Allow for dynamic responses when using the 'Sentence' trigger

[rickmoonex]

At the time of creating this request, there is no way to create 'dynamic' responses for the Sentence trigger.
The response is defined in the Sentence node and is immediately returned when said sentence is triggered. This does not allow for create templated responses, or for forwarding an error to the user.

---

My suggested fix is as follows:

Inside the handle_trigger callback of the websocket_sentence function we register a unique id for that specific Sentence trigger in a map. This id is return by the node in Node-RED.

Create a new 'Response' node that takes in the Sentence trigger id and a response value.
Create a new websocket command, maybe call it webhook_response, that listens for requests from the new Response node and maps the given response to the Sentence trigger id in the map.

Inside the handle_trigger callback of the websocket_sentence function we wait for the Sentence trigger id in the map to be assigned a response value from the websocket command. If this does not happen within a set timeout period we return a default response.

If a response node is fired while there's already a response registered we throw an error.

---

I will start by forking the repo and creating a PoC of this solution. Please give feedback on this proposed solution.

Thanks!

[zachowj]

I like all your ideas.

Here are the initial notes I put together while brainstorming how to approach this a while back. These ideas aren’t fully developed yet—just some rough thoughts I wanted to capture.

---

Option 1 (less preferred)

Allow the sentence response field to accept Jinja2 templates.

• Drawback: Users would need to learn Jinja2 to create meaningful responses.
• Inconsistency: Jinja2 is not used in any other Home Assistant nodes.

Option 2

Introduce two modes for the sentence node: Trigger and Response, similar to most entity nodes.

• Node Structure: Convert it into a dual-controller node, with one controller for each mode (similar to the device node setup).
• Response Handling: Modify the response field so that HA can either expect a response from Node-RED or use a static response.
• Implementation: Aim to replicate the functionality of the HTTP In and HTTP Response nodes.
  • Explore the possibility of using the existing WebSocket message ID to send back a response (feasibility needs further investigation).

[rickmoonex]

I'll get started with implementing Option 2.

Using the already existing Websocket message id reduces complexity but can cause problems down the line I think.

I'll start experimenting...

[zachowj]

I’m not sure if there’s been any progress on this, but I was able to find some time to complete the Home Assistant side. Now, I just need to wrap up a few remaining tasks on the Node-RED side.