rasa-pro 3.13.0.dev20250613__py3-none-any.whl → 3.13.0rc2__py3-none-any.whl

rasa/core/channels/voice_stream/genesys.py

@@ -29,6 +29,7 @@ from rasa.core.channels.voice_stream.voice_channel import (
     VoiceInputChannel,
     VoiceOutputChannel,
 )
+from rasa.shared.exceptions import InvalidConfigException
 
 """
 Genesys throws a rate limit error with too many audio messages.
@@ -91,9 +92,14 @@ class GenesysInputChannel(VoiceInputChannel):
         return "genesys"
 
     def __init__(
-        self, api_key: Text, client_secret: Optional[Text], *args: Any, **kwargs: Any
+        self,
+        server_url: str,
+        asr_config: Dict,
+        tts_config: Dict,
+        api_key: Optional[Text] = None,
+        client_secret: Optional[Text] = None,
     ) -> None:
-        super().__init__(*args, **kwargs)
+        super().__init__(server_url, asr_config, tts_config)
         self.api_key = api_key
         self.client_secret = client_secret
 
@@ -101,20 +107,33 @@ class GenesysInputChannel(VoiceInputChannel):
     def from_credentials(
         cls,
         credentials: Optional[Dict[str, Any]],
-    ) -> VoiceInputChannel:
-        if not credentials:
-            raise ValueError("No credentials given for Genesys voice channel.")
-
-        if not credentials.get("api_key"):
-            raise ValueError("No API key given for Genesys voice channel (api_key).")
-
-        return cls(
-            api_key=credentials["api_key"],
-            client_secret=credentials.get("client_secret"),
-            server_url=credentials["server_url"],
-            asr_config=credentials["asr"],
-            tts_config=credentials["tts"],
-        )
+    ) -> "GenesysInputChannel":
+        """Create a channel from credentials dictionary.
+
+        Args:
+            credentials: Dictionary containing the required credentials:
+                - server_url: URL where the server is hosted
+                - asr: ASR engine configuration
+                - tts: TTS engine configuration
+                - api_key: Required API key for Genesys authentication
+                - client_secret: Optional client secret for signature verification
+
+        Returns:
+            GenesysInputChannel instance
+        """
+        channel = super().from_credentials(credentials)
+
+        # Check required Genesys-specific credentials
+        if not credentials.get("api_key"):  # type: ignore[union-attr]
+            raise InvalidConfigException(
+                "No API key given for Genesys voice channel (api_key)."
+            )
+
+        # Update channel with Genesys-specific credentials
+        channel.api_key = credentials["api_key"]  # type: ignore[index,attr-defined]
+        channel.client_secret = credentials.get("client_secret")  # type: ignore[union-attr,attr-defined]
+
+        return channel  # type: ignore[return-value]
 
     def _ensure_channel_data_initialized(self) -> None:
         """Initialize Genesys-specific channel data if not already present.

