The problem is that my slug is unslung, running busybox for most basic *nix tools and busybox’s touch command doesn’t support the -t option, so I can’t set the modification time on the file to properly queue it.  My half-assed workaround is a shell script (because I couldn’t get the System() command to run the whole string, try as I might) that nohups a background shell process that sleeps for the difference in seconds between now and the wakeup time, then creates the call file.  Not great for far-future scheduling or for large numbers of calls and it won’t survive a reboot, but it’s a start.

My Asterisk diaplan macro passes the target time/date string as the first argument, the target channel as the second, and the ${UNIQUEID} as the third.  One further wrinkle:  busybox 1.3.1, which was what came with the unslung firmware I used, doesn’t have nohup, so it won’t work; busybox 1.10.3 is what ipkg installed into /opt/ and has nohup, but the date command in 1.10.3 didn’t seem to support –date (at all, or maybe just not properly), so nohup is a reference to the 1.10.3 busybox in /opt/bin while /bin/date is a reference to the original 1.3.1 busybox.  Stupid, but it works.

(Oh, and as I told the last person I saw wearing a t-shirt that said “If it’s stupid, but it works, then it isn’t stupid,” there’s definitely such a thing as works-but-stupid–this script hackery is such a thing.  The better solution, space permitting, is probably just to install the coreutils package so as to have proper non-busybox touch and date)

In searching some more, I found that ffmpeg, one of the best swiss-army-knife-type audio/video conversion programs, specifically supports signed 16-bit little-endian PCM.  Here’s the command to convert just about any audio (anything ffmpeg can handle) to the appropriate format for Asterisk music-on-hold:

ffmpeg -i "[input file]" -ar 8000 -ac 1 -acodec pcm_s16le -f s16le "[output file].sln"

You may also want to add a volume modifier, to make the music appropriately quiet for music-on-hold, and ffmpeg allows this to be specified in dB gain like “-vol -15dB”.

First, the dialplan.

If you’ve got a text-to-speech engine and can create smoother strings of spoken words than what I’ve strung together here, you’re better off (on my Asterisk box, I’m actually using files I made with Cepstral’s Allison-8k running on my Mac).  With my limited command set, there’s no nice way to grab the output of a system command, so I capture it into a temp file, use ReadFile() on the temp file, then delete the temp file.

That dialplan called a shellscript, get_weather.sh:

This uses curl to get the weather from Yahoo, runs it through a complicated sed program which yields two lines of forecast (as &-joined lists of sound files, suitable for running as an argument to Playback() in Asterisk), then joins the two lines with 2 seconds of silence. The “echo -n `…`” is a trick for stripping the trailing newline from the output of curl/sed (a newline inside the Playback() get interpreted as part of the filename, which it isn’t).

Finally, the sed program that processes the output from Yahoo:

The “rearrange” line does most of the work of taking apart the last bits of what Yahoo gave us; the rest of the sed program is dominated by converting months/days/numbers into references to the appropriate Asterisk sound files. [Note: Since I only had today's data with which to work, I made some guesses as to what the shortened forms of most of the months and days would be, as well as how temperatures that weren't two-digit positive numbers would be presented.]