Managing projects and models

Back in version < 1.0, it was entirely possible, and trivial, to manage projects and models. The documentation was also clear about these things. I didn’t use Rasa a whole lot before 1.0, but it was fine.

Our use case is as follows: We have multiple isolated bots and we have our own dialog manager integrated with our customer service platform. We want to replace the NLU engine to Rasa, because we tested it and it looks better than the one we’re currently using.

Currently in our own NLU engine, each bot has its own models (staging and production). Each bot is also analogous to a rasa “project”. Each bot has a “bot id” (a project id) and we can train, test and parse a message. In rasa < 1.0, we had the very same operations readily available on Rasa NLU. I understand that nowadays in Rasa 1.0, the Rasa NLU server is not able to manage projects or models, it just runs a model.

I went through the HTTP API of Rasa X and initially it seemed that the project and model functionality was migrated to Rasa X (since the API often asks for a project id), but then I haven’t found a way to create multiple projects in the HTTP API.

It looks like I need 1 Rasa or Rasa X instance per project?

Then I’ve read Support of many projects using single agent with Rasa 1.0 and Removing projects for Rasa NLU server and it is becoming very clear that I indeed need multiple instances of Rasa. Calling PUT /model doesn’t solve the problem since we have multiple bots parsing messages concurrently. We would easily run into concurrency problems.

I just want to confirm if this is true. And if it is, my feedback is that it’s very frustrating.

1 Like

Yes, I can confirm this. Thanks for your feedback (also in the other thread). As I mentioned in the corresponding GitHub issue (Removing projects · Issue #3259 · RasaHQ/rasa · GitHub), we definitely consider implementing a similar logic in the future again. However, I cannot give you any timeline. We will let you know once we decide on something.