rasa-pro 3.13.0.dev8__py3-none-any.whl → 3.13.0.dev9__py3-none-any.whl

rasa/core/information_retrieval/ingestion/faq_parser.py

@@ -0,0 +1,158 @@
+"""Utilities for parsing FAQ-style documents (Q/A pairs) used in extractive search."""
+
+import re
+from collections import defaultdict
+from typing import TYPE_CHECKING, List
+
+import structlog
+
+from rasa.shared.constants import (
+    DOCUMENT_TYPE_FAQ,
+    FAQ_DOCUMENT_ENTRY_SEPARATOR,
+    FAQ_DOCUMENT_LINE_SEPARATOR,
+    FAQ_DOCUMENT_METADATA_ANSWER,
+    FAQ_DOCUMENT_METADATA_TITLE,
+    FAQ_DOCUMENT_METADATA_TYPE,
+    FAQ_INPUT_DATA_ANSWER_LINE_PREFIX,
+    FAQ_INPUT_DATA_QUESTION_LINE_PREFIX,
+)
+
+if TYPE_CHECKING:
+    from langchain.schema import Document
+
+_FAQ_PAIR_PATTERN = re.compile(
+    rf"{re.escape(FAQ_INPUT_DATA_QUESTION_LINE_PREFIX)}\s*"
+    rf"(?P<question>.*?)\s*{FAQ_DOCUMENT_LINE_SEPARATOR}\s*"
+    rf"{re.escape(FAQ_INPUT_DATA_ANSWER_LINE_PREFIX)}\s*"
+    rf"(?P<answer>.*)",
+    re.DOTALL,
+)
+
+
+structlogger = structlog.get_logger()
+
+
+def _format_faq_documents(documents: List["Document"]) -> List["Document"]:
+    """Splits each loaded file into individual FAQs.
+
+    Args:
+        documents: Documents representing whole files containing FAQs.
+
+    Returns:
+        List of Document objects, each containing a separate FAQ.
+
+    Examples:
+        An example of a file containing FAQs:
+
+        Q: Who is Finley?
+        A: Finley is your smart assistant for the FinX App. You can add him to your
+           favorite messenger and tell him what you need help with.
+
+        Q: How does Finley work?
+        A: Finley is powered by the latest chatbot technology leveraging a unique
+           interplay of large language models and secure logic.
+
+    More details in documentation: https://rasa.com/docs/reference/config/policies/extractive-search/
+    """
+    structured_faqs = []
+    from langchain.schema import Document
+
+    for document in documents:
+        chunks = document.page_content.strip().split(FAQ_DOCUMENT_ENTRY_SEPARATOR)
+
+        for chunk in chunks:
+            match = _FAQ_PAIR_PATTERN.match(chunk.strip())
+
+            if not match:
+                structlogger.warning(
+                    "faq_parser.format_faq_documents.invalid_chunk_skipped",
+                    event_info=(
+                        "Chunk does not match expected QA format. "
+                        "Please refer to the documentation: "
+                        "https://rasa.com/docs/reference/config/"
+                        "policies/extractive-search/"
+                    ),
+                    chunk_preview=chunk[:100],
+                )
+                continue
+
+            question = match.group("question").strip()
+            answer = match.group("answer").strip()
+            title = _sanitize_title(question)
+
+            formatted_document = Document(
+                page_content=question,
+                metadata={
+                    FAQ_DOCUMENT_METADATA_TITLE: title,
+                    FAQ_DOCUMENT_METADATA_TYPE: DOCUMENT_TYPE_FAQ,
+                    FAQ_DOCUMENT_METADATA_ANSWER: answer,
+                },
+            )
+
+            structured_faqs.append(formatted_document)
+
+            structlogger.debug(
+                "faq_parser.format_faq_documents.parsed_chunk",
+                event_info="Parsed chunk.",
+                title=title,
+                question=question,
+                answer=answer,
+                parsed_chunk_preview=chunk[:100],
+            )
+
+    structlogger.debug(
+        "faq_parser.format_faq_documents.parsed_chunks",
+        event_info=(
+            f"Retrieved {len(structured_faqs)} FAQ pair(s)"
+            f"from {len(documents)} document(s)."
+        ),
+        num_structured_faqs=len(structured_faqs),
+        num_documents=len(documents),
+    )
+    _check_and_parsed_faq_documents_for_duplicates(structured_faqs)
+    return structured_faqs
+
+
+def _sanitize_title(title: str) -> str:
+    title = title.lower()
+    # Remove all whitespaces with "_"
+    title = re.sub(r"\s+", "_", title)
+    # Remove all non alpha-numeric characters
+    title = re.sub(r"[^\w]", "", title)
+    # Collapse multiple "_"
+    title = re.sub(r"_+", "_", title)
+    # Clean up edges
+    return title.strip("_")
+
+
+def _check_and_parsed_faq_documents_for_duplicates(documents: List["Document"]) -> None:
+    seen_qa_pairs = set()
+    seen_questions: defaultdict = defaultdict(list)
+
+    for doc in documents:
+        question = doc.page_content.strip()
+        answer = doc.metadata.get(FAQ_DOCUMENT_METADATA_ANSWER, "").strip()
+
+        if not question or not answer:
+            continue
+
+        if (question, answer) in seen_qa_pairs:
+            structlogger.warning(
+                "faq_parser.duplicate_qa_pair_found",
+                event_info="Duplicate QA pair found.",
+                question=question,
+                answer_preview=answer,
+            )
+            continue
+
+        if question in seen_questions and seen_questions[question] != answer:
+            structlogger.warning(
+                "faq_parser.inconsistent_answer",
+                event_info="Duplicate question with different answer found.",
+                question=question,
+                previous_answers=seen_questions[question],
+                new_answer=answer,
+            )
+
+        seen_qa_pairs.add((question, answer))
+        seen_questions[question].append(answer)

