rasa-pro 3.12.4__tar.gz → 3.12.6.dev1__tar.gz

{rasa_pro-3.12.4 → rasa_pro-3.12.6.dev1}/PKG-INFO

@@ -1,8 +1,7 @@
-Metadata-Version: 2.1
+Metadata-Version: 2.3
 Name: rasa-pro
-Version: 3.12.4
+Version: 3.12.6.dev1
 Summary: State-of-the-art open-core Conversational AI framework for Enterprises that natively leverages generative AI for effortless assistant development.
-Home-page: https://rasa.com
 Keywords: nlp,machine-learning,machine-learning-library,bot,bots,botkit,rasa conversational-agents,conversational-ai,chatbot,chatbot-framework,bot-framework
 Author: Rasa Technologies GmbH
 Author-email: hi@rasa.com
@@ -86,7 +85,6 @@ Requires-Dist: protobuf (>=4.23.3,<4.25.4)
 Requires-Dist: psutil (>=5.9.5,<6.0.0)
 Requires-Dist: psycopg2-binary (>=2.9.9,<2.10.0)
 Requires-Dist: pycountry (>=22.3.5,<23.0.0)
-Requires-Dist: pydantic (>=2.0,<3.0)
 Requires-Dist: pydot (>=1.4,<1.5)
 Requires-Dist: pykwalify (>=1.8,<1.9)
 Requires-Dist: pymilvus (>=2.4.1,<2.4.2)
@@ -118,7 +116,6 @@ Requires-Dist: scikit-learn (>=1.5.1,<1.6.0)
 Requires-Dist: scipy (>=1.13.1,<1.14.0)
 Requires-Dist: sentencepiece[sentencepiece] (>=0.1.99,<0.2.0) ; extra == "transformers" or extra == "full"
 Requires-Dist: sentry-sdk (>=1.14.0,<1.15.0)
-Requires-Dist: setuptools (>=70.0.0,<70.1.0)
 Requires-Dist: sklearn-crfsuite (>=0.3.6,<0.4.0)
 Requires-Dist: skops (>=0.10.0,<0.11.0)
 Requires-Dist: slack-sdk (>=3.27.1,<3.28.0)
@@ -135,7 +132,7 @@ Requires-Dist: tensorflow-io-gcs-filesystem (==0.34) ; sys_platform == "darwin"
 Requires-Dist: tensorflow-io-gcs-filesystem (==0.34) ; sys_platform == "linux"
 Requires-Dist: tensorflow-macos (==2.14.1) ; sys_platform == "darwin" and platform_machine == "arm64"
 Requires-Dist: tensorflow-metal (==1.1.0) ; (sys_platform == "darwin" and platform_machine == "arm64") and (extra == "metal")
-Requires-Dist: tensorflow-text (==2.14.0) ; sys_platform != "win32" and (platform_machine != "arm64" and platform_machine != "aarch64")
+Requires-Dist: tensorflow-text (==2.14.0) ; sys_platform != "win32" and platform_machine != "arm64" and platform_machine != "aarch64"
 Requires-Dist: tensorflow_hub (>=0.13.0,<0.14.0)
 Requires-Dist: terminaltables (>=3.1.10,<3.2.0)
 Requires-Dist: tiktoken (>=0.7.0,<0.8.0)
@@ -148,8 +145,8 @@ Requires-Dist: typing-utils (>=0.1.0,<0.2.0)
 Requires-Dist: ujson (>=5.8,<6.0)
 Requires-Dist: webexteamssdk (>=1.6.1,<1.7.0)
 Requires-Dist: websockets (>=10.4,<11.0)
-Requires-Dist: wheel (>=0.40.0)
 Project-URL: Documentation, https://rasa.com/docs
+Project-URL: Homepage, https://rasa.com
 Project-URL: Repository, https://github.com/rasahq/rasa
 Description-Content-Type: text/markdown
 

{rasa_pro-3.12.4 → rasa_pro-3.12.6.dev1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
 
