llama conversation response quotation marks

[prime remnant]

I have llama 3.2 working for notificatins around my house but when displayed in notifications the response from the llm is always enclosed in quotes. Is there any way I can parse the response in the yaml for the notification message? Thanks

[keen vortex]

Can you instruct the llm to not include quotes?

Or you could try

message: >-
  {{llama.response.speech.plain.speech}}

I assume this is a fragment of a script or automation that has a conversation process action before it?

[prime remnant]

I do tell the llm not to use quotes- unfortunately that results in it sometimes adding “okay i wont use quotes” lol inconsistent at best. It seems like its the result that just includes quotes by default. I’ll try the format you put above

[keen vortex]

Hmm interesting. Haven't noticed llm results including quotes when i use them elsewhere, and doesn't seem like the variable would be adding them, i tried testing that and wasn't seeing quotes. Might need to put the full automation script yaml so we can see if anything along the line may be causing it.

Also have you tried just the conversation process in dev tools and seen if the response itself has quotes coming from the llm?

[prime remnant]

no- thats a good idea. let me see what that shows

also - this method gets rejected by the yaml editor

with fixed indentation from above too

[polar dagger]

Process the response string separately and ask chatGPT. It should be able to tell you the answer

[keen vortex]

What did dev tools show?

[prime remnant]

yeah no quotes there - i might have still had the indentation wrong on the >- format. I think its working the way I want it to now. Thanks!