rasa/core/nlg/contextual_response_rephraser.py

@@ -17,6 +17,7 @@ from rasa.shared.constants import (
     MODEL_NAME_CONFIG_KEY,
     OPENAI_PROVIDER,
     PROMPT_CONFIG_KEY,
+    PROMPT_TEMPLATE_CONFIG_KEY,
     PROVIDER_CONFIG_KEY,
     TEMPERATURE_CONFIG_KEY,
     TIMEOUT_CONFIG_KEY,
@@ -38,6 +39,7 @@ from rasa.shared.utils.llm import (
     DEFAULT_OPENAI_GENERATE_MODEL_NAME,
     DEFAULT_OPENAI_MAX_GENERATED_TOKENS,
     USER,
+    check_prompt_config_keys_and_warn_if_deprecated,
     combine_custom_and_default_config,
     get_prompt_template,
     llm_factory,
@@ -110,8 +112,15 @@ class ContextualResponseRephraser(
         super().__init__(domain.responses)
 
         self.nlg_endpoint = endpoint_config
+
+        # Warn if the prompt config key is used to set the prompt template
+        check_prompt_config_keys_and_warn_if_deprecated(
+            self.nlg_endpoint.kwargs, "contextual_response_rephraser"
+        )
+
         self.prompt_template = get_prompt_template(
-            self.nlg_endpoint.kwargs.get(PROMPT_CONFIG_KEY),
+            self.nlg_endpoint.kwargs.get(PROMPT_TEMPLATE_CONFIG_KEY)
+            or self.nlg_endpoint.kwargs.get(PROMPT_CONFIG_KEY),
             DEFAULT_RESPONSE_VARIATION_PROMPT_TEMPLATE,
             log_source_component=ContextualResponseRephraser.__name__,
             log_source_method=LOG_COMPONENT_SOURCE_METHOD_INIT,

rasa/core/policies/enterprise_search_policy.py

@@ -95,6 +95,7 @@ from rasa.shared.utils.io import deep_container_fingerprint
 from rasa.shared.utils.llm import (
     DEFAULT_OPENAI_CHAT_MODEL_NAME,
     DEFAULT_OPENAI_EMBEDDING_MODEL_NAME,
+    check_prompt_config_keys_and_warn_if_deprecated,
     embedder_factory,
     get_prompt_template,
     llm_factory,
@@ -246,7 +247,9 @@ class EnterpriseSearchPolicy(LLMHealthCheckMixin, EmbeddingsHealthCheckMixin, Po
         super().__init__(config, model_storage, resource, execution_context, featurizer)
 