 [tool.poetry]
 name = "rasa-pro"
-version = "3.12.4"
+version = "3.12.6.dev1"
 description = "State-of-the-art open-core Conversational AI framework for Enterprises that natively leverages generative AI for effortless assistant development."
 authors = [ "Rasa Technologies GmbH <hi@rasa.com>",]
 maintainers = [ "Tom Bocklisch <tom@rasa.com>",]
@@ -132,7 +132,7 @@ psycopg2-binary = ">=2.9.9,<2.10.0"
 python-dateutil = "~2.8.2"
 protobuf = ">=4.23.3,< 4.25.4"
 tensorflow_hub = "^0.13.0"
-setuptools = "~70.0.0"
+#setuptools = "~70.0.0" #test if this is needed
 ujson = ">=5.8,<6.0"
 regex = "~2024.7.24"
 sentry-sdk = ">=1.14.0,<1.15.0"
@@ -156,7 +156,7 @@ langchain-community = "~0.2.19"
 jinja2 = "~3.1.6"
 openai = "~1.55.3"
 dnspython = "2.6.1"
-wheel = ">=0.40.0"
+#wheel = ">=0.40.0" #test if this is needed
 certifi = ">=2024.07.04"
 cryptography = ">=44.0.1"
 jsonpatch = "^1.33"
@@ -319,7 +319,7 @@ optional = true
 [tool.poetry.dependencies.pymongo]
 version = "~4.10.1"
 
-[tool.poetry.dev-dependencies.pytest-sanic]
+[tool.poetry.group.dev.dependencies.pytest-sanic]
 git = "https://github.com/RasaHQ/pytest-sanic"
 branch = "fix-variable-setting-on-sanic-instance-issue"
 

{rasa_pro-3.12.4 → rasa_pro-3.12.6.dev1}/rasa/core/channels/voice_ready/audiocodes.py

@@ -1,5 +1,6 @@
 import asyncio
 import copy
+import hmac
 import json
 import uuid
 from collections import defaultdict
@@ -245,8 +246,13 @@ class AudiocodesInput(InputChannel):
 
     def _check_token(self, token: Optional[Text]) -> None:
         if not token:
+            structlogger.error("audiocodes.token_not_provided")
             raise HttpUnauthorized("Authentication token required.")
 
+        if not hmac.compare_digest(str(token), str(self.token)):
+            structlogger.error("audiocodes.invalid_token", invalid_token=token)
+            raise HttpUnauthorized("Invalid authentication token.")
+
     def _get_conversation(
         self, token: Optional[Text], conversation_id: Text
     ) -> Conversation:

{rasa_pro-3.12.4 → rasa_pro-3.12.6.dev1}/rasa/core/channels/voice_stream/audiocodes.py

@@ -1,5 +1,6 @@
 import asyncio
 import base64
+import hmac
 import json
 from typing import Any, Awaitable, Callable, Dict, Optional, Text
 
@@ -103,6 +104,7 @@ class AudiocodesVoiceInputChannel(VoiceInputChannel):
 
     def __init__(
         self,
+        token: Optional[Text],
         server_url: str,
         asr_config: Dict,
         tts_config: Dict,
@@ -110,6 +112,22 @@ class AudiocodesVoiceInputChannel(VoiceInputChannel):
     ):
         mark_as_beta_feature("Audiocodes (audiocodes_stream) Channel")
         super().__init__(server_url, asr_config, tts_config, monitor_silence)
+        self.token = token
+
+    @classmethod
+    def from_credentials(
+        cls, credentials: Optional[Dict[str, Any]]
+    ) -> VoiceInputChannel:
+        if not credentials:
+            raise ValueError("No credentials given for Audiocodes voice channel.")
+
+        return cls(
+            token=credentials.get("token"),
+            server_url=credentials["server_url"],
+            asr_config=credentials["asr"],
+            tts_config=credentials["tts"],
+            monitor_silence=credentials.get("monitor_silence", False),
+        )
 
