ChatFred: OpenAI's GPT-model workflow

[TomBenz]

On 4/17/2023 at 3:26 PM, chrislemke said:

Hey @TomBenz!

Thanks for your contribution and the code example. I was actually already looking into it. The problem is, that the whisper model is huge and most of them would blow up the workflow. At the moment I am having a look at this. 

from faster_whisper import WhisperModel

model_size = "tiny"

model = model = WhisperModel(model_size, device="cpu", compute_type="int8")



segments, info = model.transcribe("SOME_FILE.mp3", beam_size=5)

print("Detected language '%s' with probability %f" % (info.language, info.language_probability))

for segment in segments:
    print("[%.2fs -> %.2fs] %s" % (segment.start, segment.end, segment.text))

 

It is not only smaller and does not use so much CPU but it is also much much faster. Try it out. You can also change the model_size to: "tiny", "tiny.en", "base", "base.en", "small", "small.en", "medium", "medium.en", "large-v1", or "large-v2".

You get transcript for first 2-3 min and then for last few seconds of the video / audio. fyi.

[chrislemke]

9 hours ago, TomBenz said:

You get transcript for first 2-3 min and then for last few seconds of the video / audio. fyi.

Hey @TomBenz!

Thanks for the update. I suggest we continue the discussion here so we don't fill up this post with our talk about your idea. Okay for you?

 

BTW: I also tested "fast_whisper" and I have to admit that I am quite fascinated by it. It is really fast and the different models provide the opportunity to adjust the quality/speed ratio in a very simple way. But even the tiny model is pretty huge (~75MB) not really sure how to put this into ChatFred without requiring additional downloads/installs.

[energyd]

Hey @chrislemke

good stuff! I recently started using ChatFred with ChatGpt-4. One thing I really miss is the interactive experience I gained through chat.openai.com. I often found myself having to going back and forth with chatGpt and work towards my final solution. ChatGpt was asked to ask clarifying question to me first before working on a solution. 

 

With ChatFred, seems I can only send standalone query to openAI and then it would lost previous context when I ask a follow up question. Do you know a way to make it work interactively (e.g. remembering the context of the conversation)? 

Edited May 9, 2023 by energyd

[wellerpond]

Hello!

 

Just installed ChatFred and get a "There is something fishy with your internet" error message. Pops up in big letters across the screen. I can open all the prompts with the default key words, but it consistently gives me the same message.

 

My Internet is functioning as normal.

 

Any thoughts?

[seann]

Heyyy

I still got this issue even if I am paying for ChatGPT.

🚨 You have reached the rate limit. Check your settings in your OpenAI dashboard.

Any reasons for this?

 

[Vero]

6 minutes ago, seann said:

Check your settings in your OpenAI dashboard.

 

What do you see in your OpenAI dashboard?

 

This is the usage dashboard:

https://platform.openai.com/account/usage

 

And the rate limits:

https://platform.openai.com/account/rate-limits

[stonefury]

I am a new user and I paid $20 for chatgpt, so I assumed I was covered. But apparently I need to also pay openai to use the API?

It's a little confusing so maybe a little info on that for users. I too was getting 🚨 You have reached the rate limit. Check your settings in your OpenAI dashboard.  I had to go and setup a payment method for the API usage and set some $$ limits per month.  It seems the $20 a month has nothing to do with this.

 

 

[Vero]

@chrislemke might have some insight into what subscription is needed for the workflow at this time?  

[Ash35]

hi - 

I am not technical at all and heard about Alfred and just downloaded it and I can already tell that due to my lack of technical skills I am probably only scratching the surface of how powerful it can be. 

 The biggest advantage I have seen so far is the new chatfred plugin which is insane, especially the fact i can send audiofiles to get transcribed! As a marketer this is huge.

but I keep getting errors on it for instance "🚨Hmmmmm......something is wrong with your request. Try again later." 

I did some research and I saw someone mentioned that I have to download python for OpenAI on my computer so i followed the instructions ChatGPT gave me and ran commands in my terminal and I updated "pip" for my computer to know from now on "I am referring to Python 3"  and I thought that would help but it did not. 

Also whenever I press enter, I just get a huge 💬  emoji and nothing happens. 

