teknologi-umum/botnet

Telegram Bot written in .NET

botnet's Issues

Bug: Image generator does not understand sentences and languages other than English

We need to rewrite and translate the prompt to english before passing it to SDXL

`/map` command should decide zoom level from place's bounding box

Roslyn eval

Add disclaimer watermark to generated images

To reduce the number of "In case of an investigation by any federal entity or similar, I do not have any involvement with this group or with the people in it, I do not know how I am here, probably added by a third party, I do not support any actions by members of this group." quote required in conversation.

Watermark template:

Image generated by AI, prompt by @username

Image Variation

Implement image variation

Add how to build and run use VSCode in README.md

Visual studio code had a lot number of users, and I as a user who develops .NET application using VSCode at daily basis it will be very helpful if there is a guidelines on how to run these applications with VSCode.

VSCode

Javascript eval

`/read`: reply to image, OCR it with IronOcr

Listen to #fyp tag in teknum's mastodon

https://streaming.mastodon.teknologiumum.com/api/v1/streaming/hashtag?tag=:fyp

Street View bot

Search place then capture streetview image

Grain: TechEmpower benchmark scraper

You type /benchmark C# C++

You get "C++ is 30% faster than C#"

Math plot

You type /plot x + y = 1

You get plotted graph

Random quote

You type /wisdom

You get a random quote rendered as image

Generating image from thread

When we ask AI to generate image based on what's discussed in the conversation so far, the bot should summarize the thread into an image generation prompt and pass it to image generator

Change rate limit of `/map` to 1 query per chat per user per 2 minutes

Current rate limit is 2 queries per chat per 2 minutes

Youtube link preview grid

You reply a youtube link with /preview or type /preview youtubeurl

You get preview image sprite from youtube seek bar

Give context to chat bot

Context includes:

• The name of user
• The username of user
• Current time in Jakarta
• Summary of top 10 hackernews topics?

Grain: Quote to Canva-ish art

You type /quote your text here

Your text will be rendered as image

Add support for vxtwitter and ddinstagram

• cleaning dirty twitter link returns clean vxtwitter link
• cleaning clean twitter link returns vxtwitter link
• cleaning dirty vxtwitter link returns clean vxtwitter link
• cleaning clean vxtwitter link returns twitter link
• cleaning dirty instagram link returns clean ddinstagram link
• cleaning clean instagram link returns ddinstagram link
• cleaning dirty ddinstagram link returns clean ddinstagram link
• cleaninf clean ddinstagram link returns instagram link

Grain: Giphy

• Giphy client #15
• RX debouncer
• LRU rate limiter
• Automatically pick small image as thumbnail #15

Handle reddit shortened tracking link

https://www.reddit.com/r/indonesia/s/y41QMsti9g

jadi dari link ini, sekarang udah gabisa replace domain name doang.

harus buka, follow redirect. terus nanti ketemu link gini

https://www.reddit.com/r/indonesia/comments/17k0h25/apa_sesusah_itu_dapat_kerja_di_perusahaan_yang/?share_id=9Wf-fXimS46Nv4jkLhl8M&utm_content=1&utm_medium=android_app&utm_name=androidcss&utm_source=share&utm_term=1

baru di cleanup, replace hostname.

https://libreddit.teknologiumum.com/r/indonesia/comments/17k0h25/apa_sesusah_itu_dapat_kerja_di_perusahaan_yang/

`/tldr` Overhaul

• Executing /tldr without replying and without argument will summarize the current telegram chat
• Replying /tldr to a link or supplying link as argument will summarize the content of the article

Security concerns:

• Fetching link is dangerous. Need design

Flip Image

Cost reduction using policies

Non TEKNUM group:

• GPT 3.5 turbo for chat
• No vision
• SD 1.4 for image generation

Private chat:

• GPT 3.5 turbo for chat
• No vision
• No image generation

TEKNUM:

• GPT 3.5, upgradeable to GPT 4 turbo when required
• GPT 4V turbo for vision
• SDXL 1.0 or DALL-E 3 standard for image generation, depending on usage

Continuing conversation after image generation

Actual Behavior

Currently, generating image using AI will stop the conversation.

Expected Behavior

The AI should continue the conversation even after image generation is done.

Dad jokes

• Fetch 5 random jokes
• Key: User id
• Cache duration: 1 hour

Do not respond to thank you or appreciation message

Grain: Inline query handler

Improvements to /map

Objectives:

• Change zoom level to 13
• Add clickable link to open the place in google maps

Hex color to image

Bug: Banned keyword not handled correctly when generating image using SDXL

The API behaves differently than what's documented

Places search

• Use google api
• Result 1: Coordinates (share loc)
• Result 2: Static map

Random Meme Generator

• Command: /meme your text here
• Select random meme template
• Add text

`/map` command should use the returned place object to provide users with more information

Google trends bot

You type: /trends keyword1 keyword2

You get rendered chart from google trends

`/map` command should be biased towards Jakarta or Java or Indonesia

Handle cleaning tokopedia short links using `/clean`

Example link: https://tokopedia.link/S0B7SJVfLtb

Grain: DDG search

• Implement Safe Search
• Implement DDG html scraper
• Display inline result as follows:

Inline Results

• Perfect Match: 1 item
• Combined Image Results: 4 items
• Search Results: 5 items
• The rest of Combined Image Results
• Fancy Texts

Anagram