ben.carrasco/freeswitch-modules
1
0
Fork 0
mirror of https://github.com/jambonz/freeswitch-modules.git synced 2025-12-19 08:27:44 +00:00
Code Issues Packages Projects Releases Wiki Activity

deepgram fixes (#3)

Browse Source 
* more license stuff

Signed-off-by: Dave Horton <daveh@beachdognet.com>

* cleanup

Signed-off-by: Dave Horton <daveh@beachdognet.com>

* cleanup

Signed-off-by: Dave Horton <daveh@beachdognet.com>

* cleanup

Signed-off-by: Dave Horton <daveh@beachdognet.com>

* deepgram fixes: proper use of bugname

---------

Signed-off-by: Dave Horton <daveh@beachdognet.com>
This commit is contained in:
Dave Horton
2023-12-26 20:35:09 -05:00
committed by GitHub
parent 0bdb4a1e32
commit 3f50aaf0d7
4 changed files with 152 additions and 147 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
mod_deepgram_transcribe/audio_pipe.cpp
View File

@@ -71,7 +71,7 @@ int AudioPipe::lws_callback(struct lws *wsi,
lwsl_err("AudioPipe::lws_service_thread LWS_CALLBACK_CLIENT_CONNECTION_ERROR: %s, response status %d\n", in ? (char *)in : "(null)", rc);
if (ap) {
ap->m_state = LWS_CLIENT_FAILED;
ap->m_callback(ap->m_uuid.c_str(), AudioPipe::CONNECT_FAIL, (char *) in, ap->isFinished());
ap->m_callback(ap->m_uuid.c_str(), ap->m_bugname.c_str(), deepgram::AudioPipe::CONNECT_FAIL, (char *) in, ap->isFinished());
}
else {
lwsl_err("AudioPipe::lws_service_thread LWS_CALLBACK_CLIENT_CONNECTION_ERROR unable to find wsi %p..\n", wsi);
@@ -82,11 +82,12 @@ int AudioPipe::lws_callback(struct lws *wsi,
case LWS_CALLBACK_CLIENT_ESTABLISHED:
{
AudioPipe* ap = findAndRemovePendingConnect(wsi);
if (ap) {
*ppAp = ap;
ap->m_vhd = vhd;
ap->m_state = LWS_CLIENT_CONNECTED;
ap->m_callback(ap->m_uuid.c_str(), AudioPipe::CONNECT_SUCCESS, NULL, ap->isFinished());
ap->m_callback(ap->m_uuid.c_str(), ap->m_bugname.c_str(), deepgram::AudioPipe::CONNECT_SUCCESS, NULL, ap->isFinished());
}
else {
lwsl_err("AudioPipe::lws_service_thread LWS_CALLBACK_CLIENT_ESTABLISHED %s unable to find wsi %p..\n", ap->m_uuid.c_str(), wsi);
@@ -96,6 +97,7 @@ int AudioPipe::lws_callback(struct lws *wsi,
case LWS_CALLBACK_CLIENT_CLOSED:
{
AudioPipe* ap = *ppAp;
if (!ap) {
lwsl_err("AudioPipe::lws_service_thread LWS_CALLBACK_CLIENT_CLOSED %s unable to find wsi %p..\n", ap->m_uuid.c_str(), wsi);
return 0;
@@ -104,12 +106,12 @@ int AudioPipe::lws_callback(struct lws *wsi,
// closed by us
lwsl_debug("%s socket closed by us\n", ap->m_uuid.c_str());
ap->m_callback(ap->m_uuid.c_str(), AudioPipe::CONNECTION_CLOSED_GRACEFULLY, NULL, ap->isFinished());
ap->m_callback(ap->m_uuid.c_str(), ap->m_bugname.c_str(), deepgram::AudioPipe::CONNECTION_CLOSED_GRACEFULLY, NULL, ap->isFinished());
}
else if (ap->m_state == LWS_CLIENT_CONNECTED) {
// closed by far end
lwsl_info("%s socket closed by far end\n", ap->m_uuid.c_str());
ap->m_callback(ap->m_uuid.c_str(), AudioPipe::CONNECTION_DROPPED, NULL, ap->isFinished());