rasa/core/channels/voice_stream/jambonz.py

@@ -0,0 +1,232 @@
+import audioop
+import json
+import uuid
+from typing import Any, Awaitable, Callable, Dict, Optional, Text, Tuple
+
+import structlog
+from sanic import (  # type: ignore[attr-defined]
+    Blueprint,
+    HTTPResponse,
+    Request,
+    Websocket,
+    response,
+)
+
+from rasa.core.channels import UserMessage, requires_basic_auth
+from rasa.core.channels.voice_ready.utils import (
+    CallParameters,
+    validate_username_password_credentials,
+)
+from rasa.core.channels.voice_stream.audio_bytes import RasaAudioBytes
+from rasa.core.channels.voice_stream.call_state import call_state
+from rasa.core.channels.voice_stream.tts.tts_engine import TTSEngine
+from rasa.core.channels.voice_stream.voice_channel import (
+    ContinueConversationAction,
+    EndConversationAction,
+    NewAudioAction,
+    VoiceChannelAction,
+    VoiceInputChannel,
+    VoiceOutputChannel,
+)
+
+logger = structlog.get_logger()
+
+JAMBONZ_STREAMS_WEBSOCKET_PATH = "webhooks/jambonz_streams/websocket"
+
+
+def map_call_params(data: Dict[Text, str]) -> CallParameters:
+    """Map the twilio stream parameters to the CallParameters dataclass."""
+    call_sid = data.get("callSid", "None")
+    from_number = data.get("from", "Unknown")
+    to_number = data.get("to")
+    return CallParameters(
+        call_id=call_sid,
+        user_phone=from_number,
+        bot_phone=to_number,
+        stream_id=call_sid,
+    )
+
+
+class JambonzStreamOutputChannel(VoiceOutputChannel):
+    @classmethod
+    def name(cls) -> str:
+        return "jambonz_stream"
+
+    async def send_audio_bytes(
+        self, recipient_id: str, audio_bytes: RasaAudioBytes
+    ) -> None:
+        """Overridden to send binary websocket messages for Jambonz.
+
+        Converts 8kHz μ-law to 8kHz L16 PCM for Jambonz streaming.
+        """
+        pcm = audioop.ulaw2lin(audio_bytes, 2)
+        await self.voice_websocket.send(pcm)
+
+    def create_marker_message(self, recipient_id: str) -> Tuple[str, str]:
+        """Create a marker message to track audio stream position."""
+        marker_id = uuid.uuid4().hex
+        return json.dumps({"type": "mark", "data": {"name": marker_id}}), marker_id
+
+
+class JambonzStreamInputChannel(VoiceInputChannel):
+    @classmethod
+    def name(cls) -> str:
+        return "jambonz_stream"
+
+    def __init__(
+        self,
+        server_url: str,
+        asr_config: Dict,
+        tts_config: Dict,
+        username: Optional[Text] = None,
+        password: Optional[Text] = None,
+    ) -> None:
+        """Initialize the channel.
+
+        Args:
+            username: Optional username for basic auth
+            password: Optional password for basic auth
+        """
+        super().__init__(server_url, asr_config, tts_config)
+        self.username = username
+        self.password = password
+
+    @classmethod
+    def from_credentials(
+        cls, credentials: Optional[Dict[Text, Any]]
+    ) -> "JambonzStreamInputChannel":
+        """Create a channel from credentials dictionary.
+
+        Args:
+            credentials: Dictionary containing the required credentials:
+                - server_url: URL where the server is hosted
+                - asr: ASR engine configuration
+                - tts: TTS engine configuration
+                - username: Optional username for basic auth
+                - password: Optional password for basic auth
+
+        Returns:
+            JambonzStreamInputChannel instance
+        """
+        # Get common credentials from parent
+        channel = super().from_credentials(credentials)
+
+        # Check optional basic auth credentials
+        username = credentials.get("username")  # type: ignore[union-attr]
+        password = credentials.get("password")  # type: ignore[union-attr]
+        validate_username_password_credentials(username, password, "Jambonz Stream")
+
+        # Update channel with auth credentials
+        channel.username = username  # type: ignore[attr-defined]
+        channel.password = password  # type: ignore[attr-defined]
+
+        return channel  # type: ignore[return-value]
+
+    def _websocket_stream_url(self) -> str:
+        """Returns the websocket stream URL."""
+        # depending on the config value, the url might contain http as a
+        # protocol or not - we'll make sure both work
+        if self.server_url.startswith("http"):
+            base_url = self.server_url.replace("http", "ws")
+        else:
+            base_url = f"wss://{self.server_url}"
+        return f"{base_url}/{JAMBONZ_STREAMS_WEBSOCKET_PATH}"
+
+    def channel_bytes_to_rasa_audio_bytes(self, input_bytes: bytes) -> RasaAudioBytes:
+        """Convert Jambonz audio bytes (L16 PCM) to Rasa audio bytes (μ-law)."""
+        ulaw = audioop.lin2ulaw(input_bytes, 2)
+        return RasaAudioBytes(ulaw)
+
+    async def collect_call_parameters(
+        self, channel_websocket: Websocket
+    ) -> Optional[CallParameters]:
+        # Wait for initial metadata message
+        message = await channel_websocket.recv()
+        logger.debug("jambonz.collect_call_parameters", message=message)
+        metadata = json.loads(message)
+        return map_call_params(metadata)
+
+    def map_input_message(self, message: Any, ws: Websocket) -> VoiceChannelAction:
+        # Handle binary audio frames
+        if isinstance(message, bytes):
+            channel_bytes = message
+            audio_bytes = self.channel_bytes_to_rasa_audio_bytes(channel_bytes)
+            return NewAudioAction(audio_bytes)
+
+        # Handle JSON messages
+        data = json.loads(message)
+        if data["type"] == "mark":
+            if data["data"]["name"] == call_state.latest_bot_audio_id:
+                # Just finished streaming last audio bytes
+                call_state.is_bot_speaking = False  # type: ignore[attr-defined]
+                if call_state.should_hangup:
+                    logger.debug(
+                        "jambonz.hangup", marker=call_state.latest_bot_audio_id
+                    )
+                    return EndConversationAction()
+            else:
+                call_state.is_bot_speaking = True  # type: ignore[attr-defined]
+        elif data["event"] == "dtmf":
+            # TODO: handle DTMF input
+            logger.debug("jambonz.dtmf.received", dtmf=data["dtmf"])
+        else:
+            logger.warning("jambonz.unexpected_message", message=data)
+
+        return ContinueConversationAction()
+
+    def create_output_channel(
+        self, voice_websocket: Websocket, tts_engine: TTSEngine
+    ) -> VoiceOutputChannel:
+        return JambonzStreamOutputChannel(
+            voice_websocket,
+            tts_engine,
+            self.tts_cache,
+        )
+
+    def blueprint(
+        self, on_new_message: Callable[[UserMessage], Awaitable[Any]]
+    ) -> Blueprint:
+        blueprint = Blueprint("jambonz_stream", __name__)
+
+        @blueprint.route("/", methods=["GET"])
+        async def health(_: Request) -> HTTPResponse:
+            return response.json({"status": "ok"})
+
+        @blueprint.route("/call_status", methods=["POST"])
+        @requires_basic_auth(self.username, self.password)
+        async def call_status(request: Request) -> HTTPResponse:
+            """Handle call status updates from Jambonz."""
+            data = request.json
+            logger.debug("jambonz.call_status.received", data=data)
+            return response.json({"status": "ok"})
+
+        @blueprint.route("/webhook", methods=["POST"])
+        @requires_basic_auth(self.username, self.password)
+        async def webhook(request: Request) -> HTTPResponse:
+            """Handle incoming webhook requests from Jambonz."""
+            data = request.json
+            logger.debug("jambonz.webhook.received", data=data)
+            return response.json(
+                [
+                    {
+                        "verb": "listen",
+                        "url": self._websocket_stream_url(),
+                        "sampleRate": 8000,
+                        "passDtmf": True,
+                        "bidirectionalAudio": {
+                            "enabled": True,
+                            "streaming": True,
+                            "sampleRate": 8000,
+                        },
+                    }
+                ]
+            )
+
+        @blueprint.websocket("/websocket", subprotocols=["audio.jambonz.org"])  # type: ignore[misc]
+        async def handle_message(request: Request, ws: Websocket) -> None:
+            try:
+                await self.run_audio_streaming(on_new_message, ws)
+            except Exception as e:
+                logger.error("jambonz.handle_message.error", error=e)
+
+        return blueprint