Can someone help? 

https://www.loom.com/share/8b564eb6e80e4148bc62af8399d49d60

[albertkinng]

10 hours ago, Ash35 said:

I am not technical

I've been using Alfred since its debut, and it's important to understand that this app can be either simple or complex depending on your needs. If you're not tech-savvy, I recommend not experimenting with things you're unfamiliar with. Installing additional programming languages like Python involves using the terminal, which can potentially harm your computer, and no one will assist you afterward. The only challenge or advantage with Alfred lies in its versatility, making it difficult to create a tutorial or provide an easy learning curve. To fully benefit from it, you need to have technical knowledge. Chatfred is useful until your free API credits run out, but you can purchase an app for audio transcription in the app store with a one-time payment. Additionally, you can enjoy ChatGPT for free indefinitely with Wavelength on your Mac, which is perfect for non-technical users.

[Andrew]

3 hours ago, albertkinng said:

Installing additional programming languages like Python involves using the terminal, which can potentially harm your computer, and no one will assist you afterward.

 

For other users reading this: Alfred has a very simple to use automatic Dependencies Resolution system for workflows installed via the Gallery. Using this method is tried and tested so safe to use, and we are here to help any users who are unable to get workflows up and running:

 

https://www.alfredapp.com/help/kb/dependencies/

 

This workflow specifically requires Python 3, which is installed automatically by Apple / macOS when first requested, so again, very simple and safe.

 

Cheers,

Andrew

[vitor]

Welcome @Ash35,

 

You don’t need to run pip to install any packages, since they are included in the workflow. The problem doesn’t seem to be with Python or any part of your installation since those are reacting normally. It’s just that you’re not getting an expected reply. The issue seems to be somewhere in the connection (e.g. an incorrect API key).

Open the debugger, perform the failing action, then copy and paste the output here. That should give more clues as to what’s happening.

[albertkinng]

22 hours ago, Andrew said:

This workflow specifically requires Python 3, which is installed automatically by Apple / macOS

If I’m not mistaken, macOS has Python 2 installed by default. 
Here’s how to upgrade to 3. Not easy for a not technical user. 
 

[link removed by mod as not relevant]

[Andrew]

@albertkinng I've removed your link as I wouldn't recommend people installing python manually.

 

Python 2 was removed from macOS 12.3 Monterey. Python 3 is now automatically installed by macOS when first accessing /usr/bin/python3. macOS does this by prompting to install Apple's developer tools, which is a fully automated process.

 

https://www.macrumors.com/2022/01/28/apple-removing-python-2-in-macos-12-3/

 

Alfred's Dependency Resolution also automatically installs Apple's developer tools as part of the Homebrew install, so you get Python 3 automatically without having to do anything manually.

[albertkinng]

On 6/11/2023 at 8:49 AM, Andrew said:

macOS does this by prompting to install Apple's developer tools

Thank you for taking down the link. The OP lacks technical knowledge and may not comprehend terms like Dependency Resolution, Homebrew, and manual installation of Python 3. If the OP can install Python 3 through Alfred, that's fine. But if they need to take additional steps to install it, they might harm their computer. It's important to pay attention when someone says they are not technically inclined. Initially, I thought I was assisting by providing installation guidance, but I appreciate you taking action.

[EricCartman]

It took about an hour for me to finally get this workflow to work as a non-technical user.

 

Just wanted to share some key pointers so others in a similar position can get this up and running asap:

1. Even as an existing paid ChatGPT WebUI user, after generating your API key, you need to set up payment to use the API needed for this workflow.
   • I had free credits which had already expired.
   • You can do so here: https://platform.openai.com/account/billing/overview
2. I used Homebrew to install the latest Python 3.11 and its dependencies.
   • This was not easy to figure out, and I followed this YouTube video: https://www.youtube.com/watch?v=6sJ0qKTyJnc
   • I suspect this Homebrew installation is not needed as some users above have mentioned it comes preinstalled with MacOS (or something like that).
   • This was not necessary (as mentioned by vitor in the next post).
3. Go to ~/Library/Application Support/Alfred/Workflow Data/ and open the file ChatFred_Error.log to self-solve problems easily.
   • There are other ways to troubleshoot as written by OP here: https://github.com/chrislemke/ChatFred/#troubleshooting-️ but I found reading this log file to be the most helpful.
4. The last hurdle I encountered was being presented with the error message The model: `gpt-4` does not exist.
   • After some research, I've found that it is required to enter a waitlist for access to the gpt-4 API: https://openai.com/waitlist/gpt-4-api
   • I will just use ChatGPT 3.5 in this workflow for the time being.

