rasa-pro 3.10.24__tar.gz → 3.10.26__tar.gz

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: rasa-pro
-Version: 3.10.24
+Version: 3.10.26
 Summary: State-of-the-art open-core Conversational AI framework for Enterprises that natively leverages generative AI for effortless assistant development.
 Keywords: nlp,machine-learning,machine-learning-library,bot,bots,botkit,rasa conversational-agents,conversational-ai,chatbot,chatbot-framework,bot-framework
 Author: Rasa Technologies GmbH
@@ -26,7 +26,7 @@ Requires-Dist: SQLAlchemy (>=2.0.22,<2.1.0)
 Requires-Dist: absl-py (>=2.0,<2.1)
 Requires-Dist: aio-pika (>=8.2.3,<8.2.4)
 Requires-Dist: aiogram (>=2.15,<2.26)
-Requires-Dist: aiohttp (>=3.9.4,<3.10)
+Requires-Dist: aiohttp (>=3.10.11,<3.11)
 Requires-Dist: apscheduler (>=3.10,<3.11)
 Requires-Dist: attrs (>=23.1,<23.2)
 Requires-Dist: azure-storage-blob (>=12.16.0,<12.17.0)
@@ -113,7 +113,7 @@ Requires-Dist: sanic-routing (>=22.8.0,<23.0.0)
 Requires-Dist: scikit-learn (>=1.1.3,<1.2) ; python_version >= "3.9" and python_version < "3.11"
 Requires-Dist: scipy (>=1.10.1,<1.11.0) ; python_version >= "3.9" and python_version < "3.11"
 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: sentry-sdk (>=2.8.0,<3)
 Requires-Dist: setuptools (>=78.1.0,<78.2.0)
 Requires-Dist: sklearn-crfsuite (>=0.3.6,<0.4.0)
 Requires-Dist: skops (>=0.10.0,<0.11.0)

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
 
 [tool.poetry]
 name = "rasa-pro"
-version = "3.10.24"
+version = "3.10.26"
 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>",]
@@ -13,17 +13,44 @@ repository = "https://github.com/rasahq/rasa"
 documentation = "https://rasa.com/docs"
 classifiers = [ "Development Status :: 5 - Production/Stable", "Intended Audience :: Developers", "Topic :: Software Development :: Libraries",]
 keywords = [ "nlp", "machine-learning", "machine-learning-library", "bot", "bots", "botkit", "rasa conversational-agents", "conversational-ai", "chatbot", "chatbot-framework", "bot-framework",]
-include = [
-  # license and readme should not be included in wheels, since they would end up in the root library folder
-  "LICENSE.txt", "README.md",
-  { path = "rasa/shared/core/training_data/visualization.html", format = ["sdist", "wheel"] },
-  { path = "rasa/cli/default_config.yml", format = ["sdist", "wheel"] },
-  { path = "rasa/shared/importers/*", format = ["sdist", "wheel"] },
-  { path = "rasa/utils/schemas/*", format = ["sdist", "wheel"] },
-  { path = "rasa/dialogue_understanding/classifiers/command_prompt_template.jinja2", format = ["sdist", "wheel"] },
-  { path = "rasa/keys", format = ["sdist", "wheel"] }
-]
 readme = "README.md"
+
+[[tool.poetry.include]]
+path = "LICENSE.txt"
+# license should not be included in wheels,
+# since it would end up in the root library folder
+format = "sdist"
+
+[[tool.poetry.include]]
+path = "README.md"
+# readme should not be included in wheels,
+# since it would end up in the root library folder
+format = "sdist"
+
+[[tool.poetry.include]]
+path = "rasa/shared/core/training_data/visualization.html"
+format = ["sdist", "wheel"]
+
+[[tool.poetry.include]]
+path = "rasa/cli/default_config.yml"
+format = ["sdist", "wheel"]
+
+[[tool.poetry.include]]
+path = "rasa/shared/importers/*"
+format = ["sdist", "wheel"]
+
+[[tool.poetry.include]]
+path = "rasa/utils/schemas/*"
+format = ["sdist", "wheel"]
+
+[[tool.poetry.include]]
+path = "rasa/dialogue_understanding/classifiers/command_prompt_template.jinja2"
+format = ["sdist", "wheel"]
+
+[[tool.poetry.include]]
+path = "rasa/keys"
+format = ["sdist", "wheel"]
+
 [[tool.poetry.packages]]
 include = "rasa"
 