     def channel_bytes_to_rasa_audio_bytes(self, input_bytes: bytes) -> RasaAudioBytes:
         return RasaAudioBytes(base64.b64decode(input_bytes))
@@ -135,6 +153,13 @@ class AudiocodesVoiceInputChannel(VoiceInputChannel):
                     )
                     if activity["name"] == "start":
                         return map_call_params(activity["parameters"])
+            elif data["type"] == "connection.validate":
+                # not part of call flow; only sent when integration is created
+                logger.info(
+                    "audiocodes_stream.collect_call_parameters.connection.validate",
+                    event_info="received request to validate integration",
+                )
+                self._send_validated(channel_websocket, data)
             else:
                 logger.warning("audiocodes_stream.unknown_message", data=data)
         return None
@@ -158,7 +183,7 @@ class AudiocodesVoiceInputChannel(VoiceInputChannel):
                 elif activity["name"] == "playFinished":
                     logger.debug("audiocodes_stream.playFinished", data=activity)
                     if call_state.should_hangup:
-                        logger.info("audiocodes.hangup")
+                        logger.info("audiocodes_stream.hangup")
                         self._send_hangup(ws, data)
                         # the conversation should continue until
                         # we receive a end message from audiocodes
@@ -180,11 +205,10 @@ class AudiocodesVoiceInputChannel(VoiceInputChannel):
         elif data["type"] == "session.end":
             logger.debug("audiocodes_stream.end", data=data)
             return EndConversationAction()
-        elif data["type"] == "connection.validate":
-            # not part of call flow; only sent when integration is created
-            self._send_validated(ws, data)
         else:
-            logger.warning("audiocodes_stream.unknown_message", data=data)
+            logger.warning(
+                "audiocodes_stream.map_input_message.unknown_message", data=data
+            )
 
         return ContinueConversationAction()
 
@@ -254,6 +278,17 @@ class AudiocodesVoiceInputChannel(VoiceInputChannel):
             self.tts_cache,
         )
 
+    def _is_token_valid(self, token: Optional[Text]) -> bool:
+        # If no token is set, always return True
+        if not self.token:
+            return True
+
+        # Token is required, but not provided
+        if not token:
+            return False
+
+        return hmac.compare_digest(str(self.token), str(token))
+
     def blueprint(
         self, on_new_message: Callable[[UserMessage], Awaitable[Any]]
     ) -> Blueprint:
@@ -266,17 +301,26 @@ class AudiocodesVoiceInputChannel(VoiceInputChannel):
 
         @blueprint.websocket("/websocket")  # type: ignore
         async def receive(request: Request, ws: Websocket) -> None:
-            # TODO: validate API key header
-            logger.info("audiocodes.receive", message="Starting audio streaming")
+            if not self._is_token_valid(request.token):
+                logger.error(
+                    "audiocodes_stream.invalid_token",
+                    invalid_token=request.token,
+                )
+                await ws.close(code=1008, reason="Invalid token")
+                return
+
+            logger.info(
+                "audiocodes_stream.receive", event_info="Started websocket connection"
+            )
             try:
                 await self.run_audio_streaming(on_new_message, ws)
             except Exception as e:
                 logger.exception(
-                    "audiocodes.receive",
+                    "audiocodes_stream.receive",
                     message="Error during audio streaming",
                     error=e,
                 )
-                # return 500 error
+                await ws.close(code=1011, reason="Error during audio streaming")
                 raise
 
         return blueprint

{rasa_pro-3.12.4 → rasa_pro-3.12.6.dev1}/rasa/core/channels/voice_stream/genesys.py

@@ -1,4 +1,7 @@
 import asyncio
+import base64
+import hashlib
+import hmac
 import json
 from typing import Any, Awaitable, Callable, Dict, Optional, Text
 
@@ -45,6 +48,7 @@ in the documentation but observed in their example app
 https://github.com/GenesysCloudBlueprints/audioconnector-server-reference-implementation
 """
 MAXIMUM_BINARY_MESSAGE_SIZE = 64000  # 64KB
+HEADER_API_KEY = "X-Api-Key"
 logger = structlog.get_logger(__name__)
 
 
@@ -86,8 +90,31 @@ class GenesysInputChannel(VoiceInputChannel):
     def name(cls) -> str:
         return "genesys"
 