ap->m_callback(ap->m_uuid.c_str(), ap->m_bugname.c_str(), deepgram::AudioPipe::CONNECTION_DROPPED, NULL, ap->isFinished());
}
ap->m_state = LWS_CLIENT_DISCONNECTED;
ap->setClosed();
@@ -125,6 +127,7 @@ int AudioPipe::lws_callback(struct lws *wsi,
case LWS_CALLBACK_CLIENT_RECEIVE:
{
AudioPipe* ap = *ppAp;
if (!ap) {
lwsl_err("AudioPipe::lws_service_thread LWS_CALLBACK_CLIENT_RECEIVE %s unable to find wsi %p..\n", ap->m_uuid.c_str(), wsi);
return 0;
@@ -171,7 +174,7 @@ int AudioPipe::lws_callback(struct lws *wsi,
if (lws_is_final_fragment(wsi)) {
if (nullptr != ap->m_recv_buf) {
std::string msg((char *)ap->m_recv_buf, ap->m_recv_buf_ptr - ap->m_recv_buf);
ap->m_callback(ap->m_uuid.c_str(), AudioPipe::MESSAGE, msg.c_str(), ap->isFinished());
ap->m_callback(ap->m_uuid.c_str(), ap->m_bugname.c_str(), deepgram::AudioPipe::MESSAGE, msg.c_str(), ap->isFinished());
if (nullptr != ap->m_recv_buf) free(ap->m_recv_buf);
}
ap->m_recv_buf = ap->m_recv_buf_ptr = nullptr;
@@ -184,6 +187,7 @@ int AudioPipe::lws_callback(struct lws *wsi,
case LWS_CALLBACK_CLIENT_WRITEABLE:
{
AudioPipe* ap = *ppAp;
if (!ap) {
lwsl_err("AudioPipe::lws_service_thread LWS_CALLBACK_CLIENT_WRITEABLE %s unable to find wsi %p..\n", ap->m_uuid.c_str(), wsi);
return 0;
@@ -251,7 +255,6 @@ static const lws_retry_bo_t retry = {
};
struct lws_context *AudioPipe::context = nullptr;
std::string AudioPipe::protocolName;
std::mutex AudioPipe::mutex_connects;
std::mutex AudioPipe::mutex_disconnects;
std::mutex AudioPipe::mutex_writes;
@@ -424,7 +427,7 @@ void AudioPipe::initialize(int loglevel, log_emit_function logger) {
lwsl_notice("AudioPipe::initialize starting\n");
std::lock_guard<std::mutex> lock(mapMutex);
std::thread t(&AudioPipe::lws_service_thread);
std::thread t(&deepgram::AudioPipe::lws_service_thread);
stopFlag = false;
t.detach();
}
@@ -437,9 +440,9 @@ bool AudioPipe::deinitialize() {
}
// instance members
AudioPipe::AudioPipe(const char* uuid, const char* host, unsigned int port, const char* path,
AudioPipe::AudioPipe(const char* uuid, const char* bugname, const char* host, unsigned int port, const char* path,
size_t bufLen, size_t minFreespace, const char* apiKey, notifyHandler_t callback) :
m_uuid(uuid), m_host(host), m_port(port), m_path(path), m_finished(false),
m_uuid(uuid), m_host(host), m_port(port), m_path(path), m_finished(false), m_bugname(bugname),
m_audio_buffer_min_freespace(minFreespace), m_audio_buffer_max_len(bufLen), m_gracefulShutdown(false),
m_audio_buffer_write_offset(LWS_PRE), m_recv_buf(nullptr), m_recv_buf_ptr(nullptr),
m_state(LWS_CLIENT_IDLE), m_wsi(nullptr), m_vhd(nullptr), m_apiKey(apiKey), m_callback(callback) {
@@ -468,7 +471,6 @@ bool AudioPipe::connect_client(struct lws_per_vhost_data *vhd) {
i.host = i.address;
i.origin = i.address;
i.ssl_connection = LCCSCF_USE_SSL;
//i.protocol = protocolName.c_str();
i.pwsi = &(m_wsi);
m_state = LWS_CLIENT_CONNECTING;