         # Check for deprecated keys and issue a warning if those are used
-        self._check_config_keys_and_warn_if_deprecated()
+        check_prompt_config_keys_and_warn_if_deprecated(
+            config, "enterprise_search_policy"
+        )
         # Check for mutual exclusivity of extractive and generative search
         self._check_and_warn_mutual_exclusivity_of_extractive_and_generative_search()
 
@@ -307,34 +310,6 @@ class EnterpriseSearchPolicy(LLMHealthCheckMixin, EmbeddingsHealthCheckMixin, Po
             self.config, LOG_COMPONENT_SOURCE_METHOD_INIT
         )
 
-    def _check_config_keys_and_warn_if_deprecated(self) -> None:
-        """Checks and warns about deprecated config parameters."""
-        if (
-            PROMPT_CONFIG_KEY in self.config
-            and PROMPT_TEMPLATE_CONFIG_KEY in self.config
-        ):
-            structlogger.warning(
-                "enterprise_search_policy.init"
-                ".both_deprecated_and_non_deprecated_config_keys_used_at_the_same_time",
-                event_info=(
-                    f"Both '{PROMPT_CONFIG_KEY}' and '{PROMPT_TEMPLATE_CONFIG_KEY}' "
-                    f"are present in the config. '{PROMPT_CONFIG_KEY}' will be ignored "
-                    f"in favor of {PROMPT_TEMPLATE_CONFIG_KEY}."
-                ),
-            )
-
-        # 'prompt' config key is deprecated in favor of 'prompt_template'
-        if PROMPT_CONFIG_KEY in self.config:
-            structlogger.warning(
-                "enterprise_search_policy.init.deprecated_config_key",
-                event_info=(
-                    f"The config parameter '{PROMPT_CONFIG_KEY}' is deprecated "
-                    "and will be removed in Rasa 4.0.0. "
-                    f"Please use the config parameter '{PROMPT_TEMPLATE_CONFIG_KEY}'"
-                    f"instead. "
-                ),
-            )
-
     def _check_and_warn_mutual_exclusivity_of_extractive_and_generative_search(
         self,
     ) -> None:
@@ -464,6 +439,7 @@ class EnterpriseSearchPolicy(LLMHealthCheckMixin, EmbeddingsHealthCheckMixin, Po
                     embeddings=embeddings,
                     index_path=path,
                     create_index=True,
+                    parse_as_faq_pairs=not self.use_llm,
                 )
         else:
             structlogger.info(
@@ -917,6 +893,9 @@ class EnterpriseSearchPolicy(LLMHealthCheckMixin, EmbeddingsHealthCheckMixin, Po
                     index_path=path,
                     docs_folder=None,
                     create_index=False,
+                    parse_as_faq_pairs=not config.get(
+                        USE_LLM_PROPERTY, DEFAULT_USE_LLM_PROPERTY
+                    ),
                 )
         else:
             vector_store = create_from_endpoint_config(

