Jump to content

ChatGPT / DALL-E - OpenAI integrations


Recommended Posts

Posted (edited)

I've made a workflow that make an External Call to com.alfredapp.vitor.openai trigger new_chat

I use it to "Summarize this:" etc with universal actions. 

I'd like to make it so that holding a modifier key overrides the engine to be gpt-4o. So i've be a var and args node after a hotkey modifier 

 

Will this work?

 

Screenshot args and vars

 Screenshot

Edited by Textdriven
Link to comment

The Alfred workflow is how I use ChatGPT. The Universal Action would be a natural place for this feature, I think! (Dunno how difficult it would be to drop images into the new input window.)

 

Cheers!

Harri (user since pretty much the beginning and Legendary Supporter :-)

Link to comment

Hi, friends. Thank you for your great work. I want to have several commands \ profiles to communicate with ChatGPT, for example:

1. Summarize text
2. Insert into the start of each line emoji
3. etc

 

Most easier may look like:
1. duplicate workflow throw button "Duplicate"

2. set API keys and other parameters

3. use it

 

Unfornetly, I have followed the error. I have experience in Python but am a full noob in Alfred. Can you please help me

[00:04:43.993] ChatGPT / DALL-E - COPY[Keyword] Processing complete
[00:04:44.010] ChatGPT / DALL-E - COPY[Keyword] Passing output '' to Arg and Vars
[00:04:44.012] ChatGPT / DALL-E - COPY[Arg and Vars] Processing complete
[00:04:44.014] ChatGPT / DALL-E - COPY[Arg and Vars] Passing output '' to Run Script
[00:04:44.015] ChatGPT / DALL-E - COPY[Run Script] Processing complete
[00:04:44.016] ChatGPT / DALL-E - COPY[Run Script] Passing output '' to Automation Task
[00:04:44.017] ChatGPT / DALL-E - COPY[Automation Task] Running task 'Does Path Exist?' with arguments (
    "{const:alfred_workflow_data}/chat.json"
)
[00:04:44.023] ChatGPT / DALL-E - COPY[Automation Task] Processing complete
[00:04:44.024] ChatGPT / DALL-E - COPY[Automation Task] Passing output 'false' to Conditional
[00:04:44.025] ChatGPT / DALL-E - COPY[Conditional] Processing complete
[00:04:44.026] ChatGPT / DALL-E - COPY[Conditional] Passing output 'false' to Write Text File
[00:04:44.026] ChatGPT / DALL-E - COPY[Write Text File] Processing complete
[00:04:44.027] ChatGPT / DALL-E - COPY[Write Text File] Passing output '/Users/redacted/Library/Application Support/Alfred/Workflow Data/{const:alfred_workflow_data}/chat.json' to Arg and Vars
[00:04:44.028] ChatGPT / DALL-E - COPY[Arg and Vars] Processing complete
[00:04:44.029] ChatGPT / DALL-E - COPY[Arg and Vars] Passing output '{const:alfred_workflow_data}
{const:alfred_workflow_cache}' to Split Arg
[00:04:44.030] ChatGPT / DALL-E - COPY[Split Arg] Processing complete
[00:04:44.030] ChatGPT / DALL-E - COPY[Split Arg] Passing output '(
    "{const:alfred_workflow_data}",
    "{const:alfred_workflow_cache}"
)' to Automation Task
[00:04:44.031] ChatGPT / DALL-E - COPY[Automation Task] Running task 'New Folder' with arguments (
    "{const:alfred_workflow_data}",
    "{const:alfred_workflow_cache}"
)
[00:04:44.057] ERROR: ChatGPT / DALL-E - COPY[Automation Task] Task Error (1): /Users/redacted/Library/Caches/com.runningwithcrayons.Alfred/Workflow Scripts/Automation Tasks/60cfcfc99f6a2b93568c621cac71599f9979bdc22241aeea13c51d1ae926e2ff: execution error: Error: Invalid path: {const:alfred_workflow_data} (-2700)

I checked for answers on ChatpGPT and Google. I probably do not understand the logic of the previous step in the workflow.


Debug info:

 

  • The Workflow: ChatGPT / DALL-E 
  • Alfred: Alfred 5.5 [2257]
  • macOS: 14.4.1 (23E224)

 

Link to comment

Hi, I installed the workflow and configured it (including the api key), But when I type the ChatGPT keyword and click Enter or Command Enter nothing happens. It does work when i type the dalle keyword. Did I miss something in the configuration?

Link to comment

@GabrielBhh Could you please fill in your Powerpack email address in your forum profile first?

 

Next, please take a look at the "How do I report an issue?" point in this post, and provide the necessary details:

 

Link to comment
Posted (edited)

I had been meaning to see whether there were any workflows around for local LLM integration with Alfred, then I realised that I'd installed this one - and Ollama exposes an OpenAI-compatible API, as do many other tools. Pleased to say it works perfectly, and there are no usage costs.

 

image.thumb.png.140aaa8c3598f68b948c2945a3fabeaa.png

 

If I were to extend this, my next step would be making model selection easier - parsing the results of either `GET /api/tags` or `curl http://localhost:11434/api/tags` to populate the `dalle_model` and `gpt_model` popup selections (filtering out `bert` embedding models, filtering `clip` models to dalle`, etc.).

Haven't fully read the thread to see how many people have asked or resolved the question of how to prevent issues when hiding the window early, but I imagine it's >1 and I thought this was interesting enough to share at this stage. :)

Edited by PKMfan420
removing weird localhost link
Link to comment
  • 2 weeks later...

