rasa-pro 3.12.0.dev10__py3-none-any.whl → 3.12.0.dev11__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.


This version of rasa-pro might be problematic. Click here for more details.

Files changed (37) hide show
  1. rasa/cli/inspect.py +20 -1
  2. rasa/cli/shell.py +3 -3
  3. rasa/core/actions/action.py +5 -6
  4. rasa/core/actions/forms.py +6 -3
  5. rasa/core/channels/__init__.py +2 -0
  6. rasa/core/channels/voice_stream/browser_audio.py +1 -0
  7. rasa/core/channels/voice_stream/call_state.py +7 -1
  8. rasa/core/channels/voice_stream/genesys.py +331 -0
  9. rasa/core/channels/voice_stream/tts/cartesia.py +16 -3
  10. rasa/core/channels/voice_stream/twilio_media_streams.py +2 -1
  11. rasa/core/channels/voice_stream/voice_channel.py +2 -1
  12. rasa/core/policies/flows/flow_executor.py +3 -41
  13. rasa/core/run.py +4 -3
  14. rasa/dialogue_understanding/generator/command_generator.py +104 -1
  15. rasa/dialogue_understanding/generator/llm_based_command_generator.py +40 -6
  16. rasa/dialogue_understanding/generator/multi_step/multi_step_llm_command_generator.py +1 -1
  17. rasa/dialogue_understanding/generator/nlu_command_adapter.py +41 -2
  18. rasa/dialogue_understanding/generator/single_step/single_step_llm_command_generator.py +1 -1
  19. rasa/dialogue_understanding/generator/utils.py +32 -1
  20. rasa/dialogue_understanding/processor/command_processor.py +10 -12
  21. rasa/dialogue_understanding_test/README.md +50 -0
  22. rasa/dialogue_understanding_test/test_case_simulation/test_case_tracker_simulator.py +3 -3
  23. rasa/model_service.py +4 -0
  24. rasa/model_training.py +24 -27
  25. rasa/shared/core/constants.py +6 -2
  26. rasa/shared/core/domain.py +11 -20
  27. rasa/shared/core/slot_mappings.py +143 -107
  28. rasa/shared/core/slots.py +4 -2
  29. rasa/telemetry.py +43 -13
  30. rasa/utils/common.py +0 -1
  31. rasa/validator.py +31 -74
  32. rasa/version.py +1 -1
  33. {rasa_pro-3.12.0.dev10.dist-info → rasa_pro-3.12.0.dev11.dist-info}/METADATA +1 -1
  34. {rasa_pro-3.12.0.dev10.dist-info → rasa_pro-3.12.0.dev11.dist-info}/RECORD +37 -36
  35. {rasa_pro-3.12.0.dev10.dist-info → rasa_pro-3.12.0.dev11.dist-info}/NOTICE +0 -0
  36. {rasa_pro-3.12.0.dev10.dist-info → rasa_pro-3.12.0.dev11.dist-info}/WHEEL +0 -0
  37. {rasa_pro-3.12.0.dev10.dist-info → rasa_pro-3.12.0.dev11.dist-info}/entry_points.txt +0 -0
rasa/cli/inspect.py CHANGED
@@ -9,6 +9,10 @@ from rasa import telemetry
9
9
  from rasa.cli import SubParsersAction
10
10
  from rasa.cli.arguments import shell as arguments
11
11
  from rasa.core import constants
12
+ from rasa.engine.storage.local_model_storage import LocalModelStorage
13
+ from rasa.exceptions import ModelNotFound
14
+ from rasa.model import get_local_model
15
+ from rasa.shared.utils.cli import print_error
12
16
  from rasa.utils.cli import remove_argument_from_parser
13
17
 
14
18
 
@@ -55,6 +59,8 @@ async def open_inspector_in_browser(server_url: Text, voice: bool = False) -> No
55
59
  def inspect(args: argparse.Namespace) -> None:
56
60
  """Inspect the bot using the most recent model."""
57
61
  import rasa.cli.run
62
+ from rasa.cli.utils import get_validated_path
63
+ from rasa.shared.constants import DEFAULT_MODELS_PATH
58
64
 