249
mod_deepgram_transcribe/audio_pipe.hpp
View File

@@ -13,133 +13,130 @@
namespace deepgram {
class AudioPipe {
public:
enum LwsState_t {
LWS_CLIENT_IDLE,
LWS_CLIENT_CONNECTING,
LWS_CLIENT_CONNECTED,
LWS_CLIENT_FAILED,
LWS_CLIENT_DISCONNECTING,
LWS_CLIENT_DISCONNECTED
class AudioPipe {
public:
enum LwsState_t {
LWS_CLIENT_IDLE,
LWS_CLIENT_CONNECTING,
LWS_CLIENT_CONNECTED,
LWS_CLIENT_FAILED,
LWS_CLIENT_DISCONNECTING,
LWS_CLIENT_DISCONNECTED
};
enum NotifyEvent_t {
CONNECT_SUCCESS,
CONNECT_FAIL,
CONNECTION_DROPPED,
CONNECTION_CLOSED_GRACEFULLY,
MESSAGE
};
typedef void (*log_emit_function)(int level, const char *line);
typedef void (*notifyHandler_t)(const char *sessionId, const char* bugname, NotifyEvent_t event, const char* message, bool finished);
struct lws_per_vhost_data {
struct lws_context *context;
struct lws_vhost *vhost;
const struct lws_protocols *protocol;
};
static void initialize(int loglevel, log_emit_function logger);
static bool deinitialize();
static bool lws_service_thread();
// constructor
AudioPipe(const char* uuid, const char* bugname, const char* host, unsigned int port, const char* path,
size_t bufLen, size_t minFreespace, const char* apiKey, notifyHandler_t callback);
~AudioPipe();
LwsState_t getLwsState(void) { return m_state; }
std::string& getApiKey(void) {
return m_apiKey;
}
void connect(void);
void bufferForSending(const char* text);
size_t binarySpaceAvailable(void) {
return m_audio_buffer_max_len - m_audio_buffer_write_offset;
}
size_t binaryMinSpace(void) {
return m_audio_buffer_min_freespace;
}
char * binaryWritePtr(void) {
return (char *) m_audio_buffer + m_audio_buffer_write_offset;
}
void binaryWritePtrAdd(size_t len) {
m_audio_buffer_write_offset += len;
}
void binaryWritePtrResetToZero(void) {
m_audio_buffer_write_offset = 0;
}
void lockAudioBuffer(void) {
m_audio_mutex.lock();
}
void unlockAudioBuffer(void) ;
void close() ;
void finish();
void waitForClose();
void setClosed() { m_promise.set_value(); }
bool isFinished() { return m_finished;}
// no default constructor or copying
AudioPipe() = delete;
AudioPipe(const AudioPipe&) = delete;
void operator=(const AudioPipe&) = delete;
private:
static int lws_callback(struct lws *wsi, enum lws_callback_reasons reason, void *user, void *in, size_t len);
static struct lws_context *context;
static std::mutex mutex_connects;
static std::mutex mutex_disconnects;
static std::mutex mutex_writes;
static std::list<AudioPipe*> pendingConnects;
static std::list<AudioPipe*> pendingDisconnects;
static std::list<AudioPipe*> pendingWrites;
static log_emit_function logger;
static std::mutex mapMutex;
static bool stopFlag;
static AudioPipe* findAndRemovePendingConnect(struct lws *wsi);
static AudioPipe* findPendingConnect(struct lws *wsi);
static void addPendingConnect(AudioPipe* ap);
static void addPendingDisconnect(AudioPipe* ap);
static void addPendingWrite(AudioPipe* ap);
static void processPendingConnects(lws_per_vhost_data *vhd);
static void processPendingDisconnects(lws_per_vhost_data *vhd);
static void processPendingWrites(void);
bool connect_client(struct lws_per_vhost_data *vhd);
LwsState_t m_state;
std::string m_uuid;
std::string m_host;
unsigned int m_port;
std::string m_path;
std::string m_metadata;
std::mutex m_text_mutex;
std::mutex m_audio_mutex;
int m_sslFlags;
struct lws *m_wsi;
uint8_t *m_audio_buffer;
size_t m_audio_buffer_max_len;
size_t m_audio_buffer_write_offset;
size_t m_audio_buffer_min_freespace;
uint8_t* m_recv_buf;
uint8_t* m_recv_buf_ptr;
size_t m_recv_buf_len;
struct lws_per_vhost_data* m_vhd;