-    def __init__(self, *args: Any, **kwargs: Any) -> None:
+    def __init__(
+        self, api_key: Text, client_secret: Optional[Text], *args: Any, **kwargs: Any
+    ) -> None:
         super().__init__(*args, **kwargs)
+        self.api_key = api_key
+        self.client_secret = client_secret
+
+    @classmethod
+    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"],
+            monitor_silence=credentials.get("monitor_silence", False),
+        )
 
     def _ensure_channel_data_initialized(self) -> None:
         """Initialize Genesys-specific channel data if not already present.
@@ -273,6 +300,93 @@ class GenesysInputChannel(VoiceInputChannel):
         logger.debug("genesys.disconnect", message=message)
         _schedule_ws_task(ws.send(json.dumps(message)))
 
+    def _calculate_signature(self, request: Request) -> str:
+        """Calculate the signature using request data."""
+        org_id = request.headers.get("Audiohook-Organization-Id")
+        session_id = request.headers.get("Audiohook-Session-Id")
+        correlation_id = request.headers.get("Audiohook-Correlation-Id")
+        api_key = request.headers.get(HEADER_API_KEY)
+
+        # order of components is important!
+        components = [
+            ("@request-target", "/webhooks/genesys/websocket"),
+            ("audiohook-session-id", session_id),
+            ("audiohook-organization-id", org_id),
+            ("audiohook-correlation-id", correlation_id),
+            (HEADER_API_KEY.lower(), api_key),
+            ("@authority", self.server_url),
+        ]
+
+        # Create signature base string
+        signing_string = ""
+        for name, value in components:
+            signing_string += f'"{name}": {value}\n'
+
+        # Add @signature-params
+        signature_input = request.headers["Signature-Input"]
+        _, params_str = signature_input.split("=", 1)
+        signing_string += f'"@signature-params": {params_str}'
+
+        # Calculate the HMAC signature
+        key_bytes = base64.b64decode(self.client_secret)
+        signature = hmac.new(
+            key_bytes, signing_string.encode("utf-8"), hashlib.sha256
+        ).digest()
+        return base64.b64encode(signature).decode("utf-8")
+
+    async def _verify_signature(self, request: Request) -> bool:
+        """Verify the HTTP message signature from Genesys."""
+        if not self.client_secret:
+            logger.info(
+                "genesys.verify_signature.no_client_secret",
+                event_info="Signature verification skipped",
+            )
+            return True  # Skip verification if no client secret
+
+        signature = request.headers.get("Signature")
+        signature_input = request.headers.get("Signature-Input")
+        if not signature or not signature_input:
+            logger.error("genesys.signature.missing_signature_header")
+            return False
+
+        try:
+            actual_signature = signature.split("=", 1)[1].strip(':"')
+            expected_signature = self._calculate_signature(request)
+            return hmac.compare_digest(
+                expected_signature.encode("utf-8"), actual_signature.encode("utf-8")
+            )
+        except Exception as e:
+            logger.exception("genesys.signature.verification_error", error=e)
+            return False
+
+    def _ensure_required_headers(self, request: Request) -> bool:
+        """Ensure required headers are present in the request."""
+        required_headers = [
+            "Audiohook-Organization-Id",
+            "Audiohook-Correlation-Id",
+            "Audiohook-Session-Id",
+            HEADER_API_KEY,
+        ]
+
+        missing_headers = [
+            header for header in required_headers if header not in request.headers
+        ]
+
+        if missing_headers:
+            logger.error(
+                "genesys.missing_required_headers",
+                missing_headers=missing_headers,
+            )
+            return False
+        return True
+
+    def _ensure_api_key(self, request: Request) -> bool:
+        """Ensure the API key is present in the request."""
+        api_key = request.headers.get(HEADER_API_KEY)
+        if not hmac.compare_digest(str(self.api_key), str(api_key)):
+            return False
+        return True
+
     def blueprint(
         self, on_new_message: Callable[[UserMessage], Awaitable[Any]]
     ) -> Blueprint:
@@ -289,23 +403,39 @@ class GenesysInputChannel(VoiceInputChannel):
                 "genesys.receive",
                 audiohook_session_id=request.headers.get("audiohook-session-id"),
             )
-            # validate required headers
-            required_headers = [
-                "audiohook-organization-id",
-                "audiohook-correlation-id",
-                "audiohook-session-id",
-                "x-api-key",
-            ]
-
-            for header in required_headers:
-                if header not in request.headers:
-                    await ws.close(1008, f"Missing required header: {header}")
-                    return
-
-            # TODO: validate API key header
+
+            # verify signature
+            if not await self._verify_signature(request):
+                logger.error("genesys.receive.invalid_signature")
+                await ws.close(code=1008, reason="Invalid signature")
+                return
+
+            # ensure required headers are present
+            if not self._ensure_required_headers(request):
+                await ws.close(code=1002, reason="Missing required headers")
+                return
+
+            # ensure API key is correct
+            if not self._ensure_api_key(request):
+                logger.error(
+                    "genesys.receive.invalid_api_key",
+                    invalid_api_key=request.headers.get(HEADER_API_KEY),
+                )
+                await ws.close(code=1008, reason="Invalid API key")
+                return
+
             # process audio streaming
             logger.info("genesys.receive", message="Starting audio streaming")
-            await self.run_audio_streaming(on_new_message, ws)
+            try:
+                await self.run_audio_streaming(on_new_message, ws)
+            except Exception as e:
+                logger.exception(
+                    "genesys.receive",
+                    message="Error during audio streaming",
+                    error=e,
+                )
+                await ws.close(code=1011, reason="Error during audio streaming")
+                raise
 
