v0.2.0

What's Changed

• Big refactor of client code: @wtlyu #140
  • Chat routes, navigate right to your favorite conversations by URL
  • state mgmt. switch from redux to recoil
  • less code, more organized
• Distinction between Fast vs. Balanced Tone Style for Bing @HyunggyuJang #127 #128
  • Resized Tone style selections fa0c56f
• CSS fixes, bug fixes, and better matching with official site #137
• You can now set your default model for Chat API completions in .env file, 45f4145
  • this will soon be an option on the frontend

Upcoming Changes

• More customizable parameters for both ChatGPT API & Bing
• Built-in multi-user support
• Test environment for CI/CD

Full Changelog: v0.1.1...v0.2.0

v0.1.1

What's Changed

Adds Tone Styles to Bing

• Tone Styles: Select between 3 options
  • Will persist across conversations
  • Tabs move along with textarea height
  • Only configurable at the start of a conversation
  • A lot of people prefer "Creative" with all convos with Bing/Sydney so I recommend this
• Minor updates:
  • Many bug squashing done by myself and @HyunggyuJang
  • adds very in-depth local install instructions thanks to @fuegovic
  • setup docker-compose environment variables for meilisearch

Immediate roadmap:

• more customGPT and Sydney parameters (response temps, changing openai models, editing system message for Sydney)
• Improve bing features (styling, params, fix formatting bugs)
• Cleanup code & writing a test suite for better CI/CD

Notable planned features:

1. Prompt templates
2. Keyboard commands
3. Folders view in Nav
4. Configurable settings per user for UI preferences

Full Changelog: v0.1.0...v0.1.1

v0.1.0

What's Changed

Adds SearchBar for Fast Message & Conversation searching

- Optional and configurable, powered by Meilisearch

• Basic usage: will search after short delay of last keystroke in searchbar
  • Shows up to 20 messages matching the query. No messages will result in Landing Page.
  • Navbar shows conversations with at least 1 hit in messages or title, sorted by most number of hits within a convo, to least
  • No conversation matches will show empty nav bar, and will show as many hits with criteria above
  • Easily clear the searchbar by hitting 'new chat'
  • Click a conversation title over a message to open a conversation and use as normal.
• Easy setup with docker, necessary environment variables shown and explained in .env.example

Click here for Full Setup Instructions

Enabling Search:

• Requires installation of free self-hosted Meilisearch or Paid Remote Plan (Remote not tested)
  • The easiest setup for this is through docker-compose, which takes care of it for you.
  • for local instructions, see the official meilisearch site
• Follow .env.example for proper env variable setup.
• Note: environment variables are DIFFERENT for Local vs. Docker use.

Local batch file instructions (windows)

@fuegovic (thank you for this): In case someone is wondering, here's the content of the batch file I used to get the search working without docker. The search function also needs to be enabled in the .env file.

@echo off

REM place the meilisearch executable at the root of the chatgpt-clone directory

start "MeiliSearch" cmd /k "meilisearch --master-key WjRvKvpelIYHa4dlcDGsdT9c0CKRbY-4VcuCQ1iZPcc

REM meilisearch is the name of the meilisearch executable ↑↑↑

REM meilisearch will generate a master key when you start meilisearch without providing one

start "ChatGPT-Clone" cmd /k "cd api && npm start"

REM this batch file goes at the root of the chatgpt-clone directory (edited)

I think that's the bare minimum to have it working, I just downloaded the meilisearch executable from github and placed it in the root folder after pulling the search-final branch (I also did the npm ci and npm build steps once after updating)

Other changes

• Clearing all conversations now requires confirmation from prompt
  
• Edit user messages in conversations with Sydney for branching the conversation #115
• Supports changing ChatGPT Browser model through .env file, see .env.example for proper env variable setup.
  • This enables use of GPT-4 for ChatGPT Plus members through the ChatGPT Browser model #114

Immediate roadmap:

• more customGPT and Sydney parameters (response temps, changing openai models, editing system message for Sydney)
• Improve bing features (styling, params, fix formatting bugs), thank you @HyunggyuJang
• Cleanup code & writing a test suite for better CI/CD

Notable planned features:

1. Prompt templates
2. Keyboard commands
3. Folders view in Nav
4. Configurable settings per user for UI preferences

Tagging for major contributions up to this point @wtlyu

Full Changelog: v0.0.6...v0.1.0

v0.0.6

What Changed

This is the latest stable release before message/conversation search goes live!