@@ -129,7 +156,7 @@ sanic-cors = "~2.2.0"
 sanic-jwt = "^1.8.0"
 sanic-routing = "^22.8.0"
 websockets = ">=10.4,<11.0"
-aiohttp = ">=3.9.4,<3.10"
+aiohttp = ">=3.10.11,<3.11"
 questionary = ">=1.10.0,<2.1.0"
 prompt-toolkit = "^3.0.28,<3.0.29"
 python-socketio = ">=5.8,<6"
@@ -144,7 +171,7 @@ tensorflow_hub = "^0.13.0"
 setuptools = "~78.1.0" # pinning this to same version as Poetry, removing does not update lock file
 ujson = ">=5.8,<6.0"
 regex = ">=2022.10.31,<2022.11"
-sentry-sdk = ">=1.14.0,<1.15.0"
+sentry-sdk = ">=2.8.0,<3"
 aio-pika = ">=8.2.3,<8.2.4"
 aiogram = ">=2.15,<2.26"
 typing-extensions = ">=4.7.1,<5.0.0"
@@ -358,6 +385,8 @@ convention = "google"
 
 
 
+
+
 [tool.poetry.group.dev.dependencies]
 ruff = ">=0.3.5,<0.4.0"
 docker = "~7.1.0"

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/rasa/core/nlg/contextual_response_rephraser.py

@@ -5,13 +5,19 @@ from jinja2 import Template
 
 from rasa import telemetry
 from rasa.core.nlg.response import TemplatedNaturalLanguageGenerator
+from rasa.core.nlg.summarize import (
+    _count_multiple_utterances_as_single_turn,
+    summarize_conversation,
+)
 from rasa.shared.constants import (
     LLM_CONFIG_KEY,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY,
     MODEL_CONFIG_KEY,
     MODEL_NAME_CONFIG_KEY,
+    OPENAI_PROVIDER,
     PROMPT_CONFIG_KEY,
     PROVIDER_CONFIG_KEY,
-    OPENAI_PROVIDER,
+    TEMPERATURE_CONFIG_KEY,
     TIMEOUT_CONFIG_KEY,
 )
 from rasa.shared.core.domain import KEY_RESPONSES_TEXT, Domain
@@ -24,11 +30,10 @@ from rasa.shared.utils.llm import (
     combine_custom_and_default_config,
     get_prompt_template,
     llm_factory,
-    try_instantiate_llm_client,
+    tracker_as_readable_transcript,
 )
 from rasa.utils.endpoints import EndpointConfig
 
-from rasa.core.nlg.summarize import summarize_conversation
 
 from rasa.utils.log_utils import log_llm
 
@@ -39,12 +44,15 @@ RESPONSE_REPHRASING_KEY = "rephrase"
 RESPONSE_REPHRASING_TEMPLATE_KEY = "rephrase_prompt"
 
 DEFAULT_REPHRASE_ALL = False
+DEFAULT_SUMMARIZE_HISTORY = True
+DEFAULT_MAX_HISTORICAL_TURNS = 5
+DEFAULT_COUNT_MULTIPLE_UTTERANCES_AS_SINGLE_TURN = True
 
 DEFAULT_LLM_CONFIG = {
     PROVIDER_CONFIG_KEY: OPENAI_PROVIDER,
     MODEL_CONFIG_KEY: DEFAULT_OPENAI_GENERATE_MODEL_NAME,
-    "temperature": 0.3,
-    "max_tokens": DEFAULT_OPENAI_MAX_GENERATED_TOKENS,
+    TEMPERATURE_CONFIG_KEY: 0.3,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY: DEFAULT_OPENAI_MAX_GENERATED_TOKENS,
     TIMEOUT_CONFIG_KEY: 5,
 }
 
@@ -56,6 +64,7 @@ its meaning. Use simple english.
 Context / previous conversation with the user:
 {{history}}
 
