(Juste) #1

Share the projects you are working on and find collaborators.

We would love to hear what you are working on and what project ideas you have. This topic is also a perfect place to share the roadblocks you are facing and find the collaborators to help you bring you great ideas to life! :slightly_smiling_face:

(Juste) pinned #2

(Juste) unlisted #3

(Juste) listed #4

(Anders) #5

My experimental project (if succesful then for production) is a bot that helps employees in the public sector report verbal/physical abuse or accidents in their workplace. The current solution is a cumbersome digital form that has to be filled out. Rasa is an excellent starting point because it is open source and allows us to handle personal information secretly.

The language is Danish so I need ner_duckling for several entities.

I need to collect:

  • likert (1 - 5 rating)
  • free-text description of the incident
  • time+date (where duckling comes in. Don’t know how to handle dates yet)
  • incident type (3 categories)
  • place (possibly duckling again)

(Guilherme Bruno) #7


Currently, we are building a ChatBot which will be called NestorGR. It is a experimental try on the tech, since our team for the project (consisting currently of:

  • me as a Intern/back-end dev. I had never really made nothing “real” with python, so LOTS of stuff to learn. NICE! :smiley:
  • my mentor, who is the deploy/back-end guy (i started to take on the back-end tasks so he could focus on the kubernetes stuff for now)
  • a front-end dev for the user interface.

The idea as of now is to build a robust and scalable app based on microservices and implement it on our company website. We are a company of about 40ish devs and handle lots of projects for our city port (IIRC, brazil’s biggest), so there is lots of contact via our website/phone regarding status on the multiple projects.

NestorGR should handle most of the information about them, so we won’t need to allocate a person specifically to deal with this kind of conversation, and in last case guide the user to the proper sector dealing with it.

In future, we plan to integrate our company Facebook page to the AI, and with it stable enough, start to provide the service as an extra for our clients, using their APIs/whatsoever.

One of our main roadblocks as of now, and I would be very thankful if you guys have some suggestions are on how to:

  • Deal with him learning new stuff as users talk to the AI (I thought on saving every conversation on a postgres db and from times to times update our models - is it the right approach?)
  • Optimize our docker containers to proper good practices, since we are handling docker/docker-compose/kubernetes for the first time, so there are images of 1.xGb, which i guess that can be reduced a lot with alpine
  • Build a UI interface to improve workflow (we had some issues with rasa-ui, but are still using it for basics)
  • Apply good practices on Python codes. As it is my first time coding with python and coming from a JS-mainly background, there are some stuff i would like to improve.

If any of you guys have real-life examples with open-source code for me to study the workflow, file structure and such, I would be thankful. I’m loving it like never before, so I want to provide the best code i can offer! hahaha

cheers and keep on the good work, mates!

(Rohit haritash) #8


Our company have an app and web based e-commerce platform. So we are planing to build a bot which can help user to look for different product with the help of chat bot. For a use case, we are planing for user authentication through the bot. So my first question is , it it feasible or right use case for a chatbot. I was exploring different e-commerce chatbots for shopping. But all of them were based on contextual -UI . Is it practical to implement a shopping workflow from product search to checkout using RASA NLU and RASA core or any contextual AI framework.

(Kiran Chandar Jayaraman) #9

I am not able to pass values from cmdline to the custom action script written in Python. Any idea on what am I missing or how to resolve it ?

(Akshit) #10

Hi @jkc.kiran This is not a thread to ask these kind of questions. Create a new topic and add more details about your issue! Good luck!

(Kiran Chandar Jayaraman) #11

Thanks @Akshit

(Shail Choksi) #12

I am working with Lichess Team (Chess playing website) to create a chatbot that works with Chess bots. Its still experimental but I got a very crude version of it working with NLU + Core.

The idea is that each chess bot can have its own bot and answers accordingly. So creating a framework for chess bot makers to make their own chatbot using Rasa’s tech.

(Michele Mastrogiovanni) #13

Since I know Kubernetes and its packager Helm I’m working on a simple chart that will make possible to deploy on a cluster a simple chatbot given a github repository for nlu, stories and custom actions.

I already developed a chart that, when installed in the Kubernetes cluster, download stories, actions and nlu and use them to train a Rasa model.

In past I also developed simple Bot for Telegram using BotKit.

The idea would be to have a out-of-the-box Rasa chart to create and deploy self-service Bots.

I’m just wondering:

  • if this contribution can be a good idea and
  • to look if some is interested in the creation of the bot (nodeJS with BotKit and Rasa interaction)

Let me know if some is interested. Bye