rasa/core/policies/intentless_policy.py

@@ -38,6 +38,7 @@ from rasa.shared.constants import (
     MODEL_NAME_CONFIG_KEY,
     OPENAI_PROVIDER,
     PROMPT_CONFIG_KEY,
+    PROMPT_TEMPLATE_CONFIG_KEY,
     PROVIDER_CONFIG_KEY,
     TEMPERATURE_CONFIG_KEY,
     TIMEOUT_CONFIG_KEY,
@@ -56,7 +57,10 @@ from rasa.shared.providers.embedding._langchain_embedding_client_adapter import
     _LangchainEmbeddingClientAdapter,
 )
 from rasa.shared.providers.llm.llm_client import LLMClient
-from rasa.shared.utils.constants import LOG_COMPONENT_SOURCE_METHOD_FINGERPRINT_ADDON
+from rasa.shared.utils.constants import (
+    LOG_COMPONENT_SOURCE_METHOD_FINGERPRINT_ADDON,
+    LOG_COMPONENT_SOURCE_METHOD_INIT,
+)
 from rasa.shared.utils.health_check.embeddings_health_check_mixin import (
     EmbeddingsHealthCheckMixin,
 )
@@ -68,6 +72,7 @@ from rasa.shared.utils.llm import (
     DEFAULT_OPENAI_EMBEDDING_MODEL_NAME,
     DEFAULT_OPENAI_MAX_GENERATED_TOKENS,
     USER,
+    check_prompt_config_keys_and_warn_if_deprecated,
     combine_custom_and_default_config,
     embedder_factory,
     get_prompt_template,
@@ -119,9 +124,12 @@ DEFAULT_EMBEDDINGS_CONFIG = {
     MODEL_CONFIG_KEY: DEFAULT_OPENAI_EMBEDDING_MODEL_NAME,
 }
 
-DEFAULT_INTENTLESS_PROMPT_TEMPLATE = importlib.resources.open_text(
+DEFAULT_INTENTLESS_PROMPT_TEMPLATE_FILE_NAME = importlib.resources.open_text(
     "rasa.core.policies", "intentless_prompt_template.jinja2"
 ).name
+DEFAULT_INTENTLESS_PROMPT_TEMPLATE = importlib.resources.read_text(
+    "rasa.core.policies", "intentless_prompt_template.jinja2"
+)
 
 INTENTLESS_PROMPT_TEMPLATE_FILE_NAME = "intentless_policy_prompt.jinja2"
 INTENTLESS_CONFIG_FILE_NAME = "config.json"
@@ -345,7 +353,7 @@ 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,
-            PROMPT_CONFIG_KEY: DEFAULT_INTENTLESS_PROMPT_TEMPLATE,
+            PROMPT_TEMPLATE_CONFIG_KEY: None,  # TODO: remove in Rasa 4.0.0
         }
 
     @staticmethod
@@ -402,11 +410,43 @@ class IntentlessPolicy(LLMHealthCheckMixin, EmbeddingsHealthCheckMixin, Policy):
         self.response_index = responses_docsearch
         self.conversation_samples_index = samples_docsearch
         self.embedder = self._create_plain_embedder(config)
-        self.prompt_template = prompt_template or rasa.shared.utils.io.read_file(
-            self.config[PROMPT_CONFIG_KEY]
+
+        # Warn if the prompt config key is used to set the prompt template
+        check_prompt_config_keys_and_warn_if_deprecated(config, "intentless_policy")
+
+        self.prompt_template = prompt_template or self._resolve_prompt_template(
+            config, LOG_COMPONENT_SOURCE_METHOD_INIT
         )
         self.trace_prompt_tokens = self.config.get("trace_prompt_tokens", False)
 