+Last user message:
 {{current_input}}
 
 Suggested AI Response: {{suggested_response}}
@@ -97,11 +106,21 @@ class ContextualResponseRephraser(TemplatedNaturalLanguageGenerator):
         self.trace_prompt_tokens = self.nlg_endpoint.kwargs.get(
             "trace_prompt_tokens", False
         )
-        try_instantiate_llm_client(
+        self.summarize_history = self.nlg_endpoint.kwargs.get(
+            "summarize_history", DEFAULT_SUMMARIZE_HISTORY
+        )
+        self.max_historical_turns = self.nlg_endpoint.kwargs.get(
+            "max_historical_turns", DEFAULT_MAX_HISTORICAL_TURNS
+        )
+
+        self.count_multiple_utterances_as_single_turn = self.nlg_endpoint.kwargs.get(
+            "count_multiple_utterances_as_single_turn",
+            DEFAULT_COUNT_MULTIPLE_UTTERANCES_AS_SINGLE_TURN,
+        )
+
+        self.llm_config = combine_custom_and_default_config(
             self.nlg_endpoint.kwargs.get(LLM_CONFIG_KEY),
             DEFAULT_LLM_CONFIG,
-            "contextual_response_rephraser.init",
-            "ContextualResponseRephraser",
         )
 
     def _last_message_if_human(self, tracker: DialogueStateTracker) -> Optional[str]:
@@ -180,10 +199,16 @@ class ContextualResponseRephraser(TemplatedNaturalLanguageGenerator):
         Returns:
         The history for the prompt.
         """
-        llm = llm_factory(
-            self.nlg_endpoint.kwargs.get(LLM_CONFIG_KEY), DEFAULT_LLM_CONFIG
+        # Count multiple utterances by bot/user as single turn in conversation history
+        turns_wrapper = (
+            _count_multiple_utterances_as_single_turn
+            if self.count_multiple_utterances_as_single_turn
+            else None
+        )
+        llm = llm_factory(self.llm_config, DEFAULT_LLM_CONFIG)
+        return await summarize_conversation(
+            tracker, llm, max_turns=5, turns_wrapper=turns_wrapper
         )
-        return await summarize_conversation(tracker, llm, max_turns=5)
 
     async def rephrase(
         self,
@@ -195,7 +220,6 @@ class ContextualResponseRephraser(TemplatedNaturalLanguageGenerator):
         Args:
             response: The response to rephrase.
             tracker: The tracker to use for the prediction.
-            model_name: The name of the model to use for the prediction.
 
         Returns:
             The response with the rephrased text.
@@ -208,8 +232,29 @@ class ContextualResponseRephraser(TemplatedNaturalLanguageGenerator):
 
         prompt_template_text = self._template_for_response_rephrasing(response)
 
+        # Last user message (=current input) should always be in prompt if available
+        last_message_by_user = getattr(tracker.latest_message, "text", "")
+        current_input = (
+            f"{USER}: {last_message_by_user}" if last_message_by_user else ""
+        )
+
+        # Only summarise conversation history if flagged
+        if self.summarize_history:
+            history = await self._create_history(tracker)
+        else:
+            # Count multiple utterances by bot/user as single turn
+            turns_wrapper = (
+                _count_multiple_utterances_as_single_turn
+                if self.count_multiple_utterances_as_single_turn
+                else None
+            )
+            max_turns = max(self.max_historical_turns, 1)
+            history = tracker_as_readable_transcript(
+                tracker, max_turns=max_turns, turns_wrapper=turns_wrapper
+            )
+
         prompt = Template(prompt_template_text).render(
-            history=await self._create_history(tracker),
+            history=history,
             suggested_response=response_text,
             current_input=current_input,
             slots=tracker.current_slot_values(),

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/rasa/core/nlg/summarize.py

@@ -1,4 +1,5 @@
-from typing import Optional
+from itertools import groupby
+from typing import Callable, List, Optional
 
 import structlog
 from jinja2 import Template
@@ -22,20 +23,47 @@ SUMMARY_PROMPT_TEMPLATE = Template(_DEFAULT_SUMMARIZER_TEMPLATE)
 MAX_TURNS_DEFAULT = 20
 