59
65
  async def after_start_hook_open_inspector(_: Sanic, __: AbstractEventLoop) -> None:
60
66
  """Hook to open the browser on server start."""
@@ -71,5 +77,18 @@ def inspect(args: argparse.Namespace) -> None:
71
77
  args.credentials = None
72
78
  args.server_listeners = [(after_start_hook_open_inspector, "after_server_start")]
73
79
 
74
- telemetry.track_inspect_started(args.connector)
80
+ model = get_validated_path(args.model, "model", DEFAULT_MODELS_PATH)
81
+
82
+ try:
83
+ model = get_local_model(model)
84
+ except ModelNotFound:
85
+ print_error(
86
+ "No model found. Train a model before running the "
87
+ "server using `rasa train`."
88
+ )
89
+ return
90
+
91
+ metadata = LocalModelStorage.metadata_from_archive(model)
92
+
93
+ telemetry.track_inspect_started(args.connector, metadata.assistant_id)
75
94
  rasa.cli.run.run(args)
rasa/cli/shell.py CHANGED
@@ -95,7 +95,7 @@ def shell_nlu(args: argparse.Namespace) -> None:
95
95
  )
96
96
  return
97
97
 
98
- telemetry.track_shell_started("nlu")
98
+ telemetry.track_shell_started("nlu", metadata.assistant_id)
99
99
  rasa.nlu.run.run_cmdline(model)
100
100
 
101
101
 
@@ -129,12 +129,12 @@ def shell(args: argparse.Namespace) -> None:
129
129
  if metadata.training_type == TrainingType.NLU:
130
130
  import rasa.nlu.run
131
131
 
132
- telemetry.track_shell_started("nlu")
132
+ telemetry.track_shell_started("nlu", metadata.assistant_id)
133
133
 
134
134
  rasa.nlu.run.run_cmdline(model)
135
135
  else:
136
136
  import rasa.cli.run
137
137
 
138
- telemetry.track_shell_started("rasa")
138
+ telemetry.track_shell_started("rasa", metadata.assistant_id)
139
139
 
140
140
  rasa.cli.run.run(args)
@@ -72,7 +72,6 @@ from rasa.shared.core.constants import (
72
72
  ACTION_UNLIKELY_INTENT_NAME,
73
73
  ACTION_VALIDATE_SLOT_MAPPINGS,
74
74
  DEFAULT_SLOT_NAMES,
75
- KEY_MAPPING_TYPE,
76
75
  KNOWLEDGE_BASE_SLOT_NAMES,
77
76
  REQUESTED_SLOT,
78
77
  USER_INTENT_OUT_OF_SCOPE,
@@ -112,6 +111,7 @@ if TYPE_CHECKING:
112
111
  from rasa.core.channels.channel import OutputChannel
113
112
  from rasa.core.nlg import NaturalLanguageGenerator
114
113
  from rasa.shared.core.events import IntentPrediction
114
+ from rasa.shared.core.slot_mappings import SlotMapping
115
115
 
116
116
  logger = logging.getLogger(__name__)
117
117
 
@@ -1222,7 +1222,7 @@ class ActionExtractSlots(Action):
1222
1222
 
1223
1223
  async def _execute_custom_action(
1224
1224
  self,
1225
- mapping: Dict[Text, Any],
1225
+ mapping: "SlotMapping",
1226
1226
  executed_custom_actions: Set[Text],
1227
1227
  output_channel: "OutputChannel",
1228
1228
  nlg: "NaturalLanguageGenerator",
@@ -1230,7 +1230,7 @@ class ActionExtractSlots(Action):
1230
1230
  domain: "Domain",
1231
1231
  calm_custom_action_names: Optional[Set[str]] = None,
1232
1232
  ) -> Tuple[List[Event], Set[Text]]:
1233
- custom_action = mapping.get("action")
1233
+ custom_action = mapping.run_action_every_turn
1234
1234
 
1235
1235
  if not custom_action or custom_action in executed_custom_actions:
1236
1236
  return [], executed_custom_actions
@@ -1331,10 +1331,9 @@ class ActionExtractSlots(Action):
1331
1331
  slot_events.append(SlotSet(slot.name, slot_value))
1332
1332
 