+    @classmethod
+    def _resolve_prompt_template(
+        cls: Any,
+        config: dict,
+        log_source_method: str,
+    ) -> str:
+        """Resolves the prompt template from the config.
+
+        Args:
+            config: The config to resolve the prompt template from.
+            log_source_method: The method from which the prompt template is resolved.
+
+        Returns:
+            The resolved prompt template.
+        """
+        # Prefer prompt template over prompt config key.
+        prompt_template_file = (
+            config.get(PROMPT_TEMPLATE_CONFIG_KEY)
+            or config.get(PROMPT_CONFIG_KEY)
+            or DEFAULT_INTENTLESS_PROMPT_TEMPLATE_FILE_NAME
+        )
+        return get_prompt_template(
+            prompt_template_file,
+            DEFAULT_INTENTLESS_PROMPT_TEMPLATE,
+            log_source_component=IntentlessPolicy.__name__,
+            log_source_method=log_source_method,
+        )
+
     @classmethod
     def _create_plain_embedder(cls, config: Dict[Text, Any]) -> Embeddings:
         """Creates an embedder that uses the OpenAI API.
@@ -945,11 +985,8 @@ class IntentlessPolicy(LLMHealthCheckMixin, EmbeddingsHealthCheckMixin, Policy):
     @classmethod
     def fingerprint_addon(cls, config: Dict[str, Any]) -> Optional[str]:
         """Add a fingerprint of intentless policy for the graph."""
-        prompt_template = get_prompt_template(
-            config.get(PROMPT_CONFIG_KEY),
-            DEFAULT_INTENTLESS_PROMPT_TEMPLATE,
-            log_source_component=IntentlessPolicy.__name__,
-            log_source_method=LOG_COMPONENT_SOURCE_METHOD_FINGERPRINT_ADDON,
+        prompt_template = cls._resolve_prompt_template(
+            config, LOG_COMPONENT_SOURCE_METHOD_FINGERPRINT_ADDON
         )
 
         llm_config = resolve_model_client_config(

rasa/dialogue_understanding/coexistence/llm_based_router.py

@@ -15,9 +15,7 @@ from rasa.dialogue_understanding.coexistence.constants import (
 )
 from rasa.dialogue_understanding.commands import Command, SetSlotCommand
 from rasa.dialogue_understanding.commands.noop_command import NoopCommand
-from rasa.dialogue_understanding.generator.constants import (
-    LLM_CONFIG_KEY,
-)
+from rasa.dialogue_understanding.generator.constants import LLM_CONFIG_KEY
 from rasa.engine.graph import ExecutionContext, GraphComponent
 from rasa.engine.recipes.default_recipe import DefaultV1Recipe
 from rasa.engine.storage.resource import Resource
@@ -28,6 +26,7 @@ from rasa.shared.constants import (
     MODEL_CONFIG_KEY,
     OPENAI_PROVIDER,
     PROMPT_CONFIG_KEY,
+    PROMPT_TEMPLATE_CONFIG_KEY,
     PROVIDER_CONFIG_KEY,
     ROUTE_TO_CALM_SLOT,
     TEMPERATURE_CONFIG_KEY,
@@ -46,6 +45,7 @@ from rasa.shared.utils.health_check.llm_health_check_mixin import LLMHealthCheck
 from rasa.shared.utils.io import deep_container_fingerprint
 from rasa.shared.utils.llm import (
     DEFAULT_OPENAI_CHAT_MODEL_NAME,
+    check_prompt_config_keys_and_warn_if_deprecated,
     get_prompt_template,
     llm_factory,
     resolve_model_client_config,
@@ -90,7 +90,7 @@ class LLMBasedRouter(LLMHealthCheckMixin, GraphComponent):
     def get_default_config() -> Dict[str, Any]:
         """The component's default config (see parent class for full docstring)."""
         return {
-            PROMPT_CONFIG_KEY: None,
+            PROMPT_TEMPLATE_CONFIG_KEY: None,  # TODO: remove in Rasa 4.0.0
             CALM_ENTRY: {STICKY: None},
             NLU_ENTRY: {
                 NON_STICKY: "handles chitchat",
@@ -111,10 +111,13 @@ class LLMBasedRouter(LLMHealthCheckMixin, GraphComponent):
             self.config.get(LLM_CONFIG_KEY), LLMBasedRouter.__name__
         )
 
