# Editing utterances in production environment

**URL:** <https://forum.rasa.com/t/editing-utterances-in-production-environment/41338>\
**Category:** Rasa Open Source\
**Created:** [March 16, 2021, 11:11am UTC](https://forum.rasa.com/t/editing-utterances-in-production-environment/41338 "2021-03-16T11:11:37Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![arushi](https://avatars.discourse-cdn.com/v4/letter/a/d6d6ee/32.png) [@arushi](https://forum.rasa.com/u/arushi)\
**Post date:** [March 16, 2021, 11:11am UTC](https://forum.rasa.com/t/editing-utterances-in-production-environment/41338/1 "2021-03-16T11:11:37Z")

</div>

Is there a way to edit(add,delete) utterances in nlu.yml file without having to redeploy the project? Can the changes come into effect without redeploying ? If yes, how?

Thanks in advance.

---

<div class="post-metadata">

**Author:** ![souvikg10](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/souvikg10/32/93_2.png) [@souvikg10](https://forum.rasa.com/u/souvikg10)\
**Post date:** [March 16, 2021, 11:24am UTC](https://forum.rasa.com/t/editing-utterances-in-production-environment/41338/2 "2021-03-16T11:24:54Z")

</div>

you have to externalize your responses in an NLG server in that case that handles utterances separately or employ a CI/CD strategy that redeploys changes in an automated way.

> **[NLG Servers](https://rasa.com/docs/rasa/nlg)**
>
> An open source machine learning framework for automated text and voice-based conversations

---

<div class="post-metadata">

**Author:** ![arushi](https://avatars.discourse-cdn.com/v4/letter/a/d6d6ee/32.png) [@arushi](https://forum.rasa.com/u/arushi)\
**Post date:** [March 18, 2021, 3:44am UTC](https://forum.rasa.com/t/editing-utterances-in-production-environment/41338/3 "2021-03-18T03:44:22Z")

</div>

Thank you for the reply. I think what you suggested in more related to the response generation. I want to add examples for intents dynamically without having the need to redeploy.

Is there an API for adding examples to nlu.yml in the production environment so that just retraining would bring in effect the changes and there is no need of redeploying?

---

<div class="post-metadata">

**Author:** ![souvikg10](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/souvikg10/32/93_2.png) [@souvikg10](https://forum.rasa.com/u/souvikg10)\
**Post date:** [March 18, 2021, 5:02am UTC](https://forum.rasa.com/t/editing-utterances-in-production-environment/41338/4 "2021-03-18T05:02:47Z")

</div>

If you are using the standard rasa server then rasa loads the model only on startup and thus it needs to be restarted, you should look into k8s deployments to better orchestrate your pods.

Another approach is to use a different framework such as flask or fast api and load models with a different endpoint using the Python’s rasa library and thus dynamically update the interpreter object for inference.

---

<div class="post-metadata">

**Author:** ![souvikg10](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/souvikg10/32/93_2.png) [@souvikg10](https://forum.rasa.com/u/souvikg10)\
**Post date:** [March 18, 2021, 6:23am UTC](https://forum.rasa.com/t/editing-utterances-in-production-environment/41338/5 "2021-03-18T06:23:34Z")

</div>

Also consider A/B deployments for ML models routing traffic X percentage at a time. K8s is your friend