rasa/core/channels/voice_stream/tts/__init__.py

@@ -0,0 +1,8 @@
+from rasa.core.channels.voice_stream.tts.tts_cache import TTSCache
+from rasa.core.channels.voice_stream.tts.tts_engine import (
+    TTSEngine,
+    TTSEngineConfig,
+    TTSError,
+)
+
+__all__ = ["TTSEngine", "TTSEngineConfig", "TTSError", "TTSCache"]

rasa/core/channels/voice_stream/twilio_media_streams.py

@@ -19,7 +19,10 @@ from rasa.core.channels.channel import (
     create_auth_requested_response_provider,
     requires_basic_auth,
 )
-from rasa.core.channels.voice_ready.utils import CallParameters
+from rasa.core.channels.voice_ready.utils import (
+    CallParameters,
+    validate_username_password_credentials,
+)
 from rasa.core.channels.voice_stream.audio_bytes import RasaAudioBytes
 from rasa.core.channels.voice_stream.call_state import call_state
 from rasa.core.channels.voice_stream.tts.tts_engine import TTSEngine
@@ -31,7 +34,6 @@ from rasa.core.channels.voice_stream.voice_channel import (
     VoiceInputChannel,
     VoiceOutputChannel,
 )
-from rasa.shared.exceptions import RasaException
 