 
+def _count_multiple_utterances_as_single_turn(transcript: List[str]) -> List[str]:
+    """Counts multiple utterances as a single turn.
+    Args:
+        transcript: the lines of the transcript
+    Returns:
+        transcript: with multiple utterances counted as a single turn
+    """
+    if not transcript:
+        return []
+
+    def get_speaker_label(line: str) -> str:
+        return line.partition(": ")[0] if ": " in line else ""
+
+    modified_transcript = [
+        f"{speaker}: {' '.join(line.partition(': ')[2] for line in group)}"
+        for speaker, group in groupby(transcript, key=get_speaker_label)
+        if speaker
+    ]
+
+    return modified_transcript
+
+
 def _create_summarization_prompt(
-    tracker: DialogueStateTracker, max_turns: Optional[int]
+    tracker: DialogueStateTracker,
+    max_turns: Optional[int],
+    turns_wrapper: Optional[Callable[[List[str]], List[str]]],
 ) -> str:
     """Creates an LLM prompt to summarize the conversation in the tracker.
 
     Args:
         tracker: tracker of the conversation to be summarized
         max_turns: maximum number of turns to summarize
+        turns_wrapper: optional function to wrap the turns
 
 
     Returns:
         The prompt to summarize the conversation.
     """
-    transcript = tracker_as_readable_transcript(tracker, max_turns=max_turns)
+    transcript = tracker_as_readable_transcript(
+        tracker, max_turns=max_turns, turns_wrapper=turns_wrapper
+    )
     return SUMMARY_PROMPT_TEMPLATE.render(
         conversation=transcript,
     )
@@ -45,6 +73,7 @@ async def summarize_conversation(
     tracker: DialogueStateTracker,
     llm: LLMClient,
     max_turns: Optional[int] = MAX_TURNS_DEFAULT,
+    turns_wrapper: Optional[Callable[[List[str]], List[str]]] = None,
 ) -> str:
     """Summarizes the dialogue using the LLM.
 
@@ -52,11 +81,12 @@ async def summarize_conversation(
         tracker: the tracker to summarize
         llm: the LLM to use for summarization
         max_turns: maximum number of turns to summarize
+        turns_wrapper: optional function to wrap the turns
 
     Returns:
         The summary of the dialogue.
     """
-    prompt = _create_summarization_prompt(tracker, max_turns)
+    prompt = _create_summarization_prompt(tracker, max_turns, turns_wrapper)
     try:
         llm_response = await llm.acompletion(prompt)
         summarization = llm_response.choices[0].strip()
@@ -65,6 +95,8 @@ async def summarize_conversation(
         )
         return summarization
     except Exception as e:
-        transcript = tracker_as_readable_transcript(tracker, max_turns=max_turns)
+        transcript = tracker_as_readable_transcript(
+            tracker, max_turns=max_turns, turns_wrapper=turns_wrapper
+        )
         structlogger.error("summarization.error", error=e)
         return transcript

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/rasa/core/policies/enterprise_search_policy.py

@@ -48,11 +48,14 @@ from rasa.graph_components.providers.responses_provider import Responses
 from rasa.shared.constants import (
     EMBEDDINGS_CONFIG_KEY,
     LLM_CONFIG_KEY,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY,
+    MAX_RETRIES_CONFIG_KEY,
     MODEL_CONFIG_KEY,
     MODEL_NAME_CONFIG_KEY,
+    OPENAI_PROVIDER,
     PROMPT_CONFIG_KEY,
     PROVIDER_CONFIG_KEY,
-    OPENAI_PROVIDER,
+    TEMPERATURE_CONFIG_KEY,
     TIMEOUT_CONFIG_KEY,
 )
 from rasa.shared.core.constants import (
@@ -117,14 +120,14 @@ DEFAULT_LLM_CONFIG = {
     PROVIDER_CONFIG_KEY: OPENAI_PROVIDER,
     MODEL_CONFIG_KEY: DEFAULT_OPENAI_CHAT_MODEL_NAME,
     TIMEOUT_CONFIG_KEY: 10,
-    "temperature": 0.0,
-    "max_tokens": 256,
-    "max_retries": 1,
+    TEMPERATURE_CONFIG_KEY: 0.0,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY: 256,
+    MAX_RETRIES_CONFIG_KEY: 1,
 }
 