1333
1333
  for mapping in slot.mappings:
1334
- mapping_type = SlotMappingType(mapping.get(KEY_MAPPING_TYPE))
1335
- should_fill_custom_slot = mapping_type == SlotMappingType.CUSTOM
1334
+ should_fill_controlled_slot = mapping.type == SlotMappingType.CONTROLLED
1336
1335
 
1337
- if should_fill_custom_slot:
1336
+ if should_fill_controlled_slot:
1338
1337
  (
1339
1338
  custom_evts,
1340
1339
  executed_custom_actions,
@@ -2,7 +2,7 @@ import copy
2
2
  import itertools
3
3
  import json
4
4
  import logging
5
- from typing import Any, Dict, List, Optional, Set, Text, Union
5
+ from typing import TYPE_CHECKING, Any, Dict, List, Optional, Set, Text, Union
6
6
 
7
7
  import structlog
8
8
 
@@ -35,6 +35,9 @@ from rasa.shared.core.slots import ListSlot
35
35
  from rasa.shared.core.trackers import DialogueStateTracker
36
36
  from rasa.utils.endpoints import EndpointConfig
37
37
 
38
+ if TYPE_CHECKING:
39
+ from rasa.shared.core.slot_mappings import SlotMapping
40
+
38
41
  logger = logging.getLogger(__name__)
39
42
  structlogger = structlog.get_logger()
40
43
 
@@ -171,7 +174,7 @@ class FormAction(LoopAction):
171
174
  return unique_entity_slot_mappings
172
175
 
173
176
  def entity_mapping_is_unique(
174
- self, slot_mapping: Dict[Text, Any], domain: Domain
177
+ self, slot_mapping: "SlotMapping", domain: Domain
175
178
  ) -> bool:
176
179
  """Verifies if the from_entity mapping is unique."""
177
180
  if not self._have_unique_entity_mappings_been_initialized:
@@ -179,7 +182,7 @@ class FormAction(LoopAction):
179
182
  self._unique_entity_mappings = self._create_unique_entity_mappings(domain)
180
183
  self._have_unique_entity_mappings_been_initialized = True
181
184
 
182
- mapping_as_string = json.dumps(slot_mapping, sort_keys=True)
185
+ mapping_as_string = json.dumps(slot_mapping.as_dict(), sort_keys=True)
183
186
  return mapping_as_string in self._unique_entity_mappings
184
187
 
185
188
  @staticmethod
@@ -32,6 +32,7 @@ from rasa.core.channels.vier_cvg import CVGInput
32
32
  from rasa.core.channels.voice_stream.twilio_media_streams import (
33
33
  TwilioMediaStreamsInputChannel,
34
34
  )
35
+ from rasa.core.channels.voice_stream.genesys import GenesysInputChannel
35
36
  from rasa.core.channels.studio_chat import StudioChatInput
36
37
 
37
38
  input_channel_classes: List[Type[InputChannel]] = [
@@ -55,6 +56,7 @@ input_channel_classes: List[Type[InputChannel]] = [
55
56
  JambonzVoiceReadyInput,
56
57
  TwilioMediaStreamsInputChannel,
57
58
  BrowserAudioInputChannel,
59
+ GenesysInputChannel,
58
60
  StudioChatInput,
59
61
  ]
60
62
 
@@ -65,6 +65,7 @@ class BrowserAudioInputChannel(VoiceInputChannel):
65
65
  def map_input_message(
66
66
  self,
67
67
  message: Any,
68
+ ws: Websocket,
68
69
  ) -> VoiceChannelAction:
69
70
  data = json.loads(message)
70
71
  if "audio" in data:
@@ -1,6 +1,6 @@
1
1
  import asyncio
2
2
  from contextvars import ContextVar
3
- from dataclasses import dataclass
3
+ from dataclasses import dataclass, field
4
4
  from typing import Optional
5
5
 
6
6
  from werkzeug.local import LocalProxy
@@ -19,6 +19,12 @@ class CallState:
19
19
  should_hangup: bool = False
20
20
  connection_failed: bool = False
21
21
 
22
+ # Genesys requires the server and client each maintain a
23
+ # monotonically increasing message sequence number.
24
+ client_sequence_number: int = 0
25
+ server_sequence_number: int = 0
26
+ audio_buffer: bytearray = field(default_factory=bytearray)
27
+
22
28
 