 if TYPE_CHECKING:
     from twilio.twiml.voice_response import VoiceResponse
@@ -122,11 +124,7 @@ class TwilioMediaStreamsInputChannel(VoiceInputChannel):
 
         username = credentials.get("username")
         password = credentials.get("password")
-        if (username is None) != (password is None):
-            raise RasaException(
-                "In TwilioMediaStreams channel, either both username and password "
-                "or neither should be provided. "
-            )
+        validate_username_password_credentials(username, password, "TwilioMediaStreams")
 
         return cls(
             credentials["server_url"],
@@ -140,6 +138,13 @@ class TwilioMediaStreamsInputChannel(VoiceInputChannel):
     def name(cls) -> str:
         return "twilio_media_streams"
 
+    def get_sender_id(self, call_parameters: CallParameters) -> str:
+        """Get the sender ID for the channel.
+
+        Twilio Media Streams uses the Stream ID as Sender ID because
+        it is required in OutputChannel.send_text_message to send messages."""
+        return call_parameters.stream_id  # type: ignore[return-value]
+
     def channel_bytes_to_rasa_audio_bytes(self, input_bytes: bytes) -> RasaAudioBytes:
         return RasaAudioBytes(base64.b64decode(input_bytes))
 

rasa/core/channels/voice_stream/voice_channel.py

@@ -35,7 +35,7 @@ from rasa.core.channels.voice_stream.util import (
     generate_silence,
 )
 from rasa.shared.core.constants import SILENCE_TIMEOUT_SLOT
-from rasa.shared.utils.cli import print_error_and_exit
+from rasa.shared.exceptions import InvalidConfigException
 from rasa.shared.utils.common import (
     class_from_module_path,
     mark_as_beta_feature,
@@ -71,6 +71,14 @@ class ContinueConversationAction(VoiceChannelAction):
 
 
 def asr_engine_from_config(asr_config: Dict) -> ASREngine:
+    if not asr_config:
+        raise ValueError("ASR configuration dictionary cannot be empty")
+
+    if "name" not in asr_config:
+        raise ValueError(
+            "ASR configuration must contain 'name' key specifying the engine type"
+        )
+
     name = str(asr_config["name"])
     asr_config = copy.copy(asr_config)
     asr_config.pop("name")
@@ -84,12 +92,12 @@ def asr_engine_from_config(asr_config: Dict) -> ASREngine:
             asr_engine_class = class_from_module_path(name)
             return asr_engine_class.from_config_dict(asr_config)
         except NameError:
-            print_error_and_exit(
+            raise InvalidConfigException(
                 f"Failed to initialize ASR Engine with type '{name}'. "
                 f"Please make sure the method `from_config_dict`is implemented."
             )
         except TypeError as e:
-            print_error_and_exit(
+            raise InvalidConfigException(
                 f"Failed to initialize ASR Engine with type '{name}'. "
                 f"Invalid configuration provided. "
                 f"Error: {e}"
@@ -97,6 +105,14 @@ def asr_engine_from_config(asr_config: Dict) -> ASREngine:
 
 
 def tts_engine_from_config(tts_config: Dict) -> TTSEngine:
+    if not tts_config:
+        raise ValueError("TTS configuration dictionary cannot be empty")
+
+    if "name" not in tts_config:
+        raise ValueError(
+            "TTS configuration must contain 'name' key specifying the engine type"
+        )
+
     name = str(tts_config["name"])
     tts_config = copy.copy(tts_config)
     tts_config.pop("name")
@@ -110,13 +126,13 @@ def tts_engine_from_config(tts_config: Dict) -> TTSEngine:
             tts_engine_class = class_from_module_path(name)
             return tts_engine_class.from_config_dict(tts_config)
         except NameError:
-            print_error_and_exit(
+            raise InvalidConfigException(
                 f"Failed to initialize TTS Engine with type '{name}'. "
                 f"Please make sure the method `from_config_dict`is implemented."
             )
         except TypeError as e:
-            print_error_and_exit(
-                f"Failed to initialize ASR Engine with type '{name}'. "
+            raise InvalidConfigException(
+                f"Failed to initialize TTS Engine with type '{name}'. "
                 f"Invalid configuration provided. "
                 f"Error: {e}"
             )
@@ -286,13 +302,18 @@ class VoiceOutputChannel(OutputChannel):
 
 
 class VoiceInputChannel(InputChannel):
+    # All children of this class require a voice license to be used.
+    requires_voice_license = True
+
     def __init__(
         self,
         server_url: str,
         asr_config: Dict,
         tts_config: Dict,
     ):
-        validate_voice_license_scope()
+        if self.requires_voice_license:
+            validate_voice_license_scope()
+
         self.server_url = server_url
         self.asr_config = asr_config
         self.tts_config = tts_config
@@ -305,6 +326,10 @@ class VoiceInputChannel(InputChannel):
             tts_config=self.tts_config,
         )
 
