web rtc for webcam, audio, etc.

This commit is contained in:
KingKDot
2026-05-13 20:19:21 -05:00
parent 05fe31f04e
commit bf093f5a7d
20 changed files with 986 additions and 199 deletions
+130 -44
View File
@@ -12,10 +12,13 @@ import (
"github.com/pion/webrtc/v4"
)
// p2pSession is a direct browser ↔ agent peer connection. SDP and ICE
// candidates are relayed via the Overlord server's existing WS channels —
// MediaMTX is NOT involved on this path.
type p2pSession struct {
kind Kind
id string
pc *webrtc.PeerConnection
track *webrtc.TrackLocalStaticSample
onICE func(ICECandidate)
onClose func()
closeOnce sync.Once
@@ -23,28 +26,58 @@ type p2pSession struct {
var (
p2pMu sync.Mutex
p2p *p2pSession
// Keyed by kind|id so the same agent can host concurrent P2P sessions
// for desktop, webcam, and audio without sessions stomping each other.
p2pSessions = map[string]*p2pSession{}
)
func StartP2POffer(ctx context.Context, sessionID, offerSDP string, opts P2POfferCallbacks) (string, error) {
func p2pSessionKey(kind Kind, sessionID string) string {
return string(kind) + "|" + sessionID
}
// StartP2POffer accepts an SDP offer from an operator (proxied through the
// server WS), creates a peer connection with the requested tracks, and
// returns the SDP answer. The offer must contain matching m-sections — the
// browser side decides via its transceivers what to ask for. Trickle ICE
// candidates emitted by Pion are delivered to opts.OnICE.
func StartP2POffer(ctx context.Context, kind Kind, sessionID string, offerSDP string, opts P2POfferCallbacks, hasVideo, hasAudio bool) (string, error) {
_ = ctx
if sessionID == "" {
return "", errors.New("webrtcpub: empty sessionID")
}
if offerSDP == "" {
return "", errors.New("webrtcpub: empty offer SDP")
}
StopP2P()
if !hasVideo && !hasAudio {
return "", errors.New("webrtcpub: at least one of hasVideo/hasAudio must be true")
}
// Replace any previous session for this kind+id (operator restart).
StopP2P(kind, sessionID)
mediaEngine := &webrtc.MediaEngine{}
if err := mediaEngine.RegisterCodec(webrtc.RTPCodecParameters{
RTPCodecCapability: webrtc.RTPCodecCapability{
MimeType: webrtc.MimeTypeH264,
ClockRate: 90000,
SDPFmtpLine: "level-asymmetry-allowed=1;packetization-mode=1;profile-level-id=42e01f",
},
PayloadType: 102,
}, webrtc.RTPCodecTypeVideo); err != nil {
return "", fmt.Errorf("register codec: %w", err)
if hasVideo {
if err := mediaEngine.RegisterCodec(webrtc.RTPCodecParameters{
RTPCodecCapability: webrtc.RTPCodecCapability{
MimeType: webrtc.MimeTypeH264,
ClockRate: 90000,
SDPFmtpLine: "level-asymmetry-allowed=1;packetization-mode=1;profile-level-id=42e01f",
},
PayloadType: 102,
}, webrtc.RTPCodecTypeVideo); err != nil {
return "", fmt.Errorf("register video codec: %w", err)
}
}
if hasAudio {
if err := mediaEngine.RegisterCodec(webrtc.RTPCodecParameters{
RTPCodecCapability: webrtc.RTPCodecCapability{
MimeType: webrtc.MimeTypePCMU,
ClockRate: 8000,
Channels: 1,
},
PayloadType: 0,
}, webrtc.RTPCodecTypeAudio); err != nil {
return "", fmt.Errorf("register audio codec: %w", err)
}
}
api := webrtc.NewAPI(webrtc.WithMediaEngine(mediaEngine))
@@ -57,29 +90,55 @@ func StartP2POffer(ctx context.Context, sessionID, offerSDP string, opts P2POffe
return "", fmt.Errorf("new peer connection: %w", err)
}
track, err := webrtc.NewTrackLocalStaticSample(
webrtc.RTPCodecCapability{MimeType: webrtc.MimeTypeH264},
"overlord-video-p2p", "overlord-desktop",
var (
videoTrack *webrtc.TrackLocalStaticSample
audioTrack *webrtc.TrackLocalStaticSample
)
if err != nil {
_ = pc.Close()
return "", fmt.Errorf("new track: %w", err)
if hasVideo {
videoTrack, err = webrtc.NewTrackLocalStaticSample(
webrtc.RTPCodecCapability{MimeType: webrtc.MimeTypeH264},
"overlord-video-p2p-"+string(kind), "overlord-"+string(kind),
)
if err != nil {
_ = pc.Close()
return "", fmt.Errorf("new video track: %w", err)
}
tx, err := pc.AddTransceiverFromTrack(videoTrack, webrtc.RTPTransceiverInit{
Direction: webrtc.RTPTransceiverDirectionSendonly,
