Wow, this is amazing. Thanks for putting it together. I'm usually allergic to the command line, but I'll give it a go! Can you add custom instructions to each GPT? And how about custom GPTs that I've created in Chat GPT? Can I still access them? And it would also be cool to organise chats in folders. Do you think they'll add that feature?
Thank you for your kind words, Ian! I'm glad you found the video helpful. Custom GPTs would ideally need to be copied over. There is something better called "Assistants" which run the same as GPTs and is part of the OpenAI API which plugs fully into LibreChat. It's all so powerful all we need is the time to set it all up perfectly! 😉 Also, organizing chats in folders would be a great feature to have.
@@CreatorMagicAI thanks! I think Tim and I are going to set up a Linode server and follow your instructions. I'm not sure I'm ready to give up ChatGPT Plus quite yet, as I really need my Custom GPTs. I'm also a little worried about the cost. You're probably right that it'll cost less than paying for the pro versions. But for me, I'd love to have a play with ChatGPT and Gemini as I've not really used the paid versions of those yet. Do you know if we can use the paid versions of Gemini with this too?
@@ianandersongray for sure, you can absolutely use paid Gemini Pro with LibreChat! The setup allows you to add API keys for pretty much any providers. I hear you on the custom instruction fear too. LibreChat has presets where you can create your own custom prompts/settings for different use cases but it may not be equal to GPTs. Good luck and give a shout if you get stuck.
Great job Mike! Clear steps & instructions on top of 2 different approaches. Question: how do you backup all chats/interactions with Libra Chat? (In case the server crashes or there's a disk corruption)
Good thought. Backup those chats like your digital life depends on it - because it just might! Should be easy as running `docker compose down` in your LibreChat directory, then back it up with `sudo tar -czvf librechat-backup.tar.gz .` That created tar.gz file then contains your whole digital life 🔮😳
This is my daily driver. Use it more than anything else. I have API costs but it's cheaper than monthly and I can use any model. I do miss Claude Projects and Artifacts though!
Hi Mike. Love this video so much as you know! Looks like the Google Gemini API isn't available in the UK and EU yet. I thought I found a workaround but I couldn't get it working. Have you got that working?
Exciting update on the LibreChat Ian - it's impressively powerful! I've set up my Linode server in the US, which gives me access to Google Gemini, and I highly recommend trying it out. Although it may not quite match the performance of ChatGPT and Claude, Gemini is still quite enjoyable to explore. Also, the latest Gemini 1.5 Pro with a 1 million context window is available with Google's Vertex AI and works seamlessly right out of the box in the UK and Europe! 🎉 Check it out here: cloud.google.com/vertex-ai
maybe lots of things have changed since you did this. i opened up port 3080 and i actually get chatgpt to respond now, but it doesn't do web search, latest data access is october 2023, and i don't see gpt-turbo like you, or any of the claude models. idk. maybe you can do an update?
thought this would be simple. i do have a server but use nginx proxy manager so not to just expose 3080 but chatgpt doesn't work though i see a list of ones (not included turbo) and don't see claude ones at all. sigh.
It took me quite a bit longer than a minute, but I got it to work. I, however, don't have access to gpt 4 or 4o. How does one get access to that? Does it require a separate API key? If we get it, how do we update that ENV file to our heart's content like you mentioned in the video?
I'm glad you were able to make it work! To access GPT-4 or 4o, you'll need to check with OpenAI and see if your API key grants access. It should do. Good luck!
You raise a valid point! API calls do cost money. So far I've been testing this and am happy with my costs plus the flexibility to have every powerful model at my fingertips without needing multiple monthly subscriptions.
@@CreatorMagicAI since i am using chat gpt 4 for most the applications, i think this solution would cost much more than 20$ per month, idk about the other apis but i highly doubt that it would in any case save money to use them over subscriptions. I would pick my words and title carefully, as misleading people might gain you views but not subscribers, I would avoid click bait. Good luck!
Yeah I'm currently paying for pro versions from Open AI, Anthropic, and until recently Perplexity Pro. I know from my token usage that Typing Mind and API is going to cost about half and have a better unified UI and chat history management.
@@tompaquette7138 did you have anything specific in mind when thinking about Home Assistant? Being a home automation system I wonder how you'd imagine this working together with it? Thanks!
@@CreatorMagicAI the LLM will see your paid list of smart devices, have a list of commands for Python on the Raspberry Pi it is hosted on, such as playing music, and can have a out loud conversation with you whether it is answering questions, being supportive, or displaying results on a specified tv. You can ask it to activate a specific smart device, create routines with it manually, or ask it to suggest routines based on the devices and their function.
for some reason it says that "docker compose up -d" is not a command. I may have messed something up in the nano process. I'm new to computer stuff like this, is there anyway I could restart the whole process, or is this just a bug you think?
@@CreatorMagicAI thank you! is there any chance you could tell me what this means? root@localhost:~/LibreChat# client_loop: send disconnect: Broken pipe
Thanks for the heads up about the security. You're right. It's a bit beyond the scope of this video but you can connect to a custom domain using an SSL certificate (that's what I'm doing) or even go for an easy solution like pointing Cloudflare Zero Trust direct to the IP. You then get the added bonus of Cloudflare security on your LibreChat instance.
Don't worry, you're not extra stupid, just API-curious! You can get the key for free on OpenAI's website. It's not connected at all to a ChatGPT Plus account and API will bill you per use which (depending on usage) can work out cheaper.
@@CreatorMagicAI haha API-curious is very funny cheers… But still using the word "free" so often on things that are billed by use is not the best move... 🤪
Well, nothing in life is truly free, right? The tool is 100% free and open-source. How you choose to use it may not be "free" (e.g. API calls to paid AI services). This could however be a completely free solution with a local LLM and that's something I intend to cover in a future video. Thanks for watching!
It sounds like you're encountering a file path issue. Make sure that the 'librechat.yaml' file is located in the correct directory. If it’s missing, you might need to create it, copy it from the example given or download it again!
