Conduit

Conduit is an LLM-based text editor, with an intuitive and powerful graphical user interface. Employ an AI assistant to help you write without typing any prompts!

Contents

• Demo
• Setup
• Documentation
• (Planned) Features

See conduit-backend for backend deployment of this technology.

Demo

Edit mode

Edit mode allows users to edit the document node and text content.

View mode

View mode allows users to view their document without any obstructions or accidental edits.

Prompt menu

There is a prompt menu associated with each node, that allows users to interact with an LLM via preset prompts.

Prompt responses

Responses to prompts for are collected underneath the node they were applied to.

Setup

Run with built-in API

This is the easiest way to run Conduit, as it uses the local API built into the frontend. This means that all files are stored/read from the system of the machine that Conduit is running on. However, there is obviously no syncing between different machines or choices of generative AI usage (there is only currently support for OpenAI locally).

1. Add the following environment variables to a .env.local file in the root directory of this project:

   Variable	Value
   NEXT_PUBLIC_INTERNAL_API_URL	http://localhost:3000/api
   STORE_LOCATION	The absolute location of the folder you want to read/write files to on your machine (e.g., /home/<user>/Documents).
   OPENAI_API_KEY	An OpenAI API key.

2. Install dependencies:

   npm i

   • If you are pulling after recent commits (and dependencies have changed/updated), reinstall dependencies:

     npm ci

3. Run in either:

   • Development mode (auto-reload without pre-rendering):

     npm run dev

   • Production mode (pre-rendering without auto-reload):

     npm run start

Run with external API

This is how Conduit works in production: a separate backend handles authentication, file storage, and generative AI usage. Any backend can be used, so long as it hosts these endpoints.

1. Add the following environment variables to a .env.local file in the root directory of this project:

   Variable	Value
   NEXT_PUBLIC_INTERNAL_API_URL	The root URL of the backend that hosts these endpoints (e.g.,http://127.0.0.1:8000).

2. Install dependencies:

   npm i

   • If you are pulling after recent commits (and dependencies have changed/updated), reinstall dependencies:

     npm ci

3. Run in either:

   • Development mode (auto-reload without pre-rendering):

     npm run dev

   • Production mode (pre-rendering without auto-reload):

     npm run start

Documentation

• API reference

(Planned) Features

1. File upload
   • Copy file content
   • Summarise file content
   • Consolidate content from multiple files
   • Individually extract text, images, LaTex, and code
2. Text highlighting
   • Re-write (in the style of X)
   • Summarise
   • Expand
   • Create examples
   • Draw diagram
   • Generate code
   • AI-generated custom suggestions, based on context
   • Translate
   • Generate todo list (steps required to complete something)
   • Implement feedback (paste in the written feedback someone has given you)
3. Multimedia
   • Summarise audio (from links)
   • Create images/diagrams of content
   • LaTeX support via KaTeX or MathJax
4. UI
   • Movable, collapsable generated material
   • Code syntax highlighting
   • Search bar
   • Drag and drop with option buttons to the side of nodes
   • Add page columns to support different page layouts
   • /create page where template document prompts can be selected, or multiple files uploaded to be summarised/based-from/etc
   • Side panel for "Research this content"
5. Spell/grammar checker
6. Customisation
   • Allow people to write their own prompts (in development, just use CSV/JSON)
7. Social media features
   • Profile pictures, friends, etc.
   • Sharable, user-created prompt libraries.