Edited June 13, 2023 by EricCartman

[vitor]

9 minutes ago, EricCartman said:

I used Homebrew to install the latest Python 3.11 and its dependencies. (…) I suspect this Homebrew installation is not needed

 

Indeed, that is not necessary. It is important to note that if it were necessary, you wouldn’t have to follow that procedure anyway because Alfred’s dependencies manager does it. That is true of every workflow in the Gallery. Alfred can set up Homebrew for you and install the necessary packages without you having to muck about with the Terminal.

For this workflow that is not necessary at all, which is why the Gallery page makes no mention of it. Compare to (for example) Calculate Anything, which tells you at the top it requires PHP and how Alfred can get that for you.

[albertkinng]

2 hours ago, EricCartman said:

Just wanted to share some key pointers so others in a similar position can get this up and running asap

Hopefully @Andrew won't delete your post. Evidently it is not easy for a non-tech user.

[EricCartman]

2 hours ago, vitor said:

 

Indeed, that is not necessary. It is important to note that if it were necessary, you wouldn’t have to follow that procedure anyway because Alfred’s dependencies manager does it. That is true of every workflow in the Gallery. Alfred can set up Homebrew for you and install the necessary packages without you having to muck about with the Terminal.

For this workflow that is not necessary at all, which is why the Gallery page makes no mention of it. Compare to (for example) Calculate Anything, which tells you at the top it requires PHP and how Alfred can get that for you.

 

Thanks for your input.

 

If that is so, the "Install Python" section (from the troubleshooting Readme from OP) needs to be updated, because running python --version yielded no results for me.

 

Actually even now (after my Homebrew Python installation) it doesn't yield any results, so this part of the Readme is definitely a bit wonky.

 

Edited June 13, 2023 by EricCartman

[vitor]

13 minutes ago, EricCartman said:

Actually even now it doesn't yield any results with the latest Python version installed via Homebrew, so this part of the Readme is definitely a bit wonky.

 

Yes, it’s because python is Python 2 while the workflow uses python3. If you try python3 --version, you should see a result.

I have submitted a pull request to delete that part of the troubleshooting.

[chrislemke]

Hey! 

Sorry for being "offline" for so long. 

Troubleshooting is always a very important part of software and I would like to improve it. Thanks to @vitor for already improving the misleading readme. I want to improve it even more and would like to receive feedback on your experience concerning troubleshooting this workflow. I already read some interesting notes.

 

Thanks in advance. 

[chrislemke]

On 5/9/2023 at 8:31 PM, energyd said:

Hey @chrislemke

good stuff! I recently started using ChatFred with ChatGpt-4. One thing I really miss is the interactive experience I gained through chat.openai.com. I often found myself having to going back and forth with chatGpt and work towards my final solution. ChatGpt was asked to ask clarifying question to me first before working on a solution. 

 

With ChatFred, seems I can only send standalone query to openAI and then it would lost previous context when I ask a follow up question. Do you know a way to make it work interactively (e.g. remembering the context of the conversation)? 

Hey @energyd!
Sorry for the delay. With "ChatGPT history length" in the workflows configuration you can tell ChatFred how long it should remember. Is this any helpful?

[Zivkitaro]

Hi. I got a new M2 Macbook pro, and when I try to install ChatFred I get a message that it can't install python3 from the server. I restarted Alfred and tried again, but now it's not asking to install it, and also just won't work. Is there a way to remove ChatFred and try it again? (I tried to delete it, didn't help).

 

The cf command will not get me to Talk to ChatGPT, actually, there's no way to summon ChatFred it will blink and vanish from the list,

Edited August 10, 2023 by Zivkitaro

[vitor]

Welcome @Zivkitaro,

 

38 minutes ago, Zivkitaro said:

I get a message that it can't install python3 from the server.

 

Where are you getting the message?

[Zivkitaro]

24 minutes ago, vitor said:

Welcome @Zivkitaro,

 

 

Where are you getting the message?

Hi Vitor.

I'll try to breakdown the flow:
1. Installed the workflow and added my API key.

2. summoned up Alfred and tapped CF

3. the system asked to install python3

4. I approved it, and it got stuck and showed a message that it couldn't install it

 

At that point, I tried to install Python3 from the Pythin website, but now when I use Alfred and type cf the option to talk to CharGPT appears and blinks away. I remain with two options: Use a InstructGPT model or Using DALL-E 2.