Goodbye everyone!
As you might have noticed, the landscape of LLMs has changed dramatically in recent months. With the introduction of models boasting a whopping 100,000+ tokens of context memory and the big players like OpenAI, Google, and Anthropic integrating RAG directly into their official websites, I feel like this repo has served its purpose.
I've also been focusing on other projects and prioritizing my personal life and mental well-being. As a result, I haven't been able to dedicate time to updating this repo, which is evident from the lack of commits in the past few months.
Looking back, this repo was initially created to address the limitations of the 4096 context GPT-3, as that context size was simply too small. However, with the advancements in LLMs and the emergence of better GUIs, I believe that higher context is now preferable to using embeddings for RAG.
If you're looking for some awesome LLM GUIs, here are my top recommendations:
(Plz don't unstar this repo🥺)
Run gpt-3.5-turbo
or any other GPT models(text-davinci-003
) with this program! \
Use gpt-4
or gpt-4-32k
to use the new GPT-4 models if you have access. \
You can switch models in the config.json
file. \
It's like https://chat.openai.com/ but in your CMD and better(in terms of memory). \
You can add custom initial prompts and save/load your chat history! \
Download and double-click the GPT3Bot.exe
or run.bat
to run the program! \
In Linux and macOS, you can run ./GPT3Bot
to run the program. \
\
Click to download: Stable Release | Development Build \
Please check the Wiki for more information: Click Me
debug_reference
to true
in config.json
!/stop
to end and save the chat./undo
to undo your last prompt./reset
to reset your entire chat./dump
to dump your chat history to a .txt file inside the dump
folder./stop
, the program will ask you to input the filename to save. You can press enter directly to skip this and not save the chat. If you input any other text and then press enter, the chat will be saved into a json in the "saved" folder. When you open the program next time, you can simply input "s"(which means saved), press enter, then type the saved chat's json file's name to load your saved chat.config.json
, can be easily modified.Written in C++ (Libraries used: Boost, cURL, nlohmann/json, libproxy, cpp-terminal, ftxui, oneTBB, clip, cpp-tiktoken, pcre2, utf8proc)