+    def get_sender_id(self, call_parameters: CallParameters) -> str:
+        """Get the sender ID for the channel."""
+        return call_parameters.call_id
+
     async def monitor_silence_timeout(self, asr_event_queue: asyncio.Queue) -> None:
         timeout = call_state.silence_timeout
         if not timeout:
@@ -327,11 +352,24 @@ class VoiceInputChannel(InputChannel):
         cls,
         credentials: Optional[Dict[str, Any]],
     ) -> InputChannel:
-        credentials = credentials or {}
+        if not credentials:
+            cls.raise_missing_credentials_exception()
+
+        if not credentials.get("server_url"):
+            raise InvalidConfigException("No server_url provided in credentials.")
+        if not credentials.get("asr"):
+            raise InvalidConfigException(
+                "No ASR configuration provided in credentials."
+            )
+        if not credentials.get("tts"):
+            raise InvalidConfigException(
+                "No TTS configuration provided in credentials."
+            )
+
         return cls(
-            credentials["server_url"],
-            credentials["asr"],
-            credentials["tts"],
+            server_url=credentials["server_url"],
+            asr_config=credentials["asr"],
+            tts_config=credentials["tts"],
         )
 
     def channel_bytes_to_rasa_audio_bytes(self, input_bytes: bytes) -> RasaAudioBytes:
@@ -353,7 +391,7 @@ class VoiceInputChannel(InputChannel):
         message = UserMessage(
             text=USER_CONVERSATION_SESSION_START,
             output_channel=output_channel,
-            sender_id=call_parameters.stream_id,
+            sender_id=self.get_sender_id(call_parameters),
             input_channel=self.name(),
             metadata=asdict(call_parameters),
         )
@@ -471,7 +509,7 @@ class VoiceInputChannel(InputChannel):
             message = UserMessage(
                 text=e.text,
                 output_channel=output_channel,
-                sender_id=call_parameters.stream_id,
+                sender_id=self.get_sender_id(call_parameters),
                 input_channel=self.name(),
                 metadata=asdict(call_parameters),
             )
@@ -484,7 +522,7 @@ class VoiceInputChannel(InputChannel):
             message = UserMessage(
                 text=USER_CONVERSATION_SILENCE_TIMEOUT,
                 output_channel=output_channel,
-                sender_id=call_parameters.stream_id,
+                sender_id=self.get_sender_id(call_parameters),
                 input_channel=self.name(),
                 metadata=asdict(call_parameters),
             )
@@ -502,7 +540,7 @@ class VoiceInputChannel(InputChannel):
         message = UserMessage(
             text=USER_CONVERSATION_SESSION_END,
             output_channel=output_channel,
-            sender_id=call_parameters.stream_id,
+            sender_id=self.get_sender_id(call_parameters),
             input_channel=self.name(),
         )
         await on_new_message(message)

rasa/core/exporter.py

@@ -16,6 +16,11 @@ from rasa.exceptions import (
     NoEventsToMigrateError,
     PublishingError,
 )
+from rasa.shared.core.events import (
+    BotUttered,
+    SlotSet,
+    UserUttered,
+)
 from rasa.shared.core.trackers import EventVerbosity
 
 logger = logging.getLogger(__name__)
@@ -43,6 +48,7 @@ class Exporter:
         tracker_store: TrackerStore,
         event_broker: EventBroker,
         endpoints_path: Text,
+        is_pii_enabled: bool = False,
         requested_conversation_ids: Optional[Text] = None,
         minimum_timestamp: Optional[float] = None,
         maximum_timestamp: Optional[float] = None,
@@ -52,6 +58,7 @@ class Exporter:
         self.tracker_store = tracker_store
 
