rasa-pro 3.12.13__tar.gz → 3.12.15__tar.gz

{rasa_pro-3.12.13 → rasa_pro-3.12.15}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: rasa-pro
-Version: 3.12.13
+Version: 3.12.15
 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

{rasa_pro-3.12.13 → rasa_pro-3.12.15}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
 
 [tool.poetry]
 name = "rasa-pro"
-version = "3.12.13"
+version = "3.12.15"
 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"
 
@@ -341,6 +368,8 @@ convention = "google"
 
 
 
+
+
 [tool.poetry.group.dev.dependencies]
 ruff = ">=0.3.5,<0.4.0"
 docker = "~7.1.0"

{rasa_pro-3.12.13 → rasa_pro-3.12.15}/rasa/cli/llm_fine_tuning.py

@@ -208,10 +208,7 @@ def prepare_llm_fine_tuning_data(args: argparse.Namespace) -> None:
         sys.exit(0)
 
     flows = asyncio.run(e2e_test_runner.agent.processor.get_flows())
-    llm_command_generator_config = _get_llm_command_generator_config(e2e_test_runner)
-    llm_command_generator: Type[LLMBasedCommandGenerator] = _get_llm_command_generator(
-        e2e_test_runner
-    )
+    _validate_llm_command_generator_present(e2e_test_runner)
 
     # set up storage context
     storage_context = create_storage_context(StorageType.FILE, output_dir)
@@ -242,11 +239,11 @@ def prepare_llm_fine_tuning_data(args: argparse.Namespace) -> None:
             rephrase_config,
             args.num_rephrases,
             flows,
-            llm_command_generator,
-            llm_command_generator_config,
+            e2e_test_runner.agent,
             storage_context,
         )
     )
+
     statistics["num_passing_rephrased_user_messages"] = sum(
         [conversation.get_number_of_rephrases(True) for conversation in conversations]
     )
@@ -257,7 +254,11 @@ def prepare_llm_fine_tuning_data(args: argparse.Namespace) -> None:
 
     # 3. create fine-tuning dataset
     log_start_of_module("LLM Data Preparation")
-    llm_fine_tuning_data = convert_to_fine_tuning_data(conversations, storage_context)
+    llm_fine_tuning_data = asyncio.run(
+        convert_to_fine_tuning_data(
+            conversations, storage_context, e2e_test_runner.agent
+        )
+    )
     statistics["num_ft_data_points"] = len(llm_fine_tuning_data)
     log_end_of_module("LLM Data Preparation", statistics)
 
@@ -311,9 +312,9 @@ def _get_llm_command_generator_config(e2e_test_runner: E2ETestRunner) -> Dict[st
     sys.exit(1)
 
 
-def _get_llm_command_generator(
+def _validate_llm_command_generator_present(
     e2e_test_runner: E2ETestRunner,
-) -> Type[LLMBasedCommandGenerator]:
+) -> None:
     train_schema = e2e_test_runner.agent.processor.model_metadata.train_schema  # type: ignore
 
     for _, node in train_schema.nodes.items():
@@ -322,7 +323,7 @@ def _get_llm_command_generator(
         ) and not node.matches_type(
             MultiStepLLMCommandGenerator, include_subtypes=True
         ):
-            return cast(Type[LLMBasedCommandGenerator], node.uses)
+            return
 
     rasa.shared.utils.cli.print_error(
         "The provided model is not trained using 'SingleStepLLMCommandGenerator' or "

{rasa_pro-3.12.13 → rasa_pro-3.12.15}/rasa/core/nlg/contextual_response_rephraser.py

@@ -5,15 +5,20 @@ from jinja2 import Template
 
 from rasa import telemetry
 from rasa.core.nlg.response import TemplatedNaturalLanguageGenerator
-from rasa.core.nlg.summarize import summarize_conversation
+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_GROUP_ID_CONFIG_KEY,
     MODEL_NAME_CONFIG_KEY,
     OPENAI_PROVIDER,
     PROMPT_CONFIG_KEY,
     PROVIDER_CONFIG_KEY,
+    TEMPERATURE_CONFIG_KEY,
     TIMEOUT_CONFIG_KEY,
 )
 from rasa.shared.core.domain import KEY_RESPONSES_TEXT, Domain
@@ -53,12 +58,13 @@ RESPONSE_SUMMARISE_CONVERSATION_KEY = "summarize_conversation"
 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,
 }
 