notifyHandler_t m_callback;
log_emit_function m_logger;
std::string m_apiKey;
bool m_gracefulShutdown;
bool m_finished;
std::string m_bugname;
std::promise<void> m_promise;
};
enum NotifyEvent_t {
CONNECT_SUCCESS,
CONNECT_FAIL,
CONNECTION_DROPPED,
CONNECTION_CLOSED_GRACEFULLY,
MESSAGE
};
typedef void (*log_emit_function)(int level, const char *line);
typedef void (*notifyHandler_t)(const char *sessionId, NotifyEvent_t event, const char* message, bool finished);
struct lws_per_vhost_data {
struct lws_context *context;
struct lws_vhost *vhost;
const struct lws_protocols *protocol;
};
static void initialize(int loglevel, log_emit_function logger);
static bool deinitialize();
static bool lws_service_thread();
// constructor
AudioPipe(const char* uuid, const char* host, unsigned int port, const char* path,
size_t bufLen, size_t minFreespace, const char* apiKey, notifyHandler_t callback);
~AudioPipe();
LwsState_t getLwsState(void) { return m_state; }
std::string& getApiKey(void) {
return m_apiKey;
}
void connect(void);
void bufferForSending(const char* text);
size_t binarySpaceAvailable(void) {
return m_audio_buffer_max_len - m_audio_buffer_write_offset;
}
size_t binaryMinSpace(void) {
return m_audio_buffer_min_freespace;
}
char * binaryWritePtr(void) {
return (char *) m_audio_buffer + m_audio_buffer_write_offset;
}
void binaryWritePtrAdd(size_t len) {
m_audio_buffer_write_offset += len;
}
void binaryWritePtrResetToZero(void) {
m_audio_buffer_write_offset = 0;
}
void lockAudioBuffer(void) {
m_audio_mutex.lock();
}
void unlockAudioBuffer(void) ;
void close() ;
void finish();
void waitForClose();
void setClosed() { m_promise.set_value(); }
bool isFinished() { return m_finished;}
// no default constructor or copying
AudioPipe() = delete;
AudioPipe(const AudioPipe&) = delete;
void operator=(const AudioPipe&) = delete;
private:
static int lws_callback(struct lws *wsi, enum lws_callback_reasons reason, void *user, void *in, size_t len);
static unsigned int nchild;
static struct lws_context *context;
static unsigned int numContexts;
static std::string protocolName;
static std::mutex mutex_connects;
static std::mutex mutex_disconnects;
static std::mutex mutex_writes;
static std::list<AudioPipe*> pendingConnects;
static std::list<AudioPipe*> pendingDisconnects;
static std::list<AudioPipe*> pendingWrites;
static log_emit_function logger;
static std::mutex mapMutex;
static bool stopFlag;
static AudioPipe* findAndRemovePendingConnect(struct lws *wsi);
static AudioPipe* findPendingConnect(struct lws *wsi);
static void addPendingConnect(AudioPipe* ap);
static void addPendingDisconnect(AudioPipe* ap);
static void addPendingWrite(AudioPipe* ap);
static void processPendingConnects(lws_per_vhost_data *vhd);
static void processPendingDisconnects(lws_per_vhost_data *vhd);
static void processPendingWrites(void);
bool connect_client(struct lws_per_vhost_data *vhd);
LwsState_t m_state;
std::string m_uuid;
std::string m_host;
unsigned int m_port;
std::string m_path;
std::string m_metadata;
std::mutex m_text_mutex;
std::mutex m_audio_mutex;
int m_sslFlags;
struct lws *m_wsi;
uint8_t *m_audio_buffer;
size_t m_audio_buffer_max_len;
size_t m_audio_buffer_write_offset;
size_t m_audio_buffer_min_freespace;
uint8_t* m_recv_buf;
uint8_t* m_recv_buf_ptr;
size_t m_recv_buf_len;
struct lws_per_vhost_data* m_vhd;
notifyHandler_t m_callback;
log_emit_function m_logger;
std::string m_apiKey;
bool m_gracefulShutdown;
bool m_finished;
std::string m_bugname;
std::promise<void> m_promise;
};
} // namespace deepgram
#endif