         return blueprint
 

{rasa_pro-3.12.4 → rasa_pro-3.12.6.dev1}/rasa/core/policies/intentless_policy.py

@@ -1,7 +1,7 @@
 import importlib.resources
 import math
 from dataclasses import dataclass, field
-from typing import TYPE_CHECKING, Any, Dict, List, Optional, Set, Text, Tuple
+from typing import TYPE_CHECKING, Any, Dict, List, Optional, Text, Tuple
 
 import structlog
 import tiktoken
@@ -18,7 +18,6 @@ from rasa.core.constants import (
     UTTER_SOURCE_METADATA_KEY,
 )
 from rasa.core.policies.policy import Policy, PolicyPrediction, SupportedData
-from rasa.dialogue_understanding.patterns.chitchat import FLOW_PATTERN_CHITCHAT
 from rasa.dialogue_understanding.stack.frames import (
     ChitChatStackFrame,
     DialogueStackFrame,
@@ -38,10 +37,9 @@ from rasa.shared.constants import (
     OPENAI_PROVIDER,
     PROMPT_CONFIG_KEY,
     PROVIDER_CONFIG_KEY,
-    REQUIRED_SLOTS_KEY,
     TIMEOUT_CONFIG_KEY,
 )
-from rasa.shared.core.constants import ACTION_LISTEN_NAME, ACTION_TRIGGER_CHITCHAT
+from rasa.shared.core.constants import ACTION_LISTEN_NAME
 from rasa.shared.core.domain import KEY_RESPONSES_TEXT, Domain
 from rasa.shared.core.events import (
     ActionExecuted,
@@ -51,6 +49,7 @@ from rasa.shared.core.events import (
 )
 from rasa.shared.core.flows import FlowsList
 from rasa.shared.core.generator import TrackerWithCachedStates
+from rasa.shared.core.policies.utils import filter_responses_for_intentless_policy
 from rasa.shared.core.trackers import DialogueStateTracker
 from rasa.shared.exceptions import FileIOException, RasaCoreException
 from rasa.shared.nlu.constants import PREDICTED_CONFIDENCE_KEY
@@ -146,59 +145,6 @@ class Conversation:
     interactions: List[Interaction] = field(default_factory=list)
 
