# Training new model on the fly using http api

**URL:** <https://forum.rasa.com/t/training-new-model-on-the-fly-using-http-api/54893>\
**Category:** Feedback on Rasa Open Source\
**Tags:** rasa\
**Created:** [August 18, 2022, 7:40am UTC](https://forum.rasa.com/t/training-new-model-on-the-fly-using-http-api/54893 "2022-08-18T07:40:23Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![anjaly](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/anjaly/32/5213_2.png) [@anjaly](https://forum.rasa.com/u/anjaly)\
**Post date:** [August 18, 2022, 7:40am UTC](https://forum.rasa.com/t/training-new-model-on-the-fly-using-http-api/54893/1 "2022-08-18T07:40:23Z")

</div>

Hi,

I am trying to train a model when I add a new question from our UI. I’m doing this using the [model/train](http://localhost:5005/model/train) http api using the below data.

intents:

- track\_parcel
- leave\_balance nlu:
- intent: track\_parcel examples: ‘- Track parcel’
- intent: leave\_balance examples: |-
  - Leave balance
  - attendance

After creating the model in the default path /models, I called the [replace model](http://localhost:5005/model) http api for loading the current model. Then I tried [http://localhost:5005/model/parse](http://localhost:5005/model/parse) http api for getting the intent triggered using the below data

{

“text”: “jjjjjjj”,

“message\_id”: “b2831e73-1407-4ba0-a861-0f30a42a2a5a”

}

But I got the track\_parcel intent with 0.9 confidence

RESPONSE for model/parse api

{

```
"text": "jjjjjjj",

"intent": {

    "id": 99692160290177574,

    "name": "track_parcel",

    "confidence": 0.9580326676368713

},

"entities": [],

"intent_ranking": [

    {

        "id": 99692160290177574,

        "name": "track_parcel",

        "confidence": 0.9580326676368713

    },

    {

        "id": -1362472849520625133,

        "name": "leave_balance",

        "confidence": 0.04196735471487045

    }

],

"response_selector": {

    "all_retrieval_intents": [],

    "default": {

        "response": {

            "id": null,

            "responses": null,

            "response_templates": null,

            "confidence": 0.0,

            "intent_response_key": null,

            "utter_action": "utter_None",

            "template_name": "utter_None"

        },

        "ranking": []

    }

}

```

}

1. why I’m getting track\_parcel intent with .9 confidence for an untrained random input? Do I need to make any changes in the pipeline. I’m using the default pipeline. Attaching the config.yml data also. [config.yml](https://forum.rasa.com/uploads/short-url/hbXCgKxPtHtSl9J8k6uGejgKsO2.yml) (1.4 KB)

2. where can I see the newly trained nlu data. I tried to unzip the files in the model but could’t find the intents and examples.

3. For handling multi language application, I see a post that we need to append the language along with the intent. Then what should be the language given in the config.yml file?

---

<div class="post-metadata">

**Author:** ![anjaly](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/anjaly/32/5213_2.png) [@anjaly](https://forum.rasa.com/u/anjaly)\
**Post date:** [August 19, 2022, 7:05am UTC](https://forum.rasa.com/t/training-new-model-on-the-fly-using-http-api/54893/2 "2022-08-19T07:05:45Z")

</div>

Can someone please help in fixing the issues. I’m stuck du to these issues.

Thanks