 DEFAULT_EMBEDDINGS_CONFIG = {
     PROVIDER_CONFIG_KEY: OPENAI_PROVIDER,
-    "model": DEFAULT_OPENAI_EMBEDDING_MODEL_NAME,
+    MODEL_CONFIG_KEY: DEFAULT_OPENAI_EMBEDDING_MODEL_NAME,
 }
 
 ENTERPRISE_SEARCH_PROMPT_FILE_NAME = "enterprise_search_policy_prompt.jinja2"

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/rasa/core/policies/intentless_policy.py

@@ -31,11 +31,13 @@ from rasa.graph_components.providers.responses_provider import Responses
 from rasa.shared.constants import (
     EMBEDDINGS_CONFIG_KEY,
     LLM_CONFIG_KEY,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY,
     MODEL_CONFIG_KEY,
     MODEL_NAME_CONFIG_KEY,
+    OPENAI_PROVIDER,
     PROMPT_CONFIG_KEY,
     PROVIDER_CONFIG_KEY,
-    OPENAI_PROVIDER,
+    TEMPERATURE_CONFIG_KEY,
     TIMEOUT_CONFIG_KEY,
 )
 from rasa.shared.core.constants import ACTION_LISTEN_NAME
@@ -106,14 +108,14 @@ NLU_ABSTENTION_THRESHOLD = "nlu_abstention_threshold"
 DEFAULT_LLM_CONFIG = {
     PROVIDER_CONFIG_KEY: OPENAI_PROVIDER,
     MODEL_CONFIG_KEY: DEFAULT_OPENAI_CHAT_MODEL_NAME,
-    "temperature": 0.0,
-    "max_tokens": DEFAULT_OPENAI_MAX_GENERATED_TOKENS,
+    TEMPERATURE_CONFIG_KEY: 0.0,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY: DEFAULT_OPENAI_MAX_GENERATED_TOKENS,
     TIMEOUT_CONFIG_KEY: 5,
 }
 
 DEFAULT_EMBEDDINGS_CONFIG = {
     PROVIDER_CONFIG_KEY: OPENAI_PROVIDER,
-    "model": DEFAULT_OPENAI_EMBEDDING_MODEL_NAME,
+    MODEL_CONFIG_KEY: DEFAULT_OPENAI_EMBEDDING_MODEL_NAME,
 }
 
 DEFAULT_INTENTLESS_PROMPT_TEMPLATE = importlib.resources.open_text(
@@ -338,8 +340,6 @@ class IntentlessPolicy(Policy):
             # ensures that the policy will not override a deterministic policy
             # which utilizes the nlu predictions confidence (e.g. Memoization).
             NLU_ABSTENTION_THRESHOLD: 0.9,
-            LLM_CONFIG_KEY: DEFAULT_LLM_CONFIG,
-            EMBEDDINGS_CONFIG_KEY: DEFAULT_EMBEDDINGS_CONFIG,
             PROMPT_CONFIG_KEY: DEFAULT_INTENTLESS_PROMPT_TEMPLATE,
         }
 
@@ -374,6 +374,13 @@ class IntentlessPolicy(Policy):
         """Constructs a new Policy object."""
         super().__init__(config, model_storage, resource, execution_context, featurizer)
 
+        self.config[LLM_CONFIG_KEY] = combine_custom_and_default_config(
+            self.config.get(LLM_CONFIG_KEY), DEFAULT_LLM_CONFIG
+        )
+        self.config[EMBEDDINGS_CONFIG_KEY] = combine_custom_and_default_config(
+            self.config.get(EMBEDDINGS_CONFIG_KEY), DEFAULT_EMBEDDINGS_CONFIG
+        )
+
         self.nlu_abstention_threshold: float = self.config[NLU_ABSTENTION_THRESHOLD]
         self.response_index = responses_docsearch
         self.conversation_samples_index = samples_docsearch

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/rasa/dialogue_understanding/coexistence/llm_based_router.py