 
-def collect_form_responses(forms: Forms) -> Set[Text]:
-    """Collect responses that belong the requested slots in forms.
-
-    Args:
-        forms: the forms from the domain
-    Returns:
-        all utterances used in forms
-    """
-    form_responses = set()
-    for _, form_info in forms.data.items():
-        for required_slot in form_info.get(REQUIRED_SLOTS_KEY, []):
-            form_responses.add(f"utter_ask_{required_slot}")
-    return form_responses
-
-
-def filter_responses(responses: Responses, forms: Forms, flows: FlowsList) -> Responses:
-    """Filters out responses that are unwanted for the intentless policy.
-
-    This includes utterances used in flows and forms.
-
-    Args:
-        responses: the responses from the domain
-        forms: the forms from the domain
-        flows: all flows
-    Returns:
-        The remaining, relevant responses for the intentless policy.
-    """
-    form_responses = collect_form_responses(forms)
-    flow_responses = flows.utterances
-    combined_responses = form_responses | flow_responses
-    filtered_responses = {
-        name: variants
-        for name, variants in responses.data.items()
-        if name not in combined_responses
-    }
-
-    pattern_chitchat = flows.flow_by_id(FLOW_PATTERN_CHITCHAT)
-
-    # The following condition is highly unlikely, but mypy requires the case
-    # of pattern_chitchat == None to be addressed
-    if not pattern_chitchat:
-        return Responses(data=filtered_responses)
-
-    # if action_trigger_chitchat, filter out "utter_free_chitchat_response"
-    has_action_trigger_chitchat = pattern_chitchat.has_action_step(
-        ACTION_TRIGGER_CHITCHAT
-    )
-    if has_action_trigger_chitchat:
-        filtered_responses.pop("utter_free_chitchat_response", None)
-
-    return Responses(data=filtered_responses)
-
-
 def action_from_response(
     text: Optional[str], responses: Dict[Text, List[Dict[Text, Any]]]
 ) -> Optional[str]:
@@ -512,7 +458,9 @@ class IntentlessPolicy(LLMHealthCheckMixin, EmbeddingsHealthCheckMixin, Policy):
         # Perform health checks of both LLM and embeddings client configs
         self._perform_health_checks(self.config, "intentless_policy.train")
 
-        responses = filter_responses(responses, forms, flows or FlowsList([]))
+        responses = filter_responses_for_intentless_policy(
+            responses, forms, flows or FlowsList([])
+        )
         telemetry.track_intentless_policy_train()
         response_texts = [r for r in extract_ai_response_examples(responses.data)]
 
@@ -947,7 +895,6 @@ class IntentlessPolicy(LLMHealthCheckMixin, EmbeddingsHealthCheckMixin, Policy):
         **kwargs: Any,
     ) -> "IntentlessPolicy":
         """Loads a trained policy (see parent class for full docstring)."""
-
         # Perform health checks of both LLM and embeddings client configs
         cls._perform_health_checks(config, "intentless_policy.load")
 

rasa_pro-3.12.6.dev1/rasa/dialogue_understanding/generator/_jinja_filters.py

@@ -0,0 +1,9 @@
+import json
+
+
+def to_json_escaped_string(s: str) -> str:
+    """
+    Serializes the input string using JSON encoding to escape special characters
+    like newlines (\\n), tabs (\\t), and quotes.
+    """
+    return json.dumps(s, ensure_ascii=False)

{rasa_pro-3.12.4 → rasa_pro-3.12.6.dev1}/rasa/dialogue_understanding/generator/constants.py

@@ -33,3 +33,7 @@ LLM_BASED_COMMAND_GENERATOR_CONFIG_FILE = "config.json"
 
 MODEL_NAME_GPT_4O_2024_11_20 = "gpt-4o-2024-11-20"
 MODEL_NAME_CLAUDE_3_5_SONNET_20240620 = "claude-3-5-sonnet-20240620"
+
+# JINJA template filters
+
+TO_JSON_ESCAPED_STRING_JINJA_FILTER = "to_json_escaped_string"

{rasa_pro-3.12.4 → rasa_pro-3.12.6.dev1}/rasa/dialogue_understanding/generator/llm_based_command_generator.py