@@ -70,6 +76,7 @@ its meaning. Use simple {{language}}.
 Context / previous conversation with the user:
 {{history}}
 
+Last user message:
 {{current_input}}
 
 Suggested AI Response: {{suggested_response}}
@@ -122,6 +129,11 @@ class ContextualResponseRephraser(
             "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 = resolve_model_client_config(
             self.nlg_endpoint.kwargs.get(LLM_CONFIG_KEY),
             ContextualResponseRephraser.__name__,
@@ -260,8 +272,16 @@ class ContextualResponseRephraser(
         Returns:
         The history for the prompt.
         """
+        # 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)
+        return await summarize_conversation(
+            tracker, llm, max_turns=5, turns_wrapper=turns_wrapper
+        )
 
     async def rephrase(
         self,
@@ -283,19 +303,26 @@ class ContextualResponseRephraser(
 
         prompt_template_text = self._template_for_response_rephrasing(response)
 
-        # Retrieve inputs for the dynamic prompt
-        latest_message = self._last_message_if_human(tracker)
-        current_input = f"{USER}: {latest_message}" if latest_message else ""
+        # 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:
-            # make sure the transcript/history contains the last user utterance
+            # 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)
-            # the history already contains the current input
-            current_input = ""
+            history = tracker_as_readable_transcript(
+                tracker, max_turns=max_turns, turns_wrapper=turns_wrapper
+            )
 
         prompt = Template(prompt_template_text).render(
             history=history,

{rasa_pro-3.12.13 → rasa_pro-3.12.15}/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
@@ -23,20 +24,49 @@ 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,
     )
@@ -46,6 +76,7 @@ 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.
 
@@ -53,11 +84,12 @@ 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()
@@ -66,6 +98,8 @@ 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-3.12.13 → rasa_pro-3.12.15}/rasa/core/policies/enterprise_search_policy.py

@@ -46,12 +46,15 @@ from rasa.graph_components.providers.forms_provider import Forms
 from rasa.graph_components.providers.responses_provider import Responses
 from rasa.shared.constants import (
     EMBEDDINGS_CONFIG_KEY,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY,
+    MAX_RETRIES_CONFIG_KEY,
     MODEL_CONFIG_KEY,
     MODEL_GROUP_ID_CONFIG_KEY,
     MODEL_NAME_CONFIG_KEY,
     OPENAI_PROVIDER,
     PROMPT_CONFIG_KEY,
     PROVIDER_CONFIG_KEY,
+    TEMPERATURE_CONFIG_KEY,
     TIMEOUT_CONFIG_KEY,
 )
 from rasa.shared.core.constants import (
@@ -135,14 +138,14 @@ 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-3.12.13 → rasa_pro-3.12.15}/rasa/core/policies/intentless_policy.py

@@ -31,12 +31,14 @@ 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,
     MODEL_CONFIG_KEY,
     MODEL_GROUP_ID_CONFIG_KEY,
     MODEL_NAME_CONFIG_KEY,
     OPENAI_PROVIDER,
     PROMPT_CONFIG_KEY,
     PROVIDER_CONFIG_KEY,
+    TEMPERATURE_CONFIG_KEY,
     TIMEOUT_CONFIG_KEY,
 )
 from rasa.shared.core.constants import ACTION_LISTEN_NAME
@@ -111,14 +113,14 @@ 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(
@@ -344,8 +346,6 @@ class IntentlessPolicy(LLMHealthCheckMixin, EmbeddingsHealthCheckMixin, 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,
         }
 
@@ -381,13 +381,19 @@ class IntentlessPolicy(LLMHealthCheckMixin, EmbeddingsHealthCheckMixin, Policy):
         super().__init__(config, model_storage, resource, execution_context, featurizer)
 
         # Resolve LLM config
-        self.config[LLM_CONFIG_KEY] = resolve_model_client_config(
-            self.config.get(LLM_CONFIG_KEY), IntentlessPolicy.__name__
+        self.config[LLM_CONFIG_KEY] = combine_custom_and_default_config(
+            resolve_model_client_config(
+                self.config.get(LLM_CONFIG_KEY), IntentlessPolicy.__name__
+            ),
+            DEFAULT_LLM_CONFIG,
         )
 
