Wanted: ideas for new starter packs!

Hey Rasa @community

This year we released two starter packs to help you build bots faster: the financial-demo and the helpdesk-assistant. Since then we’ve expanded their skillset a lot and seen lots of people do cool things with them!

Not all of you are building bots in the financial domain or the IT helpdesk domain, so we want to hear from you which starter packs you’d like to see next. If you have an idea or request, please post them in this thread :slight_smile: We’re looking for domain ideas, (e.g. retail customer service), and also specific skills you’d like to see (e.g. returning a product).

If you see an idea in this thread that would benefit you too, be sure to give it a like so we can see which ideas have the most community support!


Appointment booking bot

1 Like

FAQ bot. Converting FAQs based data in rasa training data (NLU files)


Product selection bot that advises which product to choose based on several attributes. That works for retail, travel, services, finance, etc.


A knowledge based bot that can have a collection of links of a specific website and can answer any questions when asked from that!


A restaurant order bot that can book an order with multiple items, change that order get a confirmation from the user.

Everything is fine, these two starters packs are enough to build any domain specific bots, actually I am building bot for healthcare which can answer and who can do a lot of custom actions. My request is, please design a simple responsive not UI for integrating not to the website. Somehow I can customize mr.bot’s js webcast plugin, it is not responsive in mobile screen…

1 Like

With examples the domain isn’t as important as the demonstrated functionality. Is it possible to create a bot that can structure queries and then interrogate the semantic web for general knowledge? Or demonstrate the steps involved in building a speech-2-text/text-2-speech front end? Or best practice for dealing with slang and colloquialisms? (unusual users) Or identifying and tracking specific users so that distinct customised user responses can be built over time? Great product. Thanks.


Student assistant bot at an university for attendance and class schedule details , notifications/alert . And admission query assistance for concerned parents and students


product customer service

A bot that can answer precise factual questions using a complex knowledge graph, perhaps using Neo4j.


Sounds interesting! What kind of design, architecture do you imagine, I mean what will the difference from answering questions using other ways?

As I started a prototype of this: (1) intent recognition selects (2) custom action which (3) queries Neo4j and (4) fills in slots in a sentence template using the retrieved data which is sent back to the user as (5) the response. Querying a relational database would work similarly. I assume a broader-domain information retrieval based QA architecture might be more scalable in searching and selecting a text snippet in a text corpus that answers the question. This is probably more scalable, currently, and seems more popular, but I personally have a hard time with how fuzzy and nondeterministic this approach feels. I prefer precision over recall and hope a KG-based approach can be scaled up by finding (1) efficient ways of building knowledge graphs semi-automatically (human-in-the-loop ML), and (2) applying and extending some recent research in translating natural language into code (in this case, translating a question into a cypher query).


very interesting and promising approach! The idea I am working on would require building a knowledge base about each individual user and tailoring the intent recognition and responses to each user based on that data, including custom actions. I like your idea of translating NL into code btw!

1 Like

I am looking to build a system that learns consistently from each individual who interacts with it, and keeps those learnings separate and adjusts its intent recognition and custom actions accordingly.

As far as customer service, q&a, restaurant order bots etc. – as @tswordsmith, the domain itself isn’t as important as the Rasa skills/functionality to work in a domain, plus, there are soo many examples out there of simple single-purpose basic bots in those domains. Unless of course the end-users want a “git checkout and start running an off-the-shelf basic bot”.

1 Like

Here are some ideas that I think would be useful:

  • FAQ chatbot for events, businesses etc.
  • Customer feedback collection chatbot
  • Ticket booking bot
  • Personal assistant/sales agent bot

A chatbot for e-commerce businesss. Most of the functions of the chatbot will be the same, like sharing product details, order details and tracking details. I am currently working on the same for my website. But it can be applied to any set of products, most of it is general.

Any bot with a speech feature. A detailed blog about integrating a speech system like azure speech services or google one with any existing bot which can be called a voice bot.

Also interested in healthcare bots. Used ibm watson before but they are too expensive in the long run.

any thoughts on what kind of features a healthcare bot should have? We do have an example bot, medicare locator (admittedly a little outdated), would be interested to hear whether it’s those kinds of skills you’re looking for or something else.

1 Like