Parse doesn't use latest trained model

Rasa NLU version:

0.15.0a1

Operating system (windows, osx, …):

Ubuntu 18.04.1 LTS

Content of model configuration file:

language: "en"

pipeline:
- name: "tokenizer_whitespace"
- name: "ner_crf"
- name: "ner_synonyms"
- name: "intent_featurizer_count_vectors"
- name: "intent_classifier_tensorflow_embedding"

data: {
  "rasa_nlu_data": {
    "common_examples": [
      {
        "text": "hey",
        "intent": "greet",
        "entities": []
      }
    ]
  }
}

Issue:

In the HTTP API docs (HTTP API) it says „By default the latest trained model for the project will be loaded“. However this doesn’t work for me.

My attempt to parse with curl -XPOST localhost:5000/parse -d '{"q":"test", "project": "test_1"}' returned:

{
    "intent": {
        "name": null,
        "confidence": 1
    },
    "entities": [],
    "text": "test",
    "project": "test_1",
    "model": "fallback" # This should be model_1
}

It works when I define “model” in the JSON body, however the way our application works I’d have to run an extra query (/status) to identify the latest model before every /parse.

I’ve trained the model before using curl -XPOST -H "Content-Type: application/x-yml" localhost:5000/train?project=project_1&model=model_1 (and model config above), it returned:

{
    "info": "new model trained",
    "model": "1"
}

The curl localhost:5000/status after training was:

{
    "max_training_processes": 1,
    "current_training_processes": 0,
    "available_projects": {
        "default": {
            "status": "ready",
            "error_message": null,
            "current_training_processes": 0,
            "available_models": [
                "fallback"
            ],
            "loaded_models": [
                "fallback"
            ]
        },
        "test_1": {
            "status": "ready",
            "error_message": null,
            "current_training_processes": 0,
            "available_models": [
                "fallback",
                "1"
            ],
            "loaded_models": [
                "fallback"
            ]
        }
    }
}