13
mod_deepgram_transcribe/dg_transcribe_glue.cpp
View File

@@ -265,11 +265,12 @@ namespace {
return path;
}
static void eventCallback(const char* sessionId, deepgram::AudioPipe::NotifyEvent_t event, const char* message, bool finished) {
static void eventCallback(const char* sessionId, const char* bugname,
deepgram::AudioPipe::NotifyEvent_t event, const char* message, bool finished) {
switch_core_session_t* session = switch_core_session_locate(sessionId);
if (session) {
switch_channel_t *channel = switch_core_session_get_channel(session);
switch_media_bug_t *bug = (switch_media_bug_t*) switch_channel_get_private(channel, MY_BUG_NAME);
switch_media_bug_t *bug = (switch_media_bug_t*) switch_channel_get_private(channel, bugname);
if (bug) {
private_t* tech_pvt = (private_t*) switch_core_media_bug_get_user_data(bug);
if (tech_pvt) {
@@ -343,6 +344,7 @@ namespace {
tech_pvt->channels = channels;
tech_pvt->id = ++idxCallCount;
tech_pvt->buffer_overrun_notified = 0;
strncpy(tech_pvt->bugname, bugname, MAX_BUG_LEN);
size_t buflen = LWS_PRE + (FRAME_SIZE_8000 * desiredSampling / 8000 * channels * 1000 / RTP_PACKETIZATION_PERIOD * nAudioBufferSecs);
@@ -353,7 +355,7 @@ namespace {
return SWITCH_STATUS_FALSE;
}
deepgram::AudioPipe* ap = new deepgram::AudioPipe(tech_pvt->sessionId, tech_pvt->host, tech_pvt->port, tech_pvt->path,
deepgram::AudioPipe* ap = new deepgram::AudioPipe(tech_pvt->sessionId, bugname, tech_pvt->host, tech_pvt->port, tech_pvt->path,
buflen, read_impl.decoded_bytes_per_packet, apiKey, eventCallback);
if (!ap) {
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_ERROR, "Error allocating AudioPipe\n");
@@ -400,7 +402,8 @@ extern "C" {
switch_status_t dg_transcribe_init() {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "mod_deepgram_transcribe: audio buffer (in secs): %d secs\n", nAudioBufferSecs);
int logs = LLL_ERR | LLL_WARN | LLL_NOTICE || LLL_INFO | LLL_PARSER | LLL_HEADER | LLL_EXT | LLL_CLIENT | LLL_LATENCY | LLL_DEBUG ;
int logs = LLL_ERR | LLL_WARN | LLL_NOTICE;
// | LLL_INFO | LLL_PARSER | LLL_HEADER | LLL_EXT | LLL_CLIENT | LLL_LATENCY | LLL_DEBUG ;
deepgram::AudioPipe::initialize(logs, lws_logger);
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_NOTICE, "AudioPipe::initialize completed\n");
@@ -455,7 +458,7 @@ extern "C" {
switch_status_t dg_transcribe_session_stop(switch_core_session_t *session,int channelIsClosing, char* bugname) {
switch_channel_t *channel = switch_core_session_get_channel(session);
switch_media_bug_t *bug = (switch_media_bug_t*) switch_channel_get_private(channel, MY_BUG_NAME);
switch_media_bug_t *bug = (switch_media_bug_t*) switch_channel_get_private(channel, bugname);
if (!bug) {
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_DEBUG, "dg_transcribe_session_stop: no bug - websocket conection already closed\n");
return SWITCH_STATUS_FALSE;

