Implement 2-directional RTP to play a prompt when talking to agent.

[mo-g]

No description provided.

[mo-g]

Quick note, analysing the waveform for a particular sound, it consists of three clicks, each one followed by a lower amplitude tone. Measuring the wavelength gives approximately 1000, 1000, 1250 Hz. 987.77Hz (B5) and 1244.51Hz (Eb6) seem to be adequate guesses, and are also a good ratio to the boatswain's call in TOS (low high low in that case). The clicks are not a single frequency for short duration, but are complex.

[mo-g]

Make that { freq: "B5", length: 500},{ freq: "D6", length: 1000},{ freq: "B5", length: 500}

[mo-g]

I've managed to get some basic out of tune tone generation in place, though the library needs a lot more work - it's not a priority right now. Now I need to get more of the RTP library structured so it can handle two-way signals, as well as being hopefully - better written.

[mo-g]

In case anyone's curious why this is here - it's my elintrc.json.

{
    "env": {
        "node": true,
        "es2021": true
    },
    "extends": "eslint:recommended",
    "overrides": [
    ],
    "parserOptions": {
        "ecmaVersion": "latest",
        "sourceType": "module"
    },
    "rules": {
    }
}

[mo-g]

This'll be about the 50'th time I've squashed a chunk of 50 commits where I've spent hours doing annoying debugging for a simple error.

At this point, the code is restructured to where we have a separate RTPServer class, with the agent still functional. Bonus, initial testing shows the absence of an LPF makes precious little difference to our resample as far as STT is concerned - that gives us a big speedup, which makes us under budget for ms/sample on the Haswell Ultrabook VM that is my home test server, though I still need to test it on a couple others. If anyone wants to try this on a Pi 4 server that would be great data; my only Pi 4 is busy doing other things.

Most of this task is actually still left, as a lot of the work so far has really been cleanup (some of which split off and merged into Mantissa) - the codec classes need a lot of work, threading needs to be created and then we need to create some kind of audio buffer for passing back to the device. Bonus task, I'd like to fix the agent RTP server system so they're spawned on-demand rather than on launch - as that solves a problem with playing audio to the badge.

[mo-g]

And another 50 squashed!