@@ -3,7 +3,7 @@ from functools import lru_cache
 from typing import Any, Dict, List, Optional, Set, Text, Tuple, Union
 
 import structlog
-from jinja2 import Template
+from jinja2 import Environment, Template, select_autoescape
 
 import rasa.dialogue_understanding.generator.utils
 import rasa.shared.utils.io
@@ -17,12 +17,14 @@ from rasa.dialogue_understanding.commands.handle_digressions_command import (
 )
 from rasa.dialogue_understanding.constants import KEY_MINIMIZE_NUM_CALLS
 from rasa.dialogue_understanding.generator import CommandGenerator
+from rasa.dialogue_understanding.generator._jinja_filters import to_json_escaped_string
 from rasa.dialogue_understanding.generator.constants import (
     DEFAULT_LLM_CONFIG,
     FLOW_RETRIEVAL_ACTIVE_KEY,
     FLOW_RETRIEVAL_FLOW_THRESHOLD,
     FLOW_RETRIEVAL_KEY,
     LLM_CONFIG_KEY,
+    TO_JSON_ESCAPED_STRING_JINJA_FILTER,
 )
 from rasa.dialogue_understanding.generator.flow_retrieval import FlowRetrieval
 from rasa.dialogue_understanding.stack.utils import top_flow_frame
@@ -226,12 +228,26 @@ class LLMBasedCommandGenerator(
 
     @lru_cache
     def compile_template(self, template: str) -> Template:
-        """Compile the prompt template.
-
+        """
+        Compile the prompt template and register custom filters.
         Compiling the template is an expensive operation,
         so we cache the result.
         """
-        return Template(template)
+        # Create an environment
+        # Autoescaping disabled explicitly for LLM prompt templates rendered from
+
+        # strings (safe, not HTML)
+        env = Environment(
+            autoescape=select_autoescape(
+                disabled_extensions=["jinja2"], default_for_string=False, default=True
+            )
+        )
+
+        # Register filters
+        env.filters[TO_JSON_ESCAPED_STRING_JINJA_FILTER] = to_json_escaped_string
+
+        # Return the template which can leverage registered filters
+        return env.from_string(template)
 
     @classmethod
     def load_prompt_template_from_model_storage(

{rasa_pro-3.12.4 → rasa_pro-3.12.6.dev1}/rasa/dialogue_understanding/generator/nlu_command_adapter.py

@@ -140,7 +140,7 @@ class NLUCommandAdapter(GraphComponent, CommandGenerator):
 
         if commands:
             commands = clean_up_commands(
-                commands, tracker, flows, self._execution_context
+                commands, tracker, flows, self._execution_context, domain
             )
             log_llm(
                 logger=structlogger,

{rasa_pro-3.12.4 → rasa_pro-3.12.6.dev1}/rasa/dialogue_understanding/generator/prompt_templates/command_prompt_v2_claude_3_5_sonnet_20240620_template.jinja2

@@ -33,7 +33,7 @@ Your task is to analyze the current conversation context and generate a list of
 ## Available Flows and Slots
 Use the following structured data:
 ```json
-{"flows":[{% for flow in available_flows %}{"name":"{{ flow.name }}","description":"{{ flow.description }}"{% if flow.slots %},"slots":[{% for slot in flow.slots %}{"name":"{{ slot.name }}"{% if slot.description %},"description":"{{ slot.description }}"{% endif %}{% if slot.allowed_values %},"allowed_values":{{ slot.allowed_values }}{% endif %}}{% if not loop.last %},{% endif %}{% endfor %}]{% endif %}}{% if not loop.last %},{% endif %}{% endfor %}]}
+{"flows":[{% for flow in available_flows %}{"name":"{{ flow.name }}","description":{{ flow.description | to_json_escaped_string }}{% if flow.slots %},"slots":[{% for slot in flow.slots %}{"name":"{{ slot.name }}"{% if slot.description %},"description":{{ slot.description | to_json_escaped_string }}{% endif %}{% if slot.allowed_values %},"allowed_values":{{ slot.allowed_values }}{% endif %}}{% if not loop.last %},{% endif %}{% endfor %}]{% endif %}}{% if not loop.last %},{% endif %}{% endfor %}]}
 ```
 