Wonderful workflow!

 

I would like to ask you whether it's possible to copy/paste items from Clipboard History, since as soon as I hit the hotkey for the Clipboard History, ChatGPT window is closed and the text is gone next time I open it

Link to comment
11 minutes ago, charlesether said:

When will you update this so we can make use of GPT-4o? 

 

Is it possible to modify the workflow on our own?

 

On 5/13/2024 at 10:11 PM, vitor said:

Updated to 2024.11.


Add GPT-4o model.

 

On 5/14/2024 at 1:39 PM, vitor said:

The screenshot doesn’t provide sufficient information to say. You need to override the gpt_model variable to be gpt-4o and it should work.

 

🤔

Link to comment

No output for chatgpt

 

Hi Vitor, thanks for the great work.

 

I encounter an issue.
When typing 'chatgpt' and I enter the query and hit enter, there is no output.

Debug of my error:

Quote

[12:52:05.735] Logging Started...
[12:52:07.977] ChatGPT / DALL-E[Keyword] Processing complete
[12:52:07.985] ChatGPT / DALL-E[Keyword] Passing output '' to Arg and Vars
[12:52:07.986] ChatGPT / DALL-E[Arg and Vars] Processing complete
[12:52:07.987] ChatGPT / DALL-E[Arg and Vars] Passing output '' to Run Script
[12:52:07.988] ChatGPT / DALL-E[Run Script] Processing complete
[12:52:07.989] ChatGPT / DALL-E[Run Script] Passing output '' to Automation Task
[12:52:07.989] ERROR: ChatGPT / DALL-E[Automation Task] Task not found 'com.alfredapp.automation.core/files-and-folders/path.exists'

 

 

 

With DALLE-E it works fine (see debug of DALL-E here:

Quote

[12:58:42.185] ChatGPT / DALL-E[Keyword] Processing complete
[12:58:42.192] ChatGPT / DALL-E[Keyword] Passing output 'river in z' to Text View
[12:58:42.229] ChatGPT / DALL-E[Text View] Running with argument 'river in z'
[12:58:42.342] ChatGPT / DALL-E[Text View] Script with argv 'river in z' finished
[12:58:42.345] ChatGPT / DALL-E[Text View] {"rerun":0.1,"variables":{"loaded_previous":true},"response":"**Original Prompt:** clown\n![](/Users/.../DALL-E/2024.06.19.12.46.21-98DA4A41.png)","behaviour":{"scroll":"end"}}
[12:58:42.448] ChatGPT / DALL-E[Text View] Running with argument 'river in z'

 

Again, thank you for the work. I'm looking very much forward to use this.

Best

Corina

Link to comment

@Corina Looks like you might need to update your Automation Tasks if that task is missing.

 

To do this, open Alfred's preferences to the Workflows tab, click "Alfred Gallery" at the bottom of the sidebar > Automation Task Options...

 

Once you're in, update the Tasks, and check the box to automatically keep automation tasks up to date, so that you don't need to think about it in the future when new tasks are added.

 

image.png

 

Let me know how you get on :)

 

Cheers,
Vero

Link to comment

Hi all!

How can I fix the scroll so that when adding an answer, it does not move.
I tried the scrolling behavior but the content is still jerky
This causes a lot of inconvenience when reading the answer from top to bottom

Link to comment
On 5/26/2024 at 12:33 PM, GabrielBhh said:

Hi, I installed the workflow and configured it (including the api key), But when I type the ChatGPT keyword and click Enter or Command Enter nothing happens. It does work when i type the dalle keyword. Did I miss something in the configuration?

 

I have the same issue. Did you find a way to solve this?

Link to comment
11 hours ago, intstdu said:

I have the same issue. Did you find a way to solve this?

 

@intstdu Could you check your Automation Tasks are up to date, in the same way I suggested here?

 

 

Link to comment
18 hours ago, lone said:

Hi all!

How can I fix the scroll so that when adding an answer, it does not move.
I tried the scrolling behavior but the content is still jerky
This causes a lot of inconvenience when reading the answer from top to bottom

I made a test example where I try to scroll, and it constantly rolls down
 

ezgif-1-7e52e5a1fe.gif

Link to comment

Thanks, this is great.

I've just installed it and tried it out a few times.

However, I get completely different results when I use Alfred with the ChatGPT workflow compared to going to chatgpt.com and entering the same question. For example, when I ask it to solve a certain Swift exercise, Alfred chatgpt tells me it can't do that because it doesn't know the exact contents of the exercise, while the website-chatgpt does it without problems. Also when I ask it to make a poem, Alfred does make one but with some nonsensical words and wrong punctation and capitalisation. When I enter the same question on the website, it delivers a poem that is much better. What could be the reason for this? both are on chatGPT 3.5

 

Link to comment
Just now, rees said:

What could be the reason for this?

 

That’s just how LLMs work. They are not deterministic at all, they essentially give you answers based on statistical information of the next word. You could ask them the same question twice in a row and get wildly different results. Which is one reason you should never trust them for any important or critical information. They don’t “know” anything, it’s just that the text is usually so well formed that they provide that illusion.

Link to comment
8 minutes ago, rees said:

What could be the reason for this?

Could it also be different temperature setting? Different top-p setting? Different system prompt? Plus what @vitor just said. 

 

Also, ChatGPT on the website is constantly adapting and could be a different model from what they are providing via the API with the same name. 

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...