23
29
  _call_state: ContextVar[CallState] = ContextVar("call_state")
24
30
  call_state = LocalProxy(_call_state)
@@ -0,0 +1,331 @@
1
+ import asyncio
2
+ import json
3
+ from typing import Any, Awaitable, Callable, Dict, Optional, Text
4
+
5
+ import structlog
6
+ from sanic import ( # type: ignore[attr-defined]
7
+ Blueprint,
8
+ HTTPResponse,
9
+ Request,
10
+ Websocket,
11
+ response,
12
+ )
13
+
14
+ from rasa.core.channels import UserMessage
15
+ from rasa.core.channels.voice_ready.utils import CallParameters
16
+ from rasa.core.channels.voice_stream.audio_bytes import RasaAudioBytes
17
+ from rasa.core.channels.voice_stream.call_state import (
18
+ call_state,
19
+ )
20
+ from rasa.core.channels.voice_stream.tts.tts_engine import TTSEngine
21
+ from rasa.core.channels.voice_stream.voice_channel import (
22
+ ContinueConversationAction,
23
+ EndConversationAction,
24
+ NewAudioAction,
25
+ VoiceChannelAction,
26
+ VoiceInputChannel,
27
+ VoiceOutputChannel,
28
+ )
29
+
30
+ # Not mentioned in the documentation but observed in Geneys's example
31
+ # https://github.com/GenesysCloudBlueprints/audioconnector-server-reference-implementation
32
+ MAXIMUM_BINARY_MESSAGE_SIZE = 64000 # 64KB
33
+ logger = structlog.get_logger(__name__)
34
+
35
+
36
+ def map_call_params(data: Dict[Text, Any]) -> CallParameters:
37
+ """Map the twilio stream parameters to the CallParameters dataclass."""
38
+ parameters = data["parameters"]
39
+ participant = parameters["participant"]
40
+ # sent as {"ani": "tel:+491604697810"}
41
+ ani = participant.get("ani", "")
42
+ user_phone = ani.split(":")[-1] if ani else ""
43
+
44
+ return CallParameters(
45
+ call_id=parameters.get("conversationId", ""),
46
+ user_phone=user_phone,
47
+ bot_phone=participant.get("dnis", ""),
48
+ )
49
+
50
+
51
+ class GenesysOutputChannel(VoiceOutputChannel):
52
+ @classmethod
53
+ def name(cls) -> str:
54
+ return "genesys"
55
+
56
+ async def send_audio_bytes(
57
+ self, recipient_id: str, audio_bytes: RasaAudioBytes
58
+ ) -> None:
59
+ """
60
+ Send audio bytes to the recipient with buffering.
61
+
62
+ Genesys throws a rate limit error with too many audio messages.
63
+ To avoid this, we buffer the audio messages and send them in chunks.
64
+
65
+ - global.inbound.binary.average.rate.per.second: 5
66
+ The allowed average rate per second of inbound binary data
67
+
68
+ - global.inbound.binary.max: 25
69
+ The maximum number of inbound binary data messages
70
+ that can be sent instantaneously
71
+
72
+ https://developer.genesys.cloud/organization/organization/limits#audiohook
73
+ """
74
+ call_state.audio_buffer.extend(audio_bytes)
75
+
76
+ # If we receive a non-standard chunk size, assume it's the end of a sequence
77
+ # or buffer is more than 32KB (this is half of genesys's max audio message size)
78
+ if len(audio_bytes) != 1024 or len(call_state.audio_buffer) >= (
79
+ MAXIMUM_BINARY_MESSAGE_SIZE / 2
80
+ ):
81
+ # TODO: we should send the buffer when we receive a synthesis complete event
82
+ # from TTS. This will ensure that the last audio chunk is always sent.
83
+ await self._send_audio_buffer(self.voice_websocket)
84
+
85
+ async def _send_audio_buffer(self, ws: Websocket) -> None:
86
+ """Send the audio buffer to the recipient if it's not empty."""
87
+ if call_state.audio_buffer:
88
+ buffer_bytes = bytes(call_state.audio_buffer)
89
+ await self._send_bytes_to_ws(ws, buffer_bytes)
90
+ call_state.audio_buffer.clear()
91
+
92
+ async def _send_bytes_to_ws(self, ws: Websocket, data: bytes) -> None:
93
+ """Send audio bytes to the recipient as a binary websocket message."""
94
+ if len(data) <= MAXIMUM_BINARY_MESSAGE_SIZE:
95
+ await self.voice_websocket.send(data)
96
+ else:
97
+ # split the audio into chunks
98
+ current_position = 0
99
+ while current_position < len(data):
100
+ end_position = min(
101
+ current_position + MAXIMUM_BINARY_MESSAGE_SIZE, len(data)
102
+ )
103
+ await self.voice_websocket.send(data[current_position:end_position])
104
+ current_position = end_position
105
+
106
+ async def send_marker_message(self, recipient_id: str) -> None:
107
+ """Send a message that marks positions in the audio stream."""
108
+ pass
109
+
110
+
111
+ class GenesysInputChannel(VoiceInputChannel):
112
+ @classmethod
113
+ def name(cls) -> str:
114
+ return "genesys"
115
+
116
+ def __init__(self, *args: Any, **kwargs: Any) -> None:
117
+ super().__init__(*args, **kwargs)
118
+
119
+ def _get_next_sequence(self) -> int:
120
+ """
121
+ Get the next message sequence number
122
+ Rasa == Server
123
+ Genesys == Client
124
+
125
+ Genesys requires the server and client each maintain a
126
+ monotonically increasing message sequence number.
127
+ """
128
+ cs = call_state
129
+ cs.server_sequence_number += 1 # type: ignore[attr-defined]
130
+ return cs.server_sequence_number
131
+
132
+ def _get_last_client_sequence(self) -> int:
133
+ """Get the last client(Genesys) sequence number."""
134
+ return call_state.client_sequence_number
135
+
136
+ def _update_client_sequence(self, seq: int) -> None:
137
+ """Update the client(Genesys) sequence number."""
138
+ if seq - call_state.client_sequence_number != 1:
139
+ logger.warning(
140
+ "genesys.update_client_sequence.sequence_gap",
141
+ received_seq=seq,
142
+ last_seq=call_state.client_sequence_number,
143
+ )
144
+ call_state.client_sequence_number = seq # type: ignore[attr-defined]
145
+
146
+ def channel_bytes_to_rasa_audio_bytes(self, input_bytes: bytes) -> RasaAudioBytes:
147
+ return RasaAudioBytes(input_bytes)
148
+
149
+ async def collect_call_parameters(
150
+ self, channel_websocket: Websocket
151
+ ) -> Optional[CallParameters]:
152
+ """Call Parameters are collected during the open event."""
153
+ async for message in channel_websocket:
154
+ data = json.loads(message)
155
+ self._update_client_sequence(data["seq"])
156
+ if data.get("type") == "open":
157
+ call_params = await self.handle_open(channel_websocket, data)
158
+ return call_params
159
+ else:
160
+ logger.error("genesys.receive.unexpected_initial_message", message=data)
161
+
162
+ return None
163
+
164
+ def map_input_message(
165
+ self,
166
+ message: Any,
167
+ ws: Websocket,
168
+ ) -> VoiceChannelAction:
169
+ # if message is binary, it's audio
