel

Maybe every 15 minutes or so

If a tweet contains a ticker add the ticker, user, sentiment, category, time of tweet to the database.
For every new tweet send a message containing the list of the tickers mentioned in the last 24h and delete the tickers older than 24h.

Also every x hours compare the top 10 (or more depending on rate limit) tickers with twitter's global sentiment.

First ensure that:

• #185 is working correctly to save time on deciding the ticker category