we want to know the source of the answer context

Improve the user experience by clearly distinguishing between questions and answers in the chat interface, providing visual cues or formatting to differentiate between user queries and system responses.

right now there's 1000 char chunks by default which makes no sense

Implement a first version of a way to connect to external knowledge.

Introduce redux for state-management to keep the external-connection in the App.

Implement scroll functionality for the chat page to allow users to easily view and navigate through previous messages in the chat history.

find some inspiration and a way to implement here:
https://github.com/grumpyp/AI-video-genie

As of now this is still more like a MVP.

Refactoring in every dimension is always welcomed. Feel free to open a PR!

Hitting the quota limit for the API (by uploading a file with too many tokens) for a given key throws an error. The error currently isn't handled and it presents itself an axios error on the UI.

Similar to the pdf_loader, a good start would be a loader for powerpoint files.

https://github.com/grumpyp/aixplora/blob/main/backend/loaders/pdf_loader.py

Currently, we're not supporting .xls.

python-docx needs to be added to the requirements.txt in order to use the docx_loader.
Seeing error in current build:
ModuleNotFoundError: No module named 'docx'

follow up #70

as a user I would like to save my summaries as some of them are not cheap and I also would like to re-access them once in a while.

Implement a first draft to connect AIxplora with the AIxplora knowledge.

This can then be changed via frontend to connect to other shared knowledges.

Similar to the pdf_loader, a good first start would be a .txt loader!

https://github.com/grumpyp/aixplora/blob/main/backend/loaders/pdf_loader.py

right now the chat is saved in localstorage. we want to save it in our database and work with different chat's

have seen it several times that people's API keys have rate-limits or errors,

If the error is not related to the app, we should add more verbose to it like with a toaster or something

The OpenAI limits audio files to 25 mb max, so if I upload a file with 100 mb ~ 1hr podcast, it will break.

Implement a mechanism to break it down and index the chunks.

add ability to toggle between light and dark theme, the theme icon should be in the navbar, (Menu.tsx)

right now we have a discord service, Github and an empty website, we need to fix the footer

https://github.com/grumpyp/aixplora/blob/main/frontend/src/renderer/components/Footer.tsx

Right now it's not possbile to ask questions based on a specific document. That should be possible to get better results.

We want user to be independent from OpenAI, so we need to introduce a integration for different LLMs.

make it able to choose between the model in the config, so I can save my key with a model reference.

introduce a global variable and change each model reference in the backend code to it

We want to be independent and make use of all kind of embedding algorithms.

Improve Dropzone.tsx

1. When selecting multiple files for upload, display all selected files.
2. Add cancel buttons next to each file for individual selection cancellation.
3. Current behavior: only the last file is displayed, no cancellation option.

i would like to work on this issue if you don't mind

The current README is from our PoC, now we added several loaders and other features.

A brush up would be super nice!

metadata like filetype and page/min of the related content.

the first test will be added in #55

from there on we want to include more tests to cover to main functionality.

every pr should execute the test's

We need a refactoring / cleanup for the frontend to make components and stuff reusable.

Since Node Sass always explodes when used in projects (I have a decade of experiencing this) I can't install the project.
Is it possible to use Sass or Dart-Sass instead of node-sass? Or maybe the issue lies somewhere else but from experiance its node-sass being an (s)ass. Look at their issues, like around 80% of the issues involving people not being able to install it.

npm ERR! code 1
npm ERR! path /Users/***/node_modules/frontend/node_modules/node-sass
npm ERR! command failed
npm ERR! command sh -c node scripts/build.js
npm ERR! Building: /Users/***/.nvm/versions/node/v16.16.0/bin/node /Users/***/node_modules/frontend/node_modules/node-gyp/bin/node-gyp.js rebuild --verbose --libsass_ext= --libsass_cflags= --libsass_ldflags= --libsass_library=
npm ERR! gyp info it worked if it ends with ok
npm ERR! gyp verb cli [
npm ERR! gyp verb cli   '/Users/***/.nvm/versions/node/v16.16.0/bin/node',
npm ERR! gyp verb cli   '/Users/***/node_modules/frontend/node_modules/node-gyp/bin/node-gyp.js',
npm ERR! gyp verb cli   'rebuild',
npm ERR! gyp verb cli   '--verbose',
npm ERR! gyp verb cli   '--libsass_ext=',
npm ERR! gyp verb cli   '--libsass_cflags=',
npm ERR! gyp verb cli   '--libsass_ldflags=',
npm ERR! gyp verb cli   '--libsass_library='
npm ERR! gyp verb cli ]
npm ERR! gyp info using [email protected]
npm ERR! gyp info using [email protected] | darwin | arm64
npm ERR! gyp verb command rebuild []
npm ERR! gyp verb command clean []
npm ERR! gyp verb clean removing "build" directory
npm ERR! gyp verb command configure []
npm ERR! gyp verb check python checking for Python executable "python2" in the PATH
npm ERR! gyp verb `which` succeeded python2 /Users/***/.pyenv/shims/python2
npm ERR! gyp ERR! configure error
npm ERR! gyp ERR! stack Error: Command failed: /Users/***/.pyenv/shims/python2 -c import sys; print "%s.%s.%s" % sys.version_info[:3];
npm ERR! gyp ERR! stack pyenv: python2: command not found
npm ERR! gyp ERR! stack
npm ERR! gyp ERR! stack The `python2' command exists in these Python versions:
npm ERR! gyp ERR! stack   2.7.18
npm ERR! gyp ERR! stack
npm ERR! gyp ERR! stack Note: See 'pyenv help global' for tips on allowing both
npm ERR! gyp ERR! stack       python2 and python3 to be found.
npm ERR! gyp ERR! stack
npm ERR! gyp ERR! stack     at ChildProcess.exithandler (node:child_process:398:12)
npm ERR! gyp ERR! stack     at ChildProcess.emit (node:events:527:28)
npm ERR! gyp ERR! stack     at maybeClose (node:internal/child_process:1092:16)
npm ERR! gyp ERR! stack     at Process.ChildProcess._handle.onexit (node:internal/child_process:302:5)
npm ERR! gyp ERR! System Darwin 22.5.0
npm ERR! gyp ERR! command "/Users/***/.nvm/versions/node/v16.16.0/bin/node" "/Users/***/node_modules/frontend/node_modules/node-gyp/bin/node-gyp.js" "rebuild" "--verbose" "--libsass_ext=" "--libsass_cflags=" "--libsass_ldflags=" "--libsass_library="
npm ERR! gyp ERR! cwd /Users/***/node_modules/frontend/node_modules/node-sass
npm ERR! gyp ERR! node -v v16.16.0
npm ERR! gyp ERR! node-gyp -v v3.8.0
npm ERR! gyp ERR! not ok
npm ERR! Build failed with error code: 1