170
+ if isinstance(message, bytes):
171
+ return NewAudioAction(self.channel_bytes_to_rasa_audio_bytes(message))
172
+ else:
173
+ # process text message
174
+ data = json.loads(message)
175
+ self._update_client_sequence(data["seq"])
176
+ msg_type = data.get("type")
177
+ if msg_type == "close":
178
+ logger.info("genesys.handle_close", message=data)
179
+ self.handle_close(ws, data)
180
+ return EndConversationAction()
181
+ elif msg_type == "ping":
182
+ logger.info("genesys.handle_ping", message=data)
183
+ self.handle_ping(ws, data)
184
+ elif msg_type == "playback_started":
185
+ logger.debug("genesys.handle_playback_started", message=data)
186
+ call_state.is_bot_speaking = True # type: ignore[attr-defined]
187
+ elif msg_type == "playback_completed":
188
+ logger.debug("genesys.handle_playback_completed", message=data)
189
+ call_state.is_bot_speaking = False # type: ignore[attr-defined]
190
+ if call_state.should_hangup:
191
+ logger.info("genesys.hangup")
192
+ self.disconnect(ws, data)
193
+ elif msg_type == "dtmf":
194
+ logger.info("genesys.handle_dtmf", message=data)
195
+ elif msg_type == "error":
196
+ logger.warning("genesys.handle_error", message=data)
197
+ else:
198
+ logger.warning("genesys.map_input_message.unknown_type", message=data)
199
+
200
+ return ContinueConversationAction()
201
+
202
+ def create_output_channel(
203
+ self, voice_websocket: Websocket, tts_engine: TTSEngine
204
+ ) -> VoiceOutputChannel:
205
+ return GenesysOutputChannel(
206
+ voice_websocket,
207
+ tts_engine,
208
+ self.tts_cache,
209
+ )
210
+
211
+ async def handle_open(self, ws: Websocket, message: dict) -> CallParameters:
212
+ """Handle initial open transaction from Genesys."""
213
+ call_parameters = map_call_params(message)
214
+ params = message["parameters"]
215
+ media_options = params.get("media", [])
216
+
217
+ # Send opened response
218
+ if media_options:
219
+ logger.info("genesys.handle_open", media_parameter=media_options[0])
220
+ response = {
221
+ "version": "2",
222
+ "type": "opened",
223
+ "seq": self._get_next_sequence(),
224
+ "clientseq": self._get_last_client_sequence(),
225
+ "id": message.get("id"),
226
+ "parameters": {"startPaused": False, "media": [media_options[0]]},
227
+ }
228
+ logger.debug("genesys.handle_open.opened", response=response)
229
+ await ws.send(json.dumps(response))
230
+ else:
231
+ logger.warning(
232
+ "genesys.handle_open.no_media_formats", client_message=message
233
+ )
234
+ return call_parameters
235
+
236
+ def handle_ping(self, ws: Websocket, message: dict) -> None:
237
+ """Handle ping message from Genesys."""
238
+ response = {
239
+ "version": "2",
240
+ "type": "pong",
241
+ "seq": self._get_next_sequence(),
242
+ "clientseq": message.get("seq"),
243
+ "id": message.get("id"),
244
+ "parameters": {},
245
+ }
246
+ logger.debug("genesys.handle_ping.pong", response=response)
247
+ _schedule_ws_task(ws.send(json.dumps(response)))
248
+
249
+ def handle_close(self, ws: Websocket, message: dict) -> None:
250
+ """Handle close message from Genesys."""
251
+ response = {
252
+ "version": "2",
253
+ "type": "closed",
254
+ "seq": self._get_next_sequence(),
255
+ "clientseq": self._get_last_client_sequence(),
256
+ "id": message.get("id"),
257
+ "parameters": message.get("parameters", {}),
258
+ }
259
+ logger.debug("genesys.handle_close.closed", response=response)
260
+
261
+ _schedule_ws_task(ws.send(json.dumps(response)))
262
+ _schedule_ws_task(ws.close())
263
+
264
+ def disconnect(self, ws: Websocket, data: dict) -> None:
265
+ """
266
+ Send disconnect message to Genesys.
267
+
268
+ https://developer.genesys.cloud/devapps/audiohook/protocol-reference#disconnect
269
+ It should be used to hangup the call.
270
+ Genesys will respond with a "close" message to us
271
+ that is handled by the handle_close method.
272
+ """
273
+ message = {
274
+ "version": "2",
275
+ "type": "disconnect",
276
+ "seq": self._get_next_sequence(),
277
+ "clientseq": self._get_last_client_sequence(),
278
+ "id": data.get("id"),
279
+ "parameters": {
280
+ "reason": "completed",
281
+ # arbitrary values can be sent here
282
+ },
283
+ }
284
+ logger.debug("genesys.disconnect", message=message)
285
+ _schedule_ws_task(ws.send(json.dumps(message)))
286
+
287
+ def blueprint(
288
+ self, on_new_message: Callable[[UserMessage], Awaitable[Any]]
289
+ ) -> Blueprint:
290
+ """Defines a Sanic blueprint for the voice input channel."""
291
+ blueprint = Blueprint("genesys", __name__)
292
+
293
+ @blueprint.route("/", methods=["GET"])
294
+ async def health(_: Request) -> HTTPResponse:
295
+ return response.json({"status": "ok"})
296
+
297
+ @blueprint.websocket("/websocket") # type: ignore[misc]
298
+ async def receive(request: Request, ws: Websocket) -> None:
299
+ logger.debug(
300
+ "genesys.receive",
301
+ audiohook_session_id=request.headers.get("audiohook-session-id"),
302
+ )
303
+ # validate required headers
304
+ required_headers = [
305
+ "audiohook-organization-id",
306
+ "audiohook-correlation-id",
307
+ "audiohook-session-id",
308
+ "x-api-key",
309
+ ]
310
+
311
+ for header in required_headers:
312
+ if header not in request.headers:
313
+ await ws.close(1008, f"Missing required header: {header}")
314
+ return
315
+
316
+ # TODO: validate API key header
317
+ # process audio streaming
318
+ logger.info("genesys.receive", message="Starting audio streaming")
319
+ await self.run_audio_streaming(on_new_message, ws)
320
+
321
+ return blueprint
322
+
323
+
324
+ def _schedule_ws_task(coro: Awaitable[Any]) -> None:
325
+ """Helper function to schedule a coroutine in the event loop.
326
+
327
+ Args:
328
+ coro: The coroutine to schedule
329
+ """
330
+ loop = asyncio.get_running_loop()
331
+ loop.call_soon_threadsafe(lambda: loop.create_task(coro))
@@ -1,3 +1,5 @@
1
+ import base64
2
+ import json
1
3
  import os
2
4
  from dataclasses import dataclass
3
5
  from typing import AsyncIterator, Dict, Optional
@@ -39,7 +41,7 @@ class CartesiaTTS(TTSEngine[CartesiaTTSConfig]):
39
41
  @staticmethod
40
42
  def get_tts_endpoint() -> str:
41
43
  """Create the endpoint string for cartesia."""
42
- return "https://api.cartesia.ai/tts/bytes"
44
+ return "https://api.cartesia.ai/tts/sse"
43
45
 