         self.event_broker = event_broker
+        self.is_pii_enabled = is_pii_enabled
         self.requested_conversation_ids = requested_conversation_ids
         self.minimum_timestamp = minimum_timestamp
         self.maximum_timestamp = maximum_timestamp
@@ -72,10 +79,12 @@ class Exporter:
         current_timestamp = None
 
         headers = self._get_message_headers()
+        warned_sender_ids: Set[Text] = set()
 
         async for event in self._fetch_events_within_time_range():
             # noinspection PyBroadException
             try:
+                self._check_anonymization_status(event, warned_sender_ids)
                 self._publish_with_message_headers(event, headers)
                 published_events += 1
                 current_timestamp = event["timestamp"]
@@ -282,3 +291,30 @@ class Exporter:
             events_with_conversation_id.append(event)
 
         return events_with_conversation_id
+
+    def _check_anonymization_status(
+        self, event: Dict[Text, Any], warned_sender_ids: Set[Text]
+    ) -> None:
+        """Check if the tracker store contains unanonymized events.
+
+        If it does, print a warning that these events will be published as is.
+
+        Args:
+            event: The event to check for anonymization status
+            warned_sender_ids: Set of sender IDs that have already been warned about
+        """
+        sender_id = event["sender_id"]
+        if (
+            self.is_pii_enabled
+            and sender_id not in warned_sender_ids
+            and event["event"]
+            in (UserUttered.type_name, BotUttered.type_name, SlotSet.type_name)
+            and not event.get("anonymized_at", None)
+        ):
+            rasa.shared.utils.cli.print_warning(
+                f"Retrieved un-anonymized event for sender_id {sender_id}. "
+                f"All events after this timestamp {event['timestamp']} "
+                "are not anonymized for this tracker. Proceeding with "
+                "publishing plaintext values for all events following this.",
+            )
+            warned_sender_ids.add(sender_id)

rasa/core/information_retrieval/faiss.py

@@ -12,6 +12,7 @@ from rasa.core.information_retrieval import (
     InformationRetrievalException,
     SearchResultList,
 )
+from rasa.core.information_retrieval.ingestion.faq_parser import _format_faq_documents
 from rasa.utils.endpoints import EndpointConfig
 from rasa.utils.ml_utils import persist_faiss_vector_store
 
@@ -31,10 +32,12 @@ class FAISS_Store(InformationRetrieval):
         index_path: str,
         docs_folder: Optional[str],
         create_index: Optional[bool] = False,
+        parse_as_faq_pairs: Optional[bool] = False,
     ):
         """Initializes the FAISS Store."""
         self.chunk_size = 1000
         self.chunk_overlap = 20
+        self.parse_as_faq_pairs = parse_as_faq_pairs
 
         path = Path(index_path) / "documents_faiss"
         if create_index:
@@ -86,21 +89,25 @@ class FAISS_Store(InformationRetrieval):
         if not docs_folder:
             raise ValueError("parameter `docs_folder` needs to be specified")
 
-        docs = self.load_documents(docs_folder)
-        splitter = RecursiveCharacterTextSplitter(
-            chunk_size=self.chunk_size,
-            chunk_overlap=self.chunk_overlap,
-            length_function=len,
-        )
-        doc_chunks = splitter.split_documents(docs)
+        documents = self.load_documents(docs_folder)
+
+        if not self.parse_as_faq_pairs:
+            splitter = RecursiveCharacterTextSplitter(
+                chunk_size=self.chunk_size,
+                chunk_overlap=self.chunk_overlap,
+                length_function=len,
+            )
+            parsed_documents = splitter.split_documents(documents)
+        else:
+            parsed_documents = _format_faq_documents(documents)
 
         logger.info(
             "information_retrieval.faiss_store._create_document_index",
-            len_chunks=len(doc_chunks),
+            len_chunks=len(parsed_documents),
         )
-        if doc_chunks:
-            texts = [chunk.page_content for chunk in doc_chunks]
-            metadatas = [chunk.metadata for chunk in doc_chunks]
+        if parsed_documents:
+            texts = [document.page_content for document in parsed_documents]
+            metadatas = [document.metadata for document in parsed_documents]
             return FAISS.from_texts(texts, embedding, metadatas=metadatas, ids=None)
         else:
             raise ValueError(f"No documents found at '{docs_folder}'.")