+        # Warn if the prompt config key is used to set the prompt template
+        check_prompt_config_keys_and_warn_if_deprecated(config, "llm_based_router")
+
         self.prompt_template = (
             prompt_template
             or get_prompt_template(
-                config.get(PROMPT_CONFIG_KEY),
+                config.get(PROMPT_TEMPLATE_CONFIG_KEY) or config.get(PROMPT_CONFIG_KEY),
                 DEFAULT_COMMAND_PROMPT_TEMPLATE,
                 log_source_component=LLMBasedRouter.__name__,
                 log_source_method=LOG_COMPONENT_SOURCE_METHOD_INIT,
@@ -327,7 +330,7 @@ class LLMBasedRouter(LLMHealthCheckMixin, GraphComponent):
     def fingerprint_addon(cls, config: Dict[str, Any]) -> Optional[str]:
         """Add a fingerprint of llm based router for the graph."""
         prompt_template = get_prompt_template(
-            config.get(PROMPT_CONFIG_KEY),
+            config.get(PROMPT_TEMPLATE_CONFIG_KEY) or config.get(PROMPT_CONFIG_KEY),
             DEFAULT_COMMAND_PROMPT_TEMPLATE,
             log_source_component=LLMBasedRouter.__name__,
             log_source_method=LOG_COMPONENT_SOURCE_METHOD_FINGERPRINT_ADDON,

rasa/dialogue_understanding/commands/cancel_flow_command.py

@@ -95,7 +95,9 @@ class CancelFlowCommand(Command):
         original_stack = original_tracker.stack
 
         applied_events: List[Event] = []
-        user_frame = top_user_flow_frame(original_stack)
+        user_frame = top_user_flow_frame(
+            original_stack, ignore_call_and_link_frames=False
+        )
         current_flow = user_frame.flow(all_flows) if user_frame else None
 
         if not current_flow:

rasa/dialogue_understanding/commands/correct_slots_command.py

@@ -274,9 +274,7 @@ class CorrectSlotsCommand(Command):
             # we shouldn't end up here as a correction shouldn't be triggered
             # if we are not in any flow. but just in case we do, we
             # just skip the command.
-            structlogger.warning(
-                "command_executor.correct_slots.no_active_flow", command=self
-            )
+            structlogger.warning("command_executor.correct_slots.no_active_flow")
             return []
 
         structlogger.debug("command_executor.correct_slots", command=self)

rasa/dialogue_understanding/generator/nlu_command_adapter.py

@@ -190,9 +190,9 @@ class NLUCommandAdapter(GraphComponent, CommandGenerator):
         if len(commands) > 1:
             structlogger.warning(
                 "nlu_command_adapter.predict_commands",
-                messag=f"Two many flows found that are triggered by the "
+                message=f"Too many flows found that are triggered by the "
                 f"intent '{message.get(INTENT)['name']}'. Take the first one.",
-                commands=commands,
+                commands=[command.__class__.__name__ for command in commands],
             )
             commands = [commands[0]]
 

rasa/dialogue_understanding/generator/single_step/single_step_based_llm_command_generator.py

@@ -113,7 +113,7 @@ class SingleStepBasedLLMCommandGenerator(LLMBasedCommandGenerator, ABC):
     def get_default_config() -> Dict[str, Any]:
         """The component's default config (see parent class for full docstring)."""
         return {
-            PROMPT_TEMPLATE_CONFIG_KEY: None,
+            PROMPT_TEMPLATE_CONFIG_KEY: None,  # TODO: remove in Rasa 4.0.0
             USER_INPUT_CONFIG_KEY: None,
             LLM_CONFIG_KEY: None,
             FLOW_RETRIEVAL_KEY: FlowRetrieval.get_default_config(),
@@ -212,7 +212,7 @@ class SingleStepBasedLLMCommandGenerator(LLMBasedCommandGenerator, ABC):
                 event_info=(
                     "ProviderClientAPIException occurred while predicting commands."
                 ),
-                commands=commands,
+                commands=commands,  # no PII
             )
 
         if not commands and not prior_commands:

rasa/dialogue_understanding/generator/single_step/single_step_llm_command_generator.py

