cli

cli — meowcaller managed-calling demo

A small cross-platform CLI that drives the managed meowcaller calling API with a real mic and speaker. It is a separate Go module so its audio/WhatsApp dependencies (miniaudio, whatsmeow, sqlite) stay out of the library.

The command owns only the whatsmeow login boilerplate (QR pairing + connect) and the logger — then hands the connected client to meowcaller.NewClient and does everything through the managed Call API. Mic/speaker come from the opt-in cgo subpackage meowcaller/audio/malgo, so a C compiler must be on PATH.

Commands

cli call <target>              Place a 1:1 call; your mic -> peer, peer -> your speaker.
cli play <target> <file>       Place a call and stream a .mp3/.wav/.opus file to the
                               peer (peer audio still goes to the speaker).
cli listen                     Log in and print incoming call signaling.
cli autoaccept [record.wav]    Auto-answer incoming calls, wiring mic <-> speaker, or
                               recording the peer's audio to a .wav file.
cli web                       Open a localhost call console for audio/video calls and
                              mid-call video upgrades.

<target> is a phone number (+15551234567), a phone JID, or a LID JID (...@lid). Run from this directory:

go run . call +15551234567
go run . play +15551234567 hold-music.mp3
go run . autoaccept greeting.wav
go run . web

The web command prints an ephemeral localhost URL. Open it in Chromium, pair the test client from the QR code if needed, and use the controls to dial, answer, reject, hang up, start video, accept an upgrade, or stop video. The page encodes the local camera as H.264 constrained baseline at 640x480 and 15 fps, displays peer H.264, and honors peer keyframe requests and device orientation. It also sends and displays call emoji reactions. Audio continues through the machine's default microphone and speaker.

What it shows

The managed API hides all the signaling/keying/relay/media that this example used to hand-roll:

client := meowcaller.NewClient(wa, meowcaller.WithLogger(*zerolog.Ctx(ctx)))

call, _ := client.Call(ctx, target)          // place a 1:1 call
call.Play(mic)                               // a Player streams a source (mic/file) out
call.Receive(speaker)                        // a sink takes the peer's decoded audio
call.OnReady(func() { ... })
call.OnEnd(func(reason string) { ... })
call.OnReaction(func(reaction meowcaller.CallReaction) { ... })
call.SendReaction("👍")

client.OnIncomingCall(func(c *meowcaller.Call) { c.Answer() /* or c.Reject() */ })

Audio sources decode to the codec's 16 kHz mono frames behind the scenes: meowcaller.MP3File / WAVFile / OpusFile (and malgo.Mic()); sinks are malgo.Speaker() / meowcaller.WAVRecorder / meowcaller.SinkFunc.

NOT VALIDATED: the live relay hop (DTLS handshake to WhatsApp's relay) has no test vector and is only exercisable against a real relay — expect to debug it on a real call. Everything it feeds (codec, keying, framing) is KAT-verified.

Notes

• Requires cgo (a C compiler) for miniaudio via meowcaller/audio/malgo.
• whatsmeow's session lives in a local wa-voip.db (QR pairing on first run); it is gitignored and disposable (delete it to re-pair).
• MEOW_LOG_LEVEL=trace go run . call ... surfaces the full per-frame trace across the whole stack.