         # Resolve embeddings config
-        self.config[EMBEDDINGS_CONFIG_KEY] = resolve_model_client_config(
-            self.config.get(EMBEDDINGS_CONFIG_KEY), IntentlessPolicy.__name__
+        self.config[EMBEDDINGS_CONFIG_KEY] = combine_custom_and_default_config(
+            resolve_model_client_config(
+                self.config.get(EMBEDDINGS_CONFIG_KEY), IntentlessPolicy.__name__
+            ),
+            DEFAULT_EMBEDDINGS_CONFIG,
         )
 
         self.nlu_abstention_threshold: float = self.config[NLU_ABSTENTION_THRESHOLD]

{rasa_pro-3.12.13 → rasa_pro-3.12.15}/rasa/core/utils.py

@@ -244,6 +244,10 @@ class AvailableEndpoints:
             cls._instance = cls.read_endpoints(endpoint_file)
         return cls._instance
 
+    @classmethod
+    def reset_instance(cls) -> None:
+        cls._instance = None
+
 
 def read_endpoints_from_path(
     endpoints_path: Optional[Union[Path, Text]] = None,

{rasa_pro-3.12.13 → rasa_pro-3.12.15}/rasa/dialogue_understanding/coexistence/llm_based_router.py

@@ -23,11 +23,14 @@ 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 (
+    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
@@ -66,9 +69,11 @@ 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-3.12.13 → rasa_pro-3.12.15}/rasa/dialogue_understanding/commands/clarify_command.py

@@ -119,7 +119,7 @@ class ClarifyCommand(Command):
         mapper = {
             CommandSyntaxVersion.v1: r"Clarify\(([\"\'a-zA-Z0-9_, -]*)\)",
             CommandSyntaxVersion.v2: (
-                r"""^[\s\W\d]*disambiguate flows (["'a-zA-Z0-9_, -]*)['"`]*$"""
+                r"""^[\s\W\d]*disambiguate flows (["'a-zA-Z0-9_, -]*)[\W\\n]*$"""
             ),
         }
         return mapper.get(

{rasa_pro-3.12.13 → rasa_pro-3.12.15}/rasa/dialogue_understanding/commands/set_slot_command.py

@@ -190,7 +190,7 @@ class SetSlotCommand(Command):
                 r"""SetSlot\(['"]?([a-zA-Z_][a-zA-Z0-9_-]*)['"]?, ?['"]?(.*)['"]?\)"""
             ),
             CommandSyntaxVersion.v2: (
-                r"""^[\s\W\d]*set slot ['"`]?([a-zA-Z_][a-zA-Z0-9_-]*)['"`]? ['"`]?(.+?)['"`]*$"""  # noqa: E501
+                r"""^[\s\W\d]*set slot ['"`]?([a-zA-Z_][a-zA-Z0-9_-]*)['"`]? ['"`]?(.+?)[\W\\n]*$"""  # noqa: E501
             ),
         }
         return mapper.get(

{rasa_pro-3.12.13 → rasa_pro-3.12.15}/rasa/dialogue_understanding/generator/constants.py

@@ -1,5 +1,5 @@
 from rasa.shared.constants import (
-    MAX_TOKENS_CONFIG_KEY,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY,
     MODEL_CONFIG_KEY,
     OPENAI_PROVIDER,
     PROVIDER_CONFIG_KEY,
@@ -15,7 +15,7 @@ DEFAULT_LLM_CONFIG = {
     PROVIDER_CONFIG_KEY: OPENAI_PROVIDER,
     MODEL_CONFIG_KEY: DEFAULT_OPENAI_CHAT_MODEL_NAME_ADVANCED,
     TEMPERATURE_CONFIG_KEY: 0.0,
-    MAX_TOKENS_CONFIG_KEY: DEFAULT_OPENAI_MAX_GENERATED_TOKENS,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY: DEFAULT_OPENAI_MAX_GENERATED_TOKENS,
     TIMEOUT_CONFIG_KEY: 7,
 }
 

{rasa_pro-3.12.13 → rasa_pro-3.12.15}/rasa/dialogue_understanding/generator/single_step/compact_llm_command_generator.py