@@ -21,11 +21,14 @@ from rasa.engine.recipes.default_recipe import DefaultV1Recipe
 from rasa.engine.storage.resource import Resource
 from rasa.engine.storage.storage import ModelStorage
 from rasa.shared.constants import (
-    ROUTE_TO_CALM_SLOT,
-    PROMPT_CONFIG_KEY,
-    PROVIDER_CONFIG_KEY,
+    LOGIT_BIAS_CONFIG_KEY,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY,
     MODEL_CONFIG_KEY,
     OPENAI_PROVIDER,
+    PROMPT_CONFIG_KEY,
+    PROVIDER_CONFIG_KEY,
+    ROUTE_TO_CALM_SLOT,
+    TEMPERATURE_CONFIG_KEY,
     TIMEOUT_CONFIG_KEY,
 )
 from rasa.shared.core.trackers import DialogueStateTracker
@@ -57,9 +60,11 @@ DEFAULT_LLM_CONFIG = {
     PROVIDER_CONFIG_KEY: OPENAI_PROVIDER,
     MODEL_CONFIG_KEY: DEFAULT_OPENAI_CHAT_MODEL_NAME,
     TIMEOUT_CONFIG_KEY: 7,
-    "temperature": 0.0,
-    "max_tokens": 1,
-    "logit_bias": {str(token_id): 100 for token_id in A_TO_C_TOKEN_IDS_CHATGPT},
+    TEMPERATURE_CONFIG_KEY: 0.0,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY: 1,
+    LOGIT_BIAS_CONFIG_KEY: {
+        str(token_id): 100 for token_id in A_TO_C_TOKEN_IDS_CHATGPT
+    },
 }
 
 structlogger = structlog.get_logger()

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/rasa/dialogue_understanding/generator/constants.py

@@ -1,7 +1,9 @@
 from rasa.shared.constants import (
-    PROVIDER_CONFIG_KEY,
-    OPENAI_PROVIDER,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY,
     MODEL_CONFIG_KEY,
+    OPENAI_PROVIDER,
+    PROVIDER_CONFIG_KEY,
+    TEMPERATURE_CONFIG_KEY,
     TIMEOUT_CONFIG_KEY,
 )
 from rasa.shared.utils.llm import (
@@ -12,8 +14,8 @@ from rasa.shared.utils.llm import (
 DEFAULT_LLM_CONFIG = {
     PROVIDER_CONFIG_KEY: OPENAI_PROVIDER,
     MODEL_CONFIG_KEY: DEFAULT_OPENAI_CHAT_MODEL_NAME_ADVANCED,
-    "temperature": 0.0,
-    "max_tokens": DEFAULT_OPENAI_MAX_GENERATED_TOKENS,
+    TEMPERATURE_CONFIG_KEY: 0.0,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY: DEFAULT_OPENAI_MAX_GENERATED_TOKENS,
     TIMEOUT_CONFIG_KEY: 7,
 }
 

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/rasa/dialogue_understanding/generator/multi_step/multi_step_llm_command_generator.py

@@ -246,7 +246,7 @@ class MultiStepLLMCommandGenerator(LLMBasedCommandGenerator):
         skip_question_re = re.compile(r"SkipQuestion\(\)")
         knowledge_re = re.compile(r"SearchAndReply\(\)")
         humand_handoff_re = re.compile(r"HumanHandoff\(\)")
-        clarify_re = re.compile(r"Clarify\(([\"\'a-zA-Z0-9_, ]+)\)")
+        clarify_re = re.compile(r"Clarify\(([\"\'a-zA-Z0-9_\-, ]+)\)")
         cannot_handle_re = re.compile(r"CannotHandle\(\)")
 
         for action in actions.strip().splitlines():

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/rasa/dialogue_understanding/generator/single_step/single_step_llm_command_generator.py

@@ -308,7 +308,7 @@ class SingleStepLLMCommandGenerator(LLMBasedCommandGenerator):
         skip_question_re = re.compile(r"SkipQuestion\(\)")
         knowledge_re = re.compile(r"SearchAndReply\(\)")
         humand_handoff_re = re.compile(r"HumanHandoff\(\)")
-        clarify_re = re.compile(r"Clarify\(([\"\'a-zA-Z0-9_, ]+)\)")
+        clarify_re = re.compile(r"Clarify\(([\"\'a-zA-Z0-9_\-, ]+)\)")
 
         for action in actions.strip().splitlines():
             if match := slot_set_re.search(action):

