Files
core/homeassistant/components/conversation/default_agent.py
2025-10-07 14:49:11 +02:00

1882 lines
66 KiB
Python

"""Standard conversation implementation for Home Assistant."""
from __future__ import annotations
import asyncio
from collections import OrderedDict
from collections.abc import Callable, Iterable
from dataclasses import dataclass
from enum import Enum, auto
import logging
from pathlib import Path
import time
from typing import IO, Any, cast
from hassil.expression import Expression, Group, ListReference, TextChunk
from hassil.fuzzy import FuzzyNgramMatcher, SlotCombinationInfo
from hassil.intents import (
Intent,
IntentData,
Intents,
SlotList,
TextSlotList,
TextSlotValue,
WildcardSlotList,
)
from hassil.models import MatchEntity
from hassil.ngram import Sqlite3NgramModel
from hassil.recognize import (
MISSING_ENTITY,
RecognizeResult,
recognize_all,
recognize_best,
)
from hassil.string_matcher import UnmatchedRangeEntity, UnmatchedTextEntity
from hassil.trie import Trie
from hassil.util import merge_dict, remove_punctuation
from home_assistant_intents import (
ErrorKey,
FuzzyConfig,
FuzzyLanguageResponses,
LanguageScores,
get_fuzzy_config,
get_fuzzy_language,
get_intents,
get_language_scores,
get_languages,
)
import yaml
from homeassistant.components.homeassistant.exposed_entities import (
async_listen_entity_updates,
async_should_expose,
)
from homeassistant.const import EVENT_STATE_CHANGED, MATCH_ALL
from homeassistant.core import (
Event,
EventStateChangedData,
HomeAssistant,
State,
callback,
)
from homeassistant.helpers import (
area_registry as ar,
config_validation as cv,
device_registry as dr,
entity_registry as er,
floor_registry as fr,
intent,
start as ha_start,
template,
translation,
)
from homeassistant.helpers.entity_component import EntityComponent
from homeassistant.helpers.event import async_track_state_added_domain
from homeassistant.util import language as language_util
from homeassistant.util.json import JsonObjectType, json_loads_object
from .agent_manager import get_agent_manager
from .chat_log import AssistantContent, ChatLog
from .const import DOMAIN, ConversationEntityFeature
from .entity import ConversationEntity
from .models import ConversationInput, ConversationResult
from .trace import ConversationTraceEventType, async_conversation_trace_append
from .trigger import TriggerDetails
_LOGGER = logging.getLogger(__name__)
_DEFAULT_ERROR_TEXT = "Sorry, I couldn't understand that"
_ENTITY_REGISTRY_UPDATE_FIELDS = ["aliases", "name", "original_name"]
_DEFAULT_EXPOSED_ATTRIBUTES = {"device_class"}
METADATA_CUSTOM_SENTENCE = "hass_custom_sentence"
METADATA_CUSTOM_FILE = "hass_custom_file"
METADATA_FUZZY_MATCH = "hass_fuzzy_match"
ERROR_SENTINEL = object()
def json_load(fp: IO[str]) -> JsonObjectType:
"""Wrap json_loads for get_intents."""
return json_loads_object(fp.read())
@dataclass(slots=True)
class LanguageIntents:
"""Loaded intents for a language."""
intents: Intents
intents_dict: dict[str, Any]
intent_responses: dict[str, Any]
error_responses: dict[str, Any]
language_variant: str | None
fuzzy_matcher: FuzzyNgramMatcher | None = None
fuzzy_responses: FuzzyLanguageResponses | None = None
@dataclass(slots=True)
class SentenceTriggerResult:
"""Result when matching a sentence trigger in an automation."""
sentence: str
sentence_template: str | None
matched_triggers: dict[int, RecognizeResult]
class IntentMatchingStage(Enum):
"""Stages of intent matching."""
EXPOSED_ENTITIES_ONLY = auto()
"""Match against exposed entities only."""
FUZZY = auto()
"""Use fuzzy matching to guess intent."""
UNEXPOSED_ENTITIES = auto()
"""Match against unexposed entities in Home Assistant."""
UNKNOWN_NAMES = auto()
"""Capture names that are not known to Home Assistant."""
@dataclass(frozen=True)
class IntentCacheKey:
"""Key for IntentCache."""
text: str
"""User input text."""
language: str
"""Language of text."""
satellite_id: str | None
"""Satellite id from user input."""
@dataclass(frozen=True)
class IntentCacheValue:
"""Value for IntentCache."""
result: RecognizeResult | None
"""Result of intent recognition."""
stage: IntentMatchingStage
"""Stage where result was found."""
class IntentCache:
"""LRU cache for intent recognition results."""
def __init__(self, capacity: int) -> None:
"""Initialize cache."""
self.cache: OrderedDict[IntentCacheKey, IntentCacheValue] = OrderedDict()
self.capacity = capacity
def get(self, key: IntentCacheKey) -