npm ERR! A complete log of this run can be found in:
npm ERR!     /Users/***/.npm/_logs/2023-06-12T10_13_25_623Z-debug-0.log

right now there's no auto-scroll, and once switching to the Chat route, it doesn't "load the last message"

we also decided to not implement graph in this issue, so delete this

Take this as a reference:

https://github.com/grumpyp/chroma-langchain-tutorial

Can use pubmed API, e.g.

https://www.ncbi.nlm.nih.gov/pmc/tools/textmining/

Seeing the below error when trying to pip install -r backend/requirements.txt
"RuntimeError: uvloop does not support Windows at the moment".

Right now we are using 3.5, but people may want to use gpt4 as it's definitly better.

We should integrate a way to choose it, maybe in the config,..

we could use the easy way, by depending on LangChain, or using our own implementation

https://python.langchain.com/en/latest/modules/indexes/document_loaders/examples/obsidian.html

A ´.epub` loader would be a great features to load in ebooks.

npm start from a clean install on WSL:

TypeError
Cannot read properties of undefined (reading 'ipcRenderer')
Call Stack
 ./src/renderer/index.tsx
  renderer.dev.js:66999:17
 options.factory
  renderer.dev.js:78064:31
 webpack_require
  renderer.dev.js:77510:33
 undefined
  renderer.dev.js:78649:37
 undefined
  renderer.dev.js:78652:12
 webpackUniversalModuleDefinition
  renderer.dev.js:7:11
 undefined
  renderer.dev.js:10:3

we want to query videos.

follow up ticket:

chat should return timestamps once it's relation in the vectors found the metadata "video", so people know where to watch

Ensure that when users navigate to different pages within the application, the content of the chat remains intact so that users can seamlessly switch between pages without losing their chat conversation.

right now people on windows have issues to setup the project #28 #30

Docker would be the way to go for now

please use the mantine notification

Right now we want the LLM to answer our question based on context. But what if we want it to write a Text based on the context, or a song?

Would be a great feature to make this optional to give additional input (prompt)

for larger files or audio it takes a while to index, a async way would be nice

EDIT: It's async already, but no visual on that. could be confusing if the Chat wouldn't work.

We need tests for the API and the database

feel free to work on it!

Currently we're not supporting .csv nor .xlsx

Good way to start getting into this project, feel free to work on it!

As a scientific user I want math formulas to be rendered. E.g.

The formula provided is for calculating the probability of an attacker catching up from a given deficit in a blockchain system, based on the Gambler's Ruin problem. The attacker's potential progress is assumed to follow a Poisson distribution, with an expected value of \ue0c1=zq p, where q is the probability an honest node finds the next block, p is the probability the attacker finds the next block, and z is the number of blocks behind the honest chain the attacker currently is.

To calculate the probability that the attacker could catch up now, the Poisson density for each amount of progress the attacker could have made is multiplied by the probability he could catch up from that point. This is then rearranged to avoid summing the infinite tail of the distribution.

The provided C code converts this formula into code and can be used to calculate the probability for various values of q and z. The probability drops exponentially with z, and the formula assumes p > q.

Overall, this formula and code are used to assess the security of a blockchain system and evaluate the likelihood of an attacker catching up with the honest chain.

some tasks take time to run, like uploading a file, or waiting for an answer, implement a progress bar or a loader, while the user wait for the task to end

I would like to have summarizations of documents, especially of my scientific papers I use for my thesis atm.

Is there a feature to remove files uploaded from the DB?
Right now, once I encounter a rate-limit error once, I'm unable to remove the file from the Files section