šŸš§ Ask AI updates: HTTP Node helper (beta), and self-hosted!

Announcements
,
1

Update: Ask AI in the code node was incorrectly announced and is not available in the code node, We have also now removed the beta for Ask AI in the HTTP Request node.

We have a couple of exciting updates on our ā€˜Ask AIā€™ features for you today! These are all still under development and can give unexpected results, so please keep that in mind before putting the results in production. Weā€™re sharing this with you now as weā€™re hoping to get some ā€˜real-worldā€™ feedback from our community.

Make sure you upgrade to 1.40.0 to use these features.

What is Ask AI?

With Ask AI you can get customized, user-friendly help when you need it. It is available in the following contexts:

  • The Code node uses ChatGPT to generate code, based on your prompt.
  • The HTTP node will help you generate API requests - see below.

New: Ask AI is now supported on self-hosted instances! :tada:

Many of you asked, and weā€™re happy to report that you can now use all the ā€˜Ask AIā€™ features on your self-hosted instances - including on the free Community edition! This covers both Ask AI for the Code Node and the HTTP node.

To start using it, just configure your own AI credentials - for now, only OpenAI is supported.

:construction: New: Ask AI for the HTTP node (beta)

The HTTP node is great for extending n8nā€™s functionality, but you will need to study the appropriate API documentation to use it. Our new Ask AI feature is making this a lot easier by providing detailed guidance on over 600 services.

To use:

  1. Select Ask AI.
  2. Enter the Service and Request you want to use. For example, to use the NASA API to get their picture of the day, enter NASA in Service and get picture of the day in Request.
  3. Check the parameters: the AI tries to fill them out, but you may still need to adjust or correct the configuration. The response will contain placeholders for values that need to be replaced like {YOUR_API_KEY}

Some caveats of our current implementation:

  • If the service is not defined in our knowledgebase, the response will be a hallucination from the AI (the generated code will be unreliable and not based on a real API Definition). We can, however, add services to our knowledgebase on demand. The list of known services can be found here.
  • There are some exceptional cases in which the AI will refuse to generate a working curl command, which we donā€™t have control over.

Please help us test this feature and report any issues or thoughts as a reply on this topic. Thanks!

Ask AI

19 Likes
4

Super powerful! :star2:

Just did a little fun-run with a new API (not in the knowledge base) and it did a pretty amazing job - only took me 5 minutes!

This is the API I used (to remove background from pictures)

And my prompt:

image
image698Ɨ308 16.5 KB

Then all I had to do was enter my API key (blue) and voila!

image
image1855Ɨ829 109 KB

Left the picture input and right the picture output

:+1:

5 Likes
5

Thatā€™s awesome! Thanks for this n8n team!!

Where can we ask for additional services in the knowledge base?
Iā€™d like to include MailerLite, Kiwify, ManyChat and Hotmart (in this order or priority).

6

Did you do anything to add remove.bg to itā€™s knowledge base or just asked for it?

7

have you seen/heard of gorilla? (berkeley project) theyā€™ve got a pretty extensive collection of api adapters for their modelā€¦ might be a useful source of inspiration for api integrations?

1 Like
8

Not sure what Iā€™m doing wrong but I donā€™t see the ā€œAsk AIā€ button on the HTTP module. The Code node has the Ask AI feature, but the HTTP node does not. I set up the environment variables, recreated the container, confirmed Iā€™m on 1.40.0 and confirmed the environment variables are present.

9

Wow, that Gorilla project is huge!

10

I wanted to specifically try an API that was not in the knowledge base to see how it behaves. So I just used remove.bg in my prompt :slight_smile:

1 Like
11

That is cool! Soon weā€™ll have AI building our workflows without us.

1 Like
12

I am having the exact same issue. Tried with a ā€œProject API keyā€ from OpenAI, as well as the legacy version. Unfortunately the Ask AI feature on the Code node returns: ā€œCode generation failed due to an unknown reason. Try again in a few minutes.ā€

My config:

      - N8N_AI_ENABLED=true
      - N8N_AI_PROVIDER=openai
      - N8N_AI_OPENAI_KEY="${AI_OPENAI_KEY}"
      - N8N_AI_OPENAI_MODEL=gpt-4-turbo
      - N8N_AI_PINECONE_API_KEY="${AI_PINECONE_API_KEY}"
13

Can you try switching to GPT-3.5?

14

That is what I just did before (gpt-3.5-turbo). Same error.
Which API key is supported? The new or the legacy one?

15

Iā€™ll ask around - please hold

1 Like
17

I have aswell set it up on my dev/home server. I do not even see the Ask AI field, even though configuration seems pretty straightforward.

18

@bartv Does it suport Gemini api?

19

Hey folks,

Following your feedback here and in other places we have some updates for you:

  • Ask AI (for the HTTP node) for Cloud users uses a vector store that we designed to give accurate results on over 600 APIs.
  • Self-hosted users should not set N8N_AI_PINECONE_API_KEY since they donā€™t have our data.
  • Therefore, self-hosted users will use generic GPT-4 results and will get less accurate results.
  • Weā€™re discussing how to give self-hosted users a better, non-generic GPT solution.
  • We feel the Ask AI for errors is not good enough at this time and will remove it for now.

No; all our ā€˜Ask AIā€™ features use OpenAI for now.

20

Hi,

weird. I use n8n locally and tried to activate the Ask AI feature. Itā€™s running locally with pm2.

I followed the instructions. but still canā€™t access the feature/ it doesnā€™t appear in the http request node.

21

Which version are you on? And did you enter any value for N8N_AI_PINECONE_API_KEY ?

22

Hi,

1.41. is the version.

Tried with pinecone and without.

  • Self-hosted users should not set N8N_AI_PINECONE_API_KEY since they donā€™t have our data.

Locally installed with npm on a MacBook Pro