Cisco MindMeld vs Rasa

Sorry if this is a sensitive topic, but Rasa will have to address this topic sooner or later and I am personally very interested to know what the Rasa community’s response will be to Cisco’s having open-sourced MindMeld. There were no serious competitors to Rasa in my mind before this happened. But now we have MindMeld which is also python-based, open-sourced, and has 5 years more development time under it’s belt compared to Rasa and the support of a big corporation. It does have less buzz online in terms of YouTube channel followers, etc. But, I like some of the extra features and maturity in terms of NLU and built-in knowledge base. On the other hand, I have been integrating Neo4j into my Rasa bot’s custom actions and am really loving the combination. I don’t think that MindMeld’s ElasticSearch KB will be as good as a graph database for answering precise queries.

Anyway, I am interested to hear what the Rasa community says. For now, I will be learning both in parallel, in order to come to my own conclusions.

Why do you assume that its not possible to integrate NEo4j in MindMeld ? MindMeld have a philosophy against GUI, what do you think about this ?

I’m just starting to learn about MindMeld. I expect that they probably would allow us to integrate Neo4j. I was just talking about their own default KB. I’m not sure what this philosophy against GUI is. Can you say more?

Hi @tomp - not a sensitive topic at all!

I’m very glad to see more open source tools in this space. Rasa wasn’t the first open source library for building conversational AI by any means. We also don’t get up in the morning motivated because Rasa is the only option developers have. We’re here to build the best possible tool for developers to build AI assistants.

Conversational AI is far from a solved problem, which is why we put so much emphasis on applied research and bringing those results into the code base. But not all good features have to be clever. A friend of mine who works in dialogue research at a big company told me the share your bot feature in Rasa X is a huge help to them. We’re also deeply committed to the community (and I hope that shows!) because learning from each other is just as important as functionality.

All this to say that it’s a good thing that other libraries exist, it means more people are thinking seriously about how to tackle these problems and trying different approaches.

By the way, your Neo4j integration in custom actions sounds super cool. Is it similar to this approach ?

1 Like

Thanks, @amn41. Very good.

I expect that there will be some good cross-fertilization between Rasa and MindMeld in the future, considering their close place in the CAI ecosystem.

I have not seen that KB tutorial before. There are some similarities with my nascent approach and I think there are a few things in it that I can learn from. Thanks for pointing it out.