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/rasa/dialogue_understanding/processor/command_processor_component.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
-from typing import Any, Dict, List, Text
+from typing import Any, Dict, List, Optional, Text
+
 import rasa.dialogue_understanding.processor.command_processor
 
 from rasa.engine.graph import ExecutionContext, GraphComponent
@@ -37,8 +38,8 @@ class CommandProcessorComponent(GraphComponent):
         self,
         tracker: DialogueStateTracker,
         flows: FlowsList,
-        story_graph: StoryGraph,
         domain: Domain,
+        story_graph: Optional[StoryGraph] = None,
     ) -> List[Event]:
         """Execute commands to update tracker state."""
         return rasa.dialogue_understanding.processor.command_processor.execute_commands(

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/rasa/engine/recipes/default_recipe.py

@@ -383,7 +383,9 @@ class DefaultV1Recipe(Recipe):
         return preprocessors
 
     def _get_needs_from_args(
-        self, component: Type[GraphComponent], fn_name: str
+        self,
+        component: Type[GraphComponent],
+        fn_name: str,
     ) -> Dict[str, str]:
         """Get the needed arguments from the method on the component.
 
@@ -421,6 +423,7 @@ class DefaultV1Recipe(Recipe):
         parameters = {
             name
             for name, param in sig.parameters.items()
+            # only consider parameters which are positional or keyword
             if param.kind == param.POSITIONAL_OR_KEYWORD
         }
 
@@ -737,8 +740,28 @@ class DefaultV1Recipe(Recipe):
                 predict_config, predict_nodes, train_nodes, preprocessors
             )
 
+        # The `story_graph_provider` is only needed if the intentless policy is used.
+        # If it is not used, we can remove it from the nodes as it slows down the
+        # loading time if users have a large number of stories.
+        if not self._intentless_policy_used(predict_nodes):
+            # Removes the `story_graph_provider` from the nodes
+            predict_nodes.pop("story_graph_provider", None)
+            if "command_processor" in predict_nodes:
+                # Removes story_graph from the command processor inputs
+                predict_nodes["command_processor"].needs.pop("story_graph", None)
+
         return predict_nodes
 
+    @staticmethod
+    def _intentless_policy_used(nodes: Dict[Text, SchemaNode]) -> bool:
+        """Checks if the intentless policy is used in the nodes."""
+        from rasa.core import IntentlessPolicy
+
+        for schema_node in nodes.values():
+            if schema_node.matches_type(IntentlessPolicy):
+                return True
+        return False
+
     def _add_nlu_predict_nodes(
         self,
         last_run_node: Text,
@@ -909,7 +932,8 @@ class DefaultV1Recipe(Recipe):
         predict_nodes["command_processor"] = SchemaNode(
             **DEFAULT_PREDICT_KWARGS,
             needs=self._get_needs_from_args(
-                CommandProcessorComponent, "execute_commands"
+                CommandProcessorComponent,
+                "execute_commands",
             ),
             uses=CommandProcessorComponent,
             fn="execute_commands",

{rasa_pro-3.10.24 → rasa_pro-3.10.26}/rasa/llm_fine_tuning/paraphrasing/conversation_rephraser.py

@@ -10,11 +10,13 @@ from rasa.llm_fine_tuning.paraphrasing.rephrased_user_message import (
     RephrasedUserMessage,
 )
 from rasa.shared.constants import (
-    MODEL_NAME_CONFIG_KEY,
-    MODEL_CONFIG_KEY,
     LLM_CONFIG_KEY,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY,
+    MODEL_CONFIG_KEY,
+    MODEL_NAME_CONFIG_KEY,
     PROMPT_TEMPLATE_CONFIG_KEY,
     PROVIDER_CONFIG_KEY,
+    TEMPERATURE_CONFIG_KEY,
     TIMEOUT_CONFIG_KEY,
 )
 from rasa.shared.exceptions import ProviderClientAPIException
@@ -38,8 +40,8 @@ DEFAULT_LLM_CONFIG = {
     PROVIDER_CONFIG_KEY: OPENAI_PROVIDER,
     MODEL_CONFIG_KEY: "gpt-4o-mini",
     TIMEOUT_CONFIG_KEY: 7,
-    "temperature": 0.0,
-    "max_tokens": 4096,
+    TEMPERATURE_CONFIG_KEY: 0.0,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY: 4096,
 }
 
 structlogger = structlog.get_logger()