@@ -47,7 +47,7 @@ from rasa.shared.constants import (
     AWS_BEDROCK_PROVIDER,
     AZURE_OPENAI_PROVIDER,
     EMBEDDINGS_CONFIG_KEY,
-    MAX_TOKENS_CONFIG_KEY,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY,
     PROMPT_TEMPLATE_CONFIG_KEY,
     ROUTE_TO_CALM_SLOT,
     TEMPERATURE_CONFIG_KEY,
@@ -81,7 +81,7 @@ DEFAULT_LLM_CONFIG = {
     PROVIDER_CONFIG_KEY: OPENAI_PROVIDER,
     MODEL_CONFIG_KEY: MODEL_NAME_GPT_4O_2024_11_20,
     TEMPERATURE_CONFIG_KEY: 0.0,
-    MAX_TOKENS_CONFIG_KEY: DEFAULT_OPENAI_MAX_GENERATED_TOKENS,
+    MAX_COMPLETION_TOKENS_CONFIG_KEY: DEFAULT_OPENAI_MAX_GENERATED_TOKENS,
     TIMEOUT_CONFIG_KEY: 7,
 }
 

{rasa_pro-3.12.13 → rasa_pro-3.12.15}/rasa/dialogue_understanding/processor/command_processor_component.py

@@ -1,6 +1,6 @@
 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 +37,8 @@ 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-3.12.13 → rasa_pro-3.12.15}/rasa/dialogue_understanding_test/du_test_runner.py

@@ -33,6 +33,7 @@ from rasa.e2e_test.e2e_test_runner import E2ETestRunner
 from rasa.shared.core.events import UserUttered
 from rasa.shared.core.trackers import DialogueStateTracker
 from rasa.shared.nlu.constants import PREDICTED_COMMANDS, PROMPTS
+from rasa.shared.utils.llm import create_tracker_for_user_step
 from rasa.utils.endpoints import EndpointConfig
 
 structlogger = structlog.get_logger()
@@ -178,8 +179,9 @@ class DialogueUnderstandingTestRunner:
             # create and save the tracker at the time just
             # before the user message was sent
             step_sender_id = f"{sender_id}_{user_step_index}"
-            await self._create_tracker_for_user_step(
+            await create_tracker_for_user_step(
                 step_sender_id,
+                self.agent,
                 test_case_tracker,
                 user_uttered_event_indices[user_step_index],
             )
@@ -280,26 +282,6 @@ class DialogueUnderstandingTestRunner:
 
         return user_uttered_event
 
-    async def _create_tracker_for_user_step(
-        self,
-        step_sender_id: str,
-        test_case_tracker: DialogueStateTracker,
-        index_user_uttered_event: int,
-    ) -> None:
-        """Creates a tracker for the user step."""
-        tracker = test_case_tracker.copy()
-        # modify the sender id so that the test case tracker is not overwritten
-        tracker.sender_id = step_sender_id
-
-        if tracker.events:
-            # get timestamp of the event just before the user uttered event
-            timestamp = tracker.events[index_user_uttered_event - 1].timestamp
-            # revert the tracker to the event just before the user uttered event
-            tracker = tracker.travel_back_in_time(timestamp)
-
-        # store the tracker with the unique sender id
-        await self.agent.tracker_store.save(tracker)
-
     async def _send_user_message(
         self,
         sender_id: str,

{rasa_pro-3.12.13 → rasa_pro-3.12.15}/rasa/dialogue_understanding_test/test_case_simulation/test_case_tracker_simulator.py

@@ -1,4 +1,3 @@
-from datetime import datetime
 from typing import List, Optional
 
 import structlog
@@ -24,6 +23,7 @@ from rasa.shared.core.constants import SlotMappingType
 from rasa.shared.core.events import BotUttered, SlotSet, UserUttered
 from rasa.shared.core.trackers import DialogueStateTracker
 from rasa.shared.nlu.constants import COMMANDS, ENTITIES, INTENT
+from rasa.shared.utils.llm import generate_sender_id
 
 structlogger = structlog.get_logger()
 
@@ -52,7 +52,7 @@ class TestCaseTrackerSimulator:
         self.test_case = test_case
         self.output_channel = output_channel or CollectingOutputChannel()
 
-        self.sender_id = self._generate_sender_id()
+        self.sender_id = generate_sender_id(self.test_case.name)
 
     async def simulate_test_case(
         self,
@@ -150,10 +150,6 @@ class TestCaseTrackerSimulator:
             user_uttered_event_indices=user_uttered_event_indices,
         )
 
-    def _generate_sender_id(self) -> str:
-        # add timestamp suffix to ensure sender_id is unique
-        return f"{self.test_case.name}_{datetime.now()}"
-
     @staticmethod
     async def _get_latest_user_uttered_event_index(
         tracker: DialogueStateTracker, user_uttered_event_indices: List[int]