@@ -16,7 +16,10 @@ from rasa.shared.constants import (
     PROMPT_CONFIG_KEY,
     PROMPT_TEMPLATE_CONFIG_KEY,
 )
-from rasa.shared.utils.llm import get_prompt_template
+from rasa.shared.utils.llm import (
+    check_prompt_config_keys_and_warn_if_deprecated,
+    get_prompt_template,
+)
 
 DEFAULT_COMMAND_PROMPT_TEMPLATE = importlib.resources.read_text(
     "rasa.dialogue_understanding.generator.prompt_templates",
@@ -52,16 +55,10 @@ class SingleStepLLMCommandGenerator(SingleStepBasedLLMCommandGenerator):
             **kwargs,
         )
 
-        # Set the prompt template
-        if config.get(PROMPT_CONFIG_KEY):
-            structlogger.warning(
-                "single_step_llm_command_generator.init",
-                event_info=(
-                    "The config parameter 'prompt' is deprecated "
-                    "and will be removed in Rasa 4.0.0. "
-                    "Please use the config parameter 'prompt_template' instead. "
-                ),
-            )
+        # Warn if the prompt config key is used to set the prompt template
+        check_prompt_config_keys_and_warn_if_deprecated(
+            config, "single_step_llm_command_generator"
+        )
 
     @staticmethod
     def get_component_command_syntax_version() -> CommandSyntaxVersion:

rasa/dialogue_understanding/patterns/cancel.py

@@ -90,7 +90,7 @@ class ActionCancelFlow(action.Action):
             return []
 
         if not isinstance(top, CancelPatternFlowStackFrame):
-            structlogger.warning("action.cancel_flow.no_cancel_frame", top=top)
+            structlogger.warning("action.cancel_flow.no_cancel_frame")
             return []
 
         for canceled_frame_id in top.canceled_frames:
@@ -105,7 +105,6 @@ class ActionCancelFlow(action.Action):
             else:
                 structlogger.warning(
                     "action.cancel_flow.frame_not_found",
-                    dialogue_stack=stack,
                     frame_id=canceled_frame_id,
                 )
 

rasa/dialogue_understanding/patterns/clarify.py

@@ -89,7 +89,7 @@ class ActionClarifyFlows(action.Action):
             return []
 
         if not isinstance(top, ClarifyPatternFlowStackFrame):
-            structlogger.warning("action.clarify_flows.no_clarification_frame", top=top)
+            structlogger.warning("action.clarify_flows.no_clarification_frame")
             return []
 
         options_string = self.assemble_options_string(top.names)

rasa/dialogue_understanding/patterns/correction.py

@@ -114,7 +114,8 @@ class ActionCorrectFlowSlot(action.Action):
 
         if not isinstance(top, CorrectionPatternFlowStackFrame):
             structlogger.warning(
-                "action.correct_flow_slot.no_correction_frame", top=top
+                "action.correct_flow_slot.no_correction_frame",
+                top=top,  # no PII
             )
             return []
 
@@ -246,7 +247,6 @@ def reset_stack_on_tracker_to_prior_state(
             "action.correct_flow_slot.no_target_frame_found",
             reset_step_id=reset_step_id,
             reset_flow_id=reset_flow_id,
-            stack_to_reset_to=stack_to_reset_to,
         )
         return tracker.stack
 

rasa/dialogue_understanding/processor/command_processor.py

@@ -348,7 +348,6 @@ def get_current_collect_step(
         # but no flow that triggered it. this should never happen.
         structlogger.warning(
             "command_processor.get_current_collect_step.no_flow_on_stack",
-            stack=dialogue_stack,
         )
         return None
 
@@ -358,7 +357,7 @@ def get_current_collect_step(
         # step from it
         structlogger.warning(
             "command_processor.get_current_collect_step.no_step_for_frame",
-            frame=frame_that_triggered_collect_infos,
+            frame=frame_that_triggered_collect_infos.frame_id,
         )
         return None
 
