ERPNext + N8N + ChatGPT + Whatsapp

Integration
, , ,
6

Just as a continuation, for update the progress

[DONE] User language and location detection basedly on phonenumbers
[DONE] Automatic Creation of Leads
[DONE] OpenAI Context based responses basedly on the website content
[DONE] Recovering of the Lead Chat History
[DONE] Integration with Frappe Chat
[DONE] Automatic creation of Oportunities, when the lead indicates that wanna talk to a seller
[PARTIAL] Triggering of the Exotel Pop-Up when the Lead messages the team
[PARTIAL] Creation of Chat Context Rooms (Sales, Support, Finances)
[PENDING] Information Gathering from Lead on Sales Process (Address, Documents, etc)
[PENDING] Quote approval via Whatsapp
[PENDING] Payments Notifications
[PENDING] Order Tracking
[PENDING] Ticket Opening and Tracking
[PENDING] Integration with Frappe Desk
[PENDING] Integration with Raven

That week, we had to refactor the integration with ChatGPT, to reduce costs, to add actions, and improve the context classification.

We are building a business profile, from where the chatbot can reason and think on top of that.

Next step, is work the frappe app

3 Likes
7

Guys! If someone wanna talk with the chatbot to know it’s capabilities, you can try on this phone number +55 19 953-333-525

May some responses will be inconsistents, but we are monitoring all this to ensure everything will work perfectly

1 Like
8

I would be interested to look at and give feedback

1 Like
9

I’ve a n8n deployment at work. While is fun to make things, maintenance is hell :confused:
I’m moving away from it.

1 Like
10

@mattesilver we are using many solutions in dockerized environment on top of Coolify and Portainer!

Regarding maintenance, we have zero headache!

Most of the services are realiable, and regarding upgrades, we automated all that stuff, so, no chances of downtime!

All the automation is done via N8N, so, I have literally the oposite perception you have!

2 Likes
11

Members appreciate if you can write some guidelines to use n8n with erpnext to connect with other apps or social networks etc., I am interested to learn n8n, done installation too and will proceed further.

12

How do you move changes between development, staging and production environments?

13

@mattesilver Via Github Actions!

Development is always done in Docker locally on the developer machine

Staging is made via Github Actions, when A pull request is started an N8N automation takes places, and install the automation in the staging N8N service via N8N API.

After the tester finalize his tests, his aprove the PR, and when it’s merged into production, the same process happens again, just changing the staging server to production server.

  • Keep the credentials with same name in diferent servers help a lot on that process[
  • If you have good coding skils, you can create environment variables to modify the credentials while moving from one server to another.
14

On the version I’m at I can’t even reliably export&import flows between environments and I’ve stopped upgrading when it broke my flows,without a way to fix it. I’m sure they’ve improved the API since, but I just left it alone, and moved on to python.
I might give it another look on a fresh install one day.

Anyway, I’m happy for you that your solution works for you.

1 Like
15

This is interesting. I want to explore.

16

Hi, it looks interesting to try it out.

One question, is it necessary to have Docker?

17

What kind of example would be helpful for n8n and ERPNext integration?

18

Thanks, I can share examples, I will get back to you after some research.

19

Hi !

Sorry, going back to ERPNext + ChatGPT we are doing more reasearch around that subject!

The results we got from trials are really optmistic, but, it fails in terms of context.

We had to dig more to compreend ChatGPT and other initiatives around that.

Right now, we found a properly approach, but that require build some layers of context

1 - ChatGPT need to know properly the tools available and when to use
2 - ChatGPT need to know properly the business that the ChatBot will help
3 - ChatGPT need to know properly the user that him will help

Due that, we are writing a layer of supervised learning on top of ChatGPT, so we can reuse contexts.

I think everyone that have deeply tried ChatGPT, will understand the issue we are trying to face.
After some time, the flow of the conversation get lost, and ChatGPT or start being redundant or lost the context.

This is the major issue we have with the ChatBot right now.

3 Likes
20

I think @rmeyer also have some experience in this topic. You may have something in common.

21

Hi @max_morais_dmm

It looks like a great integration. Can you please help me with one thing? How do you fill in the Child Tables from n8n? Im having a problem when creating a contact.

1 Like
22

@Caner_Alp, some times, we just do a simple HTTP Request.

When the document is complex, like the contact, that’s the most viable solution!

So, dont use the ERPNext node, use the HTTP Request Node, and pass your data as JSON on the body

23

Is that frappe app is published ? If yes please share the repo link please

24

Great concept. Are you intending to release this as a paid-add-on app to ERPnext?

There are multiple use-cases both internal and external that would benefit from this.

1 Like
25

i’m interested