44
46
  @staticmethod
45
47
  def get_request_body(text: str, config: CartesiaTTSConfig) -> Dict:
@@ -85,8 +87,19 @@ class CartesiaTTS(TTSEngine[CartesiaTTSConfig]):
85
87
  url, headers=headers, json=payload, chunked=True
86
88
  ) as response:
87
89
  if 200 <= response.status < 300:
88
- async for data in response.content.iter_chunked(1024):
89
- yield self.engine_bytes_to_rasa_audio_bytes(data)
90
+ async for chunk in response.content:
91
+ # we are looking for chunks in the response that look like
92
+ # b"data: {..., data: <base64 encoded audio bytes> ...}"
93
+ # and extract the audio bytes from that
94
+ if chunk.startswith(b"data: "):
95
+ json_bytes = chunk[5:-1]
96
+ json_data = json.loads(json_bytes.decode())
97
+ if "data" in json_data:
98
+ base64_encoded_bytes = json_data["data"]
99
+ channel_bytes = base64.b64decode(base64_encoded_bytes)
100
+ yield self.engine_bytes_to_rasa_audio_bytes(
101
+ channel_bytes
102
+ )
90
103
  return
91
104
  else:
92
105
  structlogger.error(
@@ -98,6 +98,7 @@ class TwilioMediaStreamsInputChannel(VoiceInputChannel):
98
98
  def map_input_message(
99
99
  self,
100
100
  message: Any,
101
+ ws: Websocket,
101
102
  ) -> VoiceChannelAction:
102
103
  data = json.loads(message)
103
104
  if data["event"] == "media":
@@ -142,7 +143,7 @@ class TwilioMediaStreamsInputChannel(VoiceInputChannel):
142
143
  def blueprint(
143
144
  self, on_new_message: Callable[[UserMessage], Awaitable[Any]]
144
145
  ) -> Blueprint:
145
- """Defines a Sanic bluelogger.debug."""
146
+ """Defines a Sanic blueprint for the voice input channel."""
146
147
  blueprint = Blueprint("twilio_media_streams", __name__)
147
148
 
