Elive has made a tool that can use some features of AI with zero tracking , the tool can be used in both GUI and CLI (elive-ai command), it includes really-amazing features that can help your life in daily tasks, I cannot describe here how good these features so you must try yourself all the options!
Usage: elive-assistant mode message
GUI: use with --gui if you run it from a terminal but want some GUI features
You can also use this tool as a pipe, e.g. echo 'hell0 wrold' | elive-assistant corrector
--gpt4 Use the GPT-4 model instead of the default one (GPT-3.5 turbo)
--stdout Always show results in stdout, useful for getting results from the tool
Text modifiers:
translate: Translates from any language (even if mixed) to your local language used in your Elive system
translate-improved: Translates from any language (even if mixed) to your local language, with literary improvements
translate-en: Translates any text to English, with well-formed sentences and good, elegant English
corrector: Corrects the grammar any given text (in any language)
proofread: Corrects the grammar and slightly improves any given text
improver: Rephrases a text in a deeper, more elegant and literary way, showing different options
revise: Act as a book editor, fixing the grammar, improving clarity, consistency of style, etc..
rearrange: Re-arrange the contents of a text to organize it in an uniform way
synonyms: Writes synonyms for a given word, or defines a concept. e.g. 'computer portable' -> laptop
rephrase: Rewrites a text in a different way
persuade: Rewrites a text in a more convincing way
organize: Rewrites your text in a list of points in a logical order
summarize: Summarizes a text, removing useless and unneeded parts
simplify: Simplifies a text as much as possible, making it shorter and concise (lacking emotion)
humanize: Make the sentence to sound more like human
nosell: Remove the selling sounding-like references from a sentence
Tools:
explain: Explains a concept, jargon, abbreviation, or anything!
ocrfix: Fixes text like the ones wrongly detected from a scanned document
code: generates a source code from a well defined description, ex: in ruby, show the local weather
listen: Special mode to listen your voice input, returning transcribed text as output, second argument can be the language forced to listen from
Creative:
title: Suggests a good title for a specific purpose
titlepost: Suggests a good engaging (website) post title
intro: Writes an introductory text for an article or similar
story: Narrates a story based on a description
idea: Gives ideas of how to continue a text
propose: Proposes a text with more contents and ideas, like continuing it
expand: Expands a text with more and richer contents
product: Describes a product
seo: Creates a Google-like snippet description to fill in your SEO plugin
recipe: Creates a recipe based on your given ingredients
ux: Designs an UX concept based on the described needs
copywriter: Writes about any topic from your description or details
techwriter: Writes articles related to technology
domain: Gives (web) domain name ideas
Tools and Engines:
audio: Transcript in text a given audio file
speech: Speak a text
paint: Paint / design an image based on your description
deepl: Translates from any language (even if mixed) to your local language using the deepl service
Self-Help:
therapist: Receive help from different topics related to mental health
doctor: Describe a pain or problem in detail to get health directions
inspireme: Get a personalized sentence to inspire you!
suggest-book: Describe what type of book you want to read and get a good suggestion list
elivehow: Ask how to do something in Elive (experimental)
Funny:
friend: Speak as if you were talking to a friend
girlfriend: Your personal girlfriend (lol)
yoda: Speak with Yoda from Star Wars
Environmental variables:
LANG: this variable defines your local language, or to which language actions should be spoken, like en_US.UTF-8
LANG_SOURCE: forces the language input to be recognized as X
TEXT_TONE: this variable defines the tone to write into, like formal or informal (deepl only)
Notes:
- You can write in any language and it will reply in the corresponding one. There is a single answer, but in the GUI mode you can follow a conversation.
- You can concatenate results, for example: echo 'h0l4 mznd0, cómo est4s?' | elive-assistant ocrfix | elive-assistant translate-en
Yes you have read correctly, you can even have a girlfriend there
The last time I encountered AI chatbots (around 2000 IIRC) it was like talking to morons who only rephrased the question where now I'm talking to a kindergarten kid who can google (fast).
Anyway, above answer makes abundantly clear that I'm going for 'user experience' where you're going for 'user interface'.
Where the '--gui' flag has a very different answer
It's a bit childish and sometimes outright daft.
If I ask for code in Bash it gives me 'python' (sometimes with import sys ) and if I ask for 'zsh' code it shows lots of errors and takes ages but eventually does come with some correct answers ...... albeit that I can see them as correct because I know what the answer should look like.
Try this and go make a cup of tea while waiting.
elive-ai code in zsh, copy entire directory to another
yes it uses chatgpt api, without tracking / saving any info (not conversations, one-shot answers)
In fact the "code" feature doesn't works well, I think the api works very different than the interface because the results are too different, try the same thing in chatgpt directly and you will see the code is much better quality, the "code feature in elive-ai" is just added as a small way to have such feature, if im not wrong it also request a short result like a small example of code when requesting something specific
I'm a long time user of Elive and I'm very impressed by the development team's commitment to creating an amazing Linux experience. I'm also impressed by the responsiveness of the community and the willingness to help others. Elive is an excellent choice for anyone looking for a reliable, secure, and easy-to-use operating system.
Not to bad.
Now I'm using the free keys for this Elive-AI and was wondering if anyone has tried non-free keys, if possible?
It's actually quite the shitty standard bullshit one gets to hear, no matter who or what is the subject .... we're too easily awed by: "Not really much!"
A lot like CGI in movies: Every time we think it's great and realistic and then a few years later, "How did I ever fall for that?". It evolves and we evolve with it.
Same goes for AI generated art ...... how did that ever get past serious scrutiny ? .... It's kitsch!
I agree that AI is a tool that needs leading through cleverly phrasing questions and @Thanatermesis has understood that quite well.