 --
@@ -41,7 +41,7 @@ Use the following structured data:
 ## Current State
 {% if current_flow != None %}Use the following structured data:
 ```json
-{"active_flow":"{{ current_flow }}","current_step":{"requested_slot":"{{ current_slot }}","requested_slot_description":"{{ current_slot_description }}"},"slots":[{% for slot in flow_slots %}{"name":"{{ slot.name }}","value":"{{ slot.value }}","type":"{{ slot.type }}"{% if slot.description %},"description":"{{ slot.description }}"{% endif %}{% if slot.allowed_values %},"allowed_values":"{{ slot.allowed_values }}"{% endif %}}{% if not loop.last %},{% endif %}{% endfor %}]}
+{"active_flow":"{{ current_flow }}","current_step":{"requested_slot":"{{ current_slot }}","requested_slot_description":{{ current_slot_description | to_json_escaped_string }}},"slots":[{% for slot in flow_slots %}{"name":"{{ slot.name }}","value":"{{ slot.value }}","type":"{{ slot.type }}"{% if slot.description %},"description":{{ slot.description | to_json_escaped_string }}{% endif %}{% if slot.allowed_values %},"allowed_values":"{{ slot.allowed_values }}"{% endif %}}{% if not loop.last %},{% endif %}{% endfor %}]}
 ```{% else %}
 You are currently not inside any flow.{% endif %}
 

{rasa_pro-3.12.4 → rasa_pro-3.12.6.dev1}/rasa/dialogue_understanding/generator/prompt_templates/command_prompt_v2_gpt_4o_2024_11_20_template.jinja2

@@ -6,7 +6,7 @@ Your task is to analyze the current conversation context and generate a list of
 ## Available Flows and Slots
 Use the following structured data:
 ```json
-{"flows":[{% for flow in available_flows %}{"name":"{{ flow.name }}","description":"{{ flow.description }}"{% if flow.slots %},"slots":[{% for slot in flow.slots %}{"name":"{{ slot.name }}"{% if slot.description %},"description":"{{ slot.description }}"{% endif %}{% if slot.allowed_values %},"allowed_values":{{ slot.allowed_values }}{% endif %}}{% if not loop.last %},{% endif %}{% endfor %}]{% endif %}}{% if not loop.last %},{% endif %}{% endfor %}]}
+{"flows":[{% for flow in available_flows %}{"name":"{{ flow.name }}","description":{{ flow.description | to_json_escaped_string }}{% if flow.slots %},"slots":[{% for slot in flow.slots %}{"name":"{{ slot.name }}"{% if slot.description %},"description":{{ slot.description | to_json_escaped_string }}{% endif %}{% if slot.allowed_values %},"allowed_values":{{ slot.allowed_values }}{% endif %}}{% if not loop.last %},{% endif %}{% endfor %}]{% endif %}}{% if not loop.last %},{% endif %}{% endfor %}]}
 ```
 
 ---
@@ -40,7 +40,7 @@ Use the following structured data:
 ## Current State
 {% if current_flow != None %}Use the following structured data:
 ```json
-{"active_flow":"{{ current_flow }}","current_step":{"requested_slot":"{{ current_slot }}","requested_slot_description":"{{ current_slot_description }}"},"slots":[{% for slot in flow_slots %}{"name":"{{ slot.name }}","value":"{{ slot.value }}","type":"{{ slot.type }}"{% if slot.description %},"description":"{{ slot.description }}"{% endif %}{% if slot.allowed_values %},"allowed_values":"{{ slot.allowed_values }}"{% endif %}}{% if not loop.last %},{% endif %}{% endfor %}]}
+{"active_flow":"{{ current_flow }}","current_step":{"requested_slot":"{{ current_slot }}","requested_slot_description":{{ current_slot_description | to_json_escaped_string }}},"slots":[{% for slot in flow_slots %}{"name":"{{ slot.name }}","value":"{{ slot.value }}","type":"{{ slot.type }}"{% if slot.description %},"description":{{ slot.description | to_json_escaped_string }}{% endif %}{% if slot.allowed_values %},"allowed_values":"{{ slot.allowed_values }}"{% endif %}}{% if not loop.last %},{% endif %}{% endfor %}]}
 ```{% else %}
 You are currently not inside any flow.{% endif %}