Feature suggestion: Shortcuts action

[avvocati-e-mac]

It would be useful some minimal action to intrigate with Shortcuts via the app.

[AugustDev]

Can you share some example flows that you would like to use?

[avvocati-e-mac]

Sure! A minimum of context: I am an Italian lawyer passionate about computer science and Apple tools, I am not a programmer. I've already created some shortcuts to interact with Ollama's APIs (see the image below). This allows me, for example, to send the PDF text or a web page to the LLM and get a response that then, possibly, can be further passed to another shortcuts. If Enchanted LLM could donate some similar action to Shortcuts I would avoid the whole problem of managing APIs in Shortcuts and expand what the app can do.  If I understood correctly, it is only possible to use it for chat but not to work on documents or other sources outside the LLM. Thank you for your availability, love your work! Filippo PS the image shows Shortcuts in Italian but Mistral summarized and answered in English and summarize my notes to use Ollama and Mistral via terminal. On 3 marzo 2024 a 02:17:15, Augustinas Malinauskas ***@***.***) scritto: Can you share some example flows that you would like to use? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.

[AugustDev]

Hi @avvocati-e-mac thank you so much for sharing this interesting workflow. It seems the images are missing from the message, it would be great if you can share it.

I think I get the rough idea. Few questions

• You would like an Action which accepts text or image, processes it via Enchanted(Ollama) and returns the response back?
• What exactly do you send to Ollama right now - text and image?
• How do you currently process PDF?

Best

[avvocati-e-mac]

Sorry @AugustDev , I replied via email 😅

Here is the image (I hope… 🤞🏻)

For your question:

• Yes, I would like an Action which accepts text or image, processes it via Enchanted(Ollama) and returns the response back;
• I tested only text;
• I use the builtin action of Shortcuts to extract the text of the PDF and then pass it to the API, I didn't chunck the text;

I rapidly check and via API it's possible to send image in coded in base64 that also Shortcuts can handle link here. I'll try it as soon as I can.

Al the best and thanks for the response!

[AugustDev]

Thank you for the detail explanation @avvocati-e-mac. It should be fairly easy to support and seems like an interesting idea.

Enchanted already supports images for models that allow it (llama and baklava currently).

Let me give it a go next week.