Test, test, and test some more

[jpagh]

Need to test the voice interactions and improve the prompts in order to improve the caller's experience.

Follow up comments should be commits related to improving the prompts and voice experience.

[jpagh]

feat: enhance conversation flow in: reduce narration; service area; adverse party; household composition

[jpagh]

fix: improved the household composition question to make it smoother and more easily answered

[jpagh]

feat: added county to caller address; populate in legalserver: county of residence; service area as county of dispute
This was kind of bad to realize that I was putting the service area FIPS as the county of residence and not as the county of dispute. It's fixed now, and we ask the county of residence along with the caller's address, but whoops!

feat: create a note when assets listing is empty and total value is 0
This is just to ensure that it's known that the bot did not record any assets, not that the omission of the note is potentially an error.

feat: updated validation rules so that no income was stored as a proper "no household income" entry, improving the legalserver intake data
This is similar to the above, absence of income isn't an error. This ensures that the first income field in the intake is correctly filled out as "No Household Income".

[jpagh]

I tried gpt-4o-mini for the LLM and gpt-4o-mini-transcribe for STT just to see, and it wasn't good. They work fine for the Python test client but didn't handle income and some of the other more complicated nodes/function calls as well.