148
149
  @blueprint.route("/", methods=["GET"])
@@ -315,6 +315,7 @@ class VoiceInputChannel(InputChannel):
315
315
  def map_input_message(
316
316
  self,
317
317
  message: Any,
318
+ ws: Websocket,
318
319
  ) -> VoiceChannelAction:
319
320
  """Map a channel input message to a voice channel action."""
320
321
  raise NotImplementedError
@@ -340,7 +341,7 @@ class VoiceInputChannel(InputChannel):
340
341
  async def consume_audio_bytes() -> None:
341
342
  async for message in channel_websocket:
342
343
  is_bot_speaking_before = call_state.is_bot_speaking
343
- channel_action = self.map_input_message(message)
344
+ channel_action = self.map_input_message(message, channel_websocket)
344
345
  is_bot_speaking_after = call_state.is_bot_speaking
345
346
 
346
347
  if not is_bot_speaking_before and is_bot_speaking_after:
@@ -54,7 +54,9 @@ from rasa.dialogue_understanding.stack.utils import (
54
54
  user_flows_on_the_stack,
55
55
  )
56
56
  from rasa.shared.constants import RASA_PATTERN_HUMAN_HANDOFF
57
- from rasa.shared.core.constants import ACTION_LISTEN_NAME, SlotMappingType
57
+ from rasa.shared.core.constants import (
58
+ ACTION_LISTEN_NAME,
59
+ )
58
60
  from rasa.shared.core.events import (
59
61
  Event,
60
62
  FlowCompleted,
@@ -564,38 +566,6 @@ def cancel_flow_and_push_internal_error(stack: DialogueStack, flow_name: str) ->
564
566
  stack.push(InternalErrorPatternFlowStackFrame())
565
567
 