By request, there is now also a community discord server

• Significant improvements across the board to existing features, the biggest of which is better Markdown handling
  • Cursors follows text better and correctly blinks as expected
    • If there is ever any choppiness, this is due to the AI stream which is not easily controlled and random.
  • Code blocks more stable
• Sydney is once again available through new jailbreaks (though formatting still needs improvement in edge cases)
• Uses latest api client, allowing usage of GPT-4 (must be manually set at the moment)
• Better docker config with options #108

Following soon after this release (tomorrow):

• Search all messages and conversations feature (optional and requires meilisearch, provided through docker)

Immediate roadmap:

• more customgpt/bing parameterss (response temps and changing openai models)
• improve bing features (smoother streaming, style params, fix formatting bugs)
• cleanup code & writing a test suite for better CI/CD

Notable planned features:

1. Prompt templates
2. Keyboard commands

As always open to more integrations, contributions, suggestions as they come up and time allows

v0.0.5

Changes

• Adds Stop generating & Regenerate Message options #88
• Skeleton for Multi-User support (bare bones integration, currently DIY) 2ad9b33 8029ed7
• Models without credentials are omitted from model menu 29ee765
• Better titling (and titling as a separate request from messages) f1e3bbd
• Various UI fixes/additions
  o Current model highlighted at the top of the conversation
  o Better textarea and button styling for mobile

On my immediate to-dos for this weekend:

1. search messages feature (75% done already)
2. more customgpt params (including changing openai models)
3. rewrite code/markdown handling (I have many things in mind to significantly clean this up) #67 #69

Stretch Goals if time allows:

1. improve bing features (smoother streaming, style params, fix formatting bugs)
2. cleanup code & writing a test suite for better CI/CD

v0.0.4

Many thanks to @wtlyu as they initiated and coded a lot of the features in this release, see a lot of the work done here: #55

Changes

• Resubmit messages & Branching messages

• Mirrors official ChatGPT feature: edit sent message to branch messages (works only with ChatGPT)

v004demo.mp4

• Paginated conversation list

• Model menu icons

Immediate Roadmap

• Message search
• Multi-user system
• Bing AI Styling (for conversation style, suggested responses, etc.) - In progress
• Address issues & feature requests
• See complete roadmap in readme.md

v0.0.3

Changes

This is to mark a new tag before a major PR merge, to properly document the latest changes before.

• Mobile Styling added #43
• Build docker container with one command now from root dir #56
• Auto focus support added to Chat Input #58
• Model menu key issue fixed #54
• Titles now generates in user language #40
• Host name and proxy address configuration support added #26
• Added Bing styling (citations with links that open in new tabs) #23
• Spinner indicating messages are loading
• Unsupported code language errors & misc. reported errors fixed

v0.0.2

Changes

• Adds Sydney (jailbroken Bing AI) to the model menu, thanks to DavesDevFails for bringing it up: #13.
  • No limits on Bing AI through Sydney
  • Keep token usage small (words/messages/convo) to avoid any Jailbreak errors
• Bing/Sydney now correctly cite links, with sources appended to message. More styling to come.
• Minor fixes/improvements
  • Ensures new conversations are created on all model state changes
  • Model menu won't close on model deletion
• Re-enabled ChatGPT free since it might be working for most people (not for me)
• First outside contributions! Thank you HyunggyuJang for #11 & #12

v0.0.1

The app is dockerized as of this first release for quick setup! See full setup instructions here.

Features

• Response streaming identical to ChatGPT through server-sent events
• UI from original ChatGPT, including Dark mode
• AI model selection (official ChatGPT API, BingAI, ChatGPT Free)
• Create and Save custom ChatGPTs*
• Dockerized and easy setup instructions

^* ChatGPT can be 'customized' by setting a system message or prompt prefix and alternate 'role' to the API request

Roadmap

Warning

This is a work in progress. I'm building this in public. FYI there is still a lot of tech debt to cleanup. You can follow the progress here or on my Linkedin.

Here are my recently completed and planned features:

• Persistent conversation
• Rename, delete conversations
• UI Error handling
• Bing AI integration
• AI model change handling (start new convos within existing, remembers last selected)
• Code block handling (highlighting, markdown, clipboard, language detection)
• Markdown handling
• Customize prompt prefix/label (custom ChatGPT using official API)
• Server convo pagination (limit fetch and load more with 'show more' button)
• Config file for easy startup (docker compose)
• Build test suite for CI/CD
• Conversation Search (by title)
• Resubmit/edit sent messages
• Semantic Search Option (requires more tokens)
• Bing AI Styling (for suggested responses, convo end, etc.)
• Prompt Templates/Search
• Refactor/clean up code (tech debt)
• Optional use of local storage for credentials
• Mobile styling (half-finished)
• Deploy demo

Also worth noting, the method to access the Free Version is no longer working, so I've removed it from model selection until further notice.