@@ -724,7 +723,7 @@ def clean_up_chitchat_command(
         )
         structlogger.warn(
             "command_processor.clean_up_chitchat_command.pattern_chitchat_not_found",
-            command=resulting_commands[0],
+            command=resulting_commands[0],  # no PII
         )
         return resulting_commands
 
@@ -742,7 +741,7 @@ def clean_up_chitchat_command(
         )
         structlogger.warn(
             "command_processor.clean_up_chitchat_command.replace_chitchat_answer_with_cannot_handle",
-            command=resulting_commands[0],
+            command=resulting_commands[0],  # no PII
             pattern_chitchat_uses_action_trigger_chitchat=has_action_trigger_chitchat,
             defined_intentless_policy_in_config=defines_intentless_policy,
         )

rasa/dialogue_understanding/stack/utils.py

@@ -209,7 +209,9 @@ def get_collect_steps_excluding_ask_before_filling_for_active_flow(
         All collect steps that are part of the current active flow,
         excluding the collect steps that have to be asked before filling.
     """
-    active_frame = top_user_flow_frame(dialogue_stack)
+    active_frame = top_user_flow_frame(
+        dialogue_stack, ignore_call_and_link_frames=False
+    )
     if active_frame is None:
         return set()
     active_flow = active_frame.flow(all_flows)

rasa/engine/graph.py

@@ -500,9 +500,9 @@ class GraphNode:
                 structlogger.warning(
                     "graph.node.input_not_resolved",
                     node_name=self._node_name,
-                    input_name=i,
+                    input_name=i,  # no PII
                     event_info=(
-                        "Node input was not resolved, there is no putput. "
+                        "Node input was not resolved, there is no output. "
                         "Another component should have provided this as an output."
                     ),
                 )

rasa/llm_fine_tuning/paraphrasing/conversation_rephraser.py

@@ -22,11 +22,7 @@ from rasa.shared.constants import (
 from rasa.shared.exceptions import ProviderClientAPIException
 from rasa.shared.providers.mappings import OPENAI_PROVIDER
 from rasa.shared.utils.constants import LOG_COMPONENT_SOURCE_METHOD_INIT
-from rasa.shared.utils.llm import (
-    USER,
-    get_prompt_template,
-    llm_factory,
-)
+from rasa.shared.utils.llm import USER, get_prompt_template, llm_factory
 
 SEPARATOR = "\n\n"
 BACKUP_SEPARATOR = "\nUSER:"

rasa/shared/constants.py

@@ -295,6 +295,7 @@ CONTEXT = "context"
 
 RASA_PATTERN_INTERNAL_ERROR = "pattern_internal_error"
 RASA_PATTERN_HUMAN_HANDOFF = "pattern_human_handoff"
+RASA_PATTERN_CHITCHAT = "pattern_chitchat"
 
 RASA_INTERNAL_ERROR_PREFIX = "rasa_internal_error_"
 RASA_PATTERN_INTERNAL_ERROR_DEFAULT = RASA_INTERNAL_ERROR_PREFIX + "default"
@@ -345,3 +346,13 @@ ROLE_SYSTEM = "system"
 # Used for key values in ValidateSlotPatternFlowStackFrame
 REFILL_UTTER = "refill_utter"
 REJECTIONS = "rejections"
+
+# Constants for extractive search FAQ parsing (QA pairs from input documents)
+FAQ_DOCUMENT_METADATA_TITLE = "title"
+FAQ_DOCUMENT_METADATA_ANSWER = "answer"
+FAQ_DOCUMENT_METADATA_TYPE = "type"
+DOCUMENT_TYPE_FAQ = "faq"
+FAQ_INPUT_DATA_QUESTION_LINE_PREFIX = "Q:"
+FAQ_INPUT_DATA_ANSWER_LINE_PREFIX = "A:"
+FAQ_DOCUMENT_ENTRY_SEPARATOR = "\n\n"
+FAQ_DOCUMENT_LINE_SEPARATOR = "\n"