566
568
 
567
- def validate_custom_slot_mappings(
568
- step: CollectInformationFlowStep,
569
- stack: DialogueStack,
570
- tracker: DialogueStateTracker,
571
- available_actions: List[str],
572
- flow_name: str,
573
- ) -> bool:
574
- """Validate a slot with custom mappings.
575
-
576
- If invalid, trigger pattern_internal_error and return False.
577
- """
578
- slot = tracker.slots.get(step.collect, None)
579
- slot_mappings = slot.mappings if slot else []
580
- for mapping in slot_mappings:
581
- if (
582
- mapping.get("type") == SlotMappingType.CUSTOM.value
583
- and mapping.get("action") is None
584
- ):
585
- # this is a slot that must be filled by a custom action
586
- # check if collect_action exists
587
- if step.collect_action not in available_actions:
588
- structlogger.error(
589
- "flow.step.run.collect_action_not_found_for_custom_slot_mapping",
590
- action=step.collect_action,
591
- collect=step.collect,
592
- )
593
- cancel_flow_and_push_internal_error(stack, flow_name)
594
- return False
595
-
596
- return True
597
-
598
-
599
569
  def attach_stack_metadata_to_events(
600
570
  step_id: str,
601
571
  flow_id: str,
@@ -792,14 +762,6 @@ def _run_collect_information_step(
792
762
  # if we return any other FlowStepResult, the assistant will stay silent
793
763
  # instead of triggering the internal error pattern
794
764
  return ContinueFlowWithNextStep(events=initial_events)
795
- is_mapping_valid = validate_custom_slot_mappings(
796
- step, stack, tracker, available_actions, flow_name
797
- )
798
-
799
- if not is_mapping_valid:
800
- # if we return any other FlowStepResult, the assistant will stay silent
801
- # instead of triggering the internal error pattern
802
- return ContinueFlowWithNextStep(events=initial_events)
803
765
 
804
766
  structlogger.debug("flow.step.run.collect")
805
767
  trigger_pattern_ask_collect_information(
rasa/core/run.py CHANGED
@@ -283,9 +283,10 @@ def serve_application(
283
283
  endpoints.lock_store if endpoints else None
284
284
  )
285
285
 
286
- telemetry.track_server_start(
287
- input_channels, endpoints, model_path, number_of_workers, enable_api
288
- )
286
+ if not inspect:
287
+ telemetry.track_server_start(
288
+ input_channels, endpoints, model_path, number_of_workers, enable_api
289
+ )
289
290
 
290
291
  rasa.utils.common.update_sanic_log_level(
291
292
  log_file, use_syslog, syslog_address, syslog_port, syslog_protocol