})
if err != nil {
_ = pc.Close()
return "", fmt.Errorf("add video transceiver: %w", err)
}
if sender := tx.Sender(); sender != nil {
go drainRTCP(sender)
}
}
transceiver, err := pc.AddTransceiverFromTrack(track, webrtc.RTPTransceiverInit{
Direction: webrtc.RTPTransceiverDirectionSendonly,
})
if err != nil {
_ = pc.Close()
return "", fmt.Errorf("add transceiver: %w", err)
}
if sender := transceiver.Sender(); sender != nil {
go drainRTCP(sender)
if hasAudio {
audioTrack, err = webrtc.NewTrackLocalStaticSample(
webrtc.RTPCodecCapability{MimeType: webrtc.MimeTypePCMU, ClockRate: 8000, Channels: 1},
"overlord-audio-p2p-"+string(kind), "overlord-"+string(kind),
)
if err != nil {
_ = pc.Close()
return "", fmt.Errorf("new audio track: %w", err)
}
tx, err := pc.AddTransceiverFromTrack(audioTrack, webrtc.RTPTransceiverInit{
Direction: webrtc.RTPTransceiverDirectionSendonly,
})
if err != nil {
_ = pc.Close()
return "", fmt.Errorf("add audio transceiver: %w", err)
}
if sender := tx.Sender(); sender != nil {
go drainRTCP(sender)
}
}
sess := &p2pSession{
kind: kind,
id: sessionID,
pc: pc,
track: track,
onICE: opts.OnICE,
onClose: opts.OnClose,
}
@@ -108,7 +167,7 @@ func StartP2POffer(ctx context.Context, sessionID, offerSDP string, opts P2POffe
})
})
pc.OnConnectionStateChange(func(state webrtc.PeerConnectionState) {
log.Printf("webrtcpub: P2P peer state=%s", state)
log.Printf("webrtcpub: P2P[%s/%s] peer state=%s", kind, sessionID, state)
switch state {
case webrtc.PeerConnectionStateFailed, webrtc.PeerConnectionStateClosed, webrtc.PeerConnectionStateDisconnected:
sess.closeAndUnregister()
@@ -132,21 +191,29 @@ func StartP2POffer(ctx context.Context, sessionID, offerSDP string, opts P2POffe
return "", fmt.Errorf("set local desc: %w", err)
}
key := p2pSessionKey(kind, sessionID)
p2pMu.Lock()
p2p = sess
p2pSessions[key] = sess
p2pMu.Unlock()
registerWriter(p2pWriterID(sessionID), &trackWriter{t: track})
writerID := p2pWriterID(kind, sessionID)
if videoTrack != nil {
registerVideoWriter(kind, writerID, &h264TrackWriter{t: videoTrack})
}
if audioTrack != nil {
registerAudioWriter(kind, writerID, newPCMUAudioWriter(audioTrack))
}
return answer.SDP, nil
}
func AddP2PICECandidate(sessionID string, c ICECandidate) {
// AddP2PICECandidate hands a remote ICE candidate to a specific P2P session.
func AddP2PICECandidate(kind Kind, sessionID string, c ICECandidate) {
if c.Candidate == "" {
return
}
p2pMu.Lock()
sess := p2p
sess := p2pSessions[p2pSessionKey(kind, sessionID)]
p2pMu.Unlock()
if sess == nil || sess.id != sessionID {
if sess == nil {
return
}
mid := c.SDPMid
@@ -160,32 +227,51 @@ func AddP2PICECandidate(sessionID string, c ICECandidate) {
}
}
// StopP2P tears down the active P2P session, if any.
func StopP2P() {
// StopP2P tears down a specific P2P session.
func StopP2P(kind Kind, sessionID string) {
p2pMu.Lock()
sess := p2p
p2p = nil
sess := p2pSessions[p2pSessionKey(kind, sessionID)]
delete(p2pSessions, p2pSessionKey(kind, sessionID))
p2pMu.Unlock()
if sess != nil {
sess.closeAndUnregister()
}
}
// StopAllP2P tears down every active P2P session. Used on shutdown.
func StopAllP2P() {
p2pMu.Lock()
keys := make([]string, 0, len(p2pSessions))
sessions := make([]*p2pSession, 0, len(p2pSessions))
for k, s := range p2pSessions {
keys = append(keys, k)
sessions = append(sessions, s)
}
for _, k := range keys {
delete(p2pSessions, k)
}
p2pMu.Unlock()
for _, s := range sessions {
s.closeAndUnregister()
}
}
func (s *p2pSession) closeAndUnregister() {
s.closeOnce.Do(func() {
unregisterWriter(p2pWriterID(s.id))
unregisterWriter(s.kind, p2pWriterID(s.kind, s.id))
_ = s.pc.Close()
if s.onClose != nil {
s.onClose()
}
})
p2pMu.Lock()
if p2p == s {
p2p = nil
key := p2pSessionKey(s.kind, s.id)
if p2pSessions[key] == s {
delete(p2pSessions, key)
}
p2pMu.Unlock()
}
func p2pWriterID(sessionID string) string {
return "p2p:" + sessionID
func p2pWriterID(kind Kind, sessionID string) string {
return "p2p:" + string(kind) + ":" + sessionID
}