I just asked:
elive-ai code in yad, create radiolist with 2 choices
Where I'd just researched the correct answer for another thingy:
yad --radiolist --list --geometry=300x150+700+700 --column=Tick --column="Choose ..." false "Do not ask me again." false "Remind me next time."
Albeit the answer goes a long way with some options that might come in handy one day ..... it still struggles.
Here's a similar as your query ... but invoking the devil :
On a serious note on the privacy aspects of, in this case elive-ai.
Where it's true that 'elive-ai' does not chat nor does it use a web-browser to access the AI-bot, it does require you to have a personal key that's tied to your email and telephone-number.
The keys (you're allowed 2 as a free user) can be tied specifically to you personally and like when doing searches on i.e Google that indirect data can be tied to you personally.
So the "non-tracked & on-demand" comment can be taken with a big grain of salt.
Even every question asked, when combined and tied to a single entity will create a personal digital profile that can be accessed if required or wanted.
So despite the initial WoW! and the nice option in Elive to enable this function, allowing for some playfulness ....... take a minute to think about what it could mean in the long run.
And ....... remember it's practically the same as search engines like 'Google', 'Bing', 'Alexa', etc, etc where the quality of the question defines the answer and that the user is not the customer but the product.
Here's an article that doesn't try to scare it's readers but does explain potential future risks.
As for Elive, I still think it a bad choice to have 'Google Chrome' with 'Google-search' as the default. Like with 'elive-ai' it should be 'opt-in' and not 'opt-out', leaving it a personal choice whether to use those or not.
There is an option to opt out of data-collection but that is down to requesting it from OpenAI and hoping they will actually do so;
The other option, for if you've had enough is to delete the account with it's data ...... again by request.
Here's how:
I already write, that using in Elive some google things by default is sad...
Can be a little like using Linux but still give datas & hardware power to Minus$oft...
So these things can be improved in Elive, for example:
force using new api key each month
using an alternative to chatgpt (which at the moment there's not, but there will be pretty soon)
Note also that the uses developed on elive-ai is not something like for "dialogue with the entity" (and thus learning from you), I use it many times and its basically to help me improving texts, rephrasing them, correcting grammar errors, etc... so more like a "tool" than a "friend"
Maybe I can add some feature in the elive-installer (another "fine-tune window") for privacy-related options (like the one with the "non-free software" options), including a selection of things like the default search engine (even if I always said the user can always set / configure the default search engine on their browsers)
Update: the last version includes some improvements and a better engine set by default (gpt-3.5-turbo), which has slightly better results, especially for the "code" mode which didn't worked so well
Don't be silly. Comparing/saving the questions and bringing that back to a single entity is a very absolute way of creating an image of someone's mind or the state the person is in, as well as the surrounding social group....the algorithm is clever enough for that.
In Exemplis: Let someone setup a query-form and you'll know exactly what he personally wants or prefers.
Rotating the key isn't going to help unless you care to use different e-mail and phone-numbers and be aware that the sharing of keys will be stopped by OpenAi, as by their own words in the key-generator:
<quote
Do not share your API key with others, or expose it in the browser or other client-side code. In order to protect the security of your account, OpenAI may also automatically rotate any API key that we've found has leaked publicly.
/quote>
meaning they know exactly who's who.
That's just it .... it's a curated answer that you get and you have no power into how it's curated .... the algorithm defines what your answer/world looks like and ultimately what you should believe, taking away all personal form of control or choice.
With a text search you still have an opportunity to go through the findings and decide what you want or need.
We're looking abuse in the face and are actually rejoicing in the same way people cheer a dictator. The arguments made pro ChatGPT/OpenAI are the same as those used to support totalitairian leaders:
trains run on time
Overt corruption changed
No long debates to get a decision done
Others will be worse
He's very clever/wise
etc, etc.
Be assured that OpenAI, despite it's name is NOT open at all ...... they just want/need the community to help them set up shop, same as Google (and M$) did when it started out.
AND:
I personally have not seen any (and I really mean any) given answer that goes above general, non-creative ramblings unless you give the AI good and exact arguments and facts to go on.
The trouble is if you can give that: You already know or gave the answer yourself!
IMO it's "SnakeOil" big-tech is trying to sell to the world making it's inhabitants stupider by the day.
what about that? sharing the API keys between all the users? everytime is used, a random api from other users are used
unfortunately this can create a profile or a ban of the user's email but sounds like a perfect solution
I wonder if could be possible to create a single API key to be used by all the users of Elive that sounds like a better solution, it will be also not needed to insert an api key but using the default one (is not this the solution duckduckgo used, @triantares ?)
I think this can be a good way to solve the privacy issues
I think you need to try more all the options of elive-ai in a real usage/need betatesting, elive-ai is not something like "I need an answer / search result / conversate", is just a tool that can improve your life in many aspects, for example the text corrector or grammar improver is just amazing in an unbeatable way (no other tool can do a job so good), the title suggester is very good too, the doctor mode is just surprisingly helpful, the summarizer & simplifier is quite useful too
@triantares : for the Duchy one, based on wikipedia is not "based", when you know that even wiki's creator is disgusted about how it turns... @Thanatermesis : So what about Eltrans? Still needed I think!?
Yes @Franc , eltrans is needed before i used reverse-engineered google-translator to improve the english sentences, now elive-ai (corrector mode, sometimes proofreader mode) improves them a lot too but in a different way (more like a grammar correction)
, the tool can be used in both GUI and CLI (elive-ai command), it includes really-amazing features that can help your life in daily tasks, I cannot describe here how good these features so you must try yourself all the options! 
:
before i used reverse-engineered google-translator to improve the english sentences, now elive-ai (corrector mode, sometimes proofreader mode) improves them a lot too but in a different way (more like a grammar correction)