15
mod_deepgram_transcribe/mod_deepgram_transcribe.c
View File

@@ -45,7 +45,6 @@ static switch_bool_t capture_callback(switch_media_bug_t *bug, void *user_data,
{
private_t *tech_pvt = (private_t*) switch_core_media_bug_get_user_data(bug);
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "Got SWITCH_ABC_TYPE_CLOSE.\n");
dg_transcribe_session_stop(session, 1, tech_pvt->bugname);
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "Finished SWITCH_ABC_TYPE_CLOSE.\n");
}
@@ -74,7 +73,7 @@ static switch_status_t start_capture(switch_core_session_t *session, switch_medi
void *pUserData;
uint32_t samples_per_second;
if (switch_channel_get_private(channel, MY_BUG_NAME)) {
if (switch_channel_get_private(channel, bugname)) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "removing bug from previous transcribe\n");
do_stop(session, bugname);
}
@@ -94,7 +93,7 @@ static switch_status_t start_capture(switch_core_session_t *session, switch_medi
if ((status = switch_core_media_bug_add(session, "dg_transcribe", NULL, capture_callback, pUserData, 0, flags, &bug)) != SWITCH_STATUS_SUCCESS) {
return status;
}
switch_channel_set_private(channel, MY_BUG_NAME, bug);
switch_channel_set_private(channel, bugname, bug);
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "added media bug for dg transcribe\n");
return SWITCH_STATUS_SUCCESS;
@@ -105,13 +104,16 @@ static switch_status_t do_stop(switch_core_session_t *session, char* bugname)
switch_status_t status = SWITCH_STATUS_SUCCESS;
switch_channel_t *channel = switch_core_session_get_channel(session);
switch_media_bug_t *bug = switch_channel_get_private(channel, MY_BUG_NAME);
switch_media_bug_t *bug = switch_channel_get_private(channel, bugname);
if (bug) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "Received user command command to stop transcribe.\n");
status = dg_transcribe_session_stop(session, 0, bugname);
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "stopped transcribe.\n");
}
else {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_INFO, "do_stop: did not find bugname %s.\n", bugname);
}
return status;
}
@@ -141,7 +143,7 @@ SWITCH_STANDARD_API(dg_transcribe_function)
if ((lsession = switch_core_session_locate(argv[0]))) {
if (!strcasecmp(argv[1], "stop")) {
char *bugname = argc > 2 ? argv[2] : MY_BUG_NAME;
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "stop transcribing\n");
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "stop transcribing %s\n", bugname);
status = do_stop(lsession, bugname);
} else if (!strcasecmp(argv[1], "start")) {
char* lang = argv[2];
@@ -151,7 +153,8 @@ SWITCH_STANDARD_API(dg_transcribe_function)
flags |= SMBF_WRITE_STREAM ;
flags |= SMBF_STEREO;
}
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "start transcribing %s %s\n", lang, interim ? "interim": "complete");
switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "start transcribing %s %s %s\n",
lang, interim ? "interim": "complete", bugname);
status = start_capture(lsession, flags, lang, interim, bugname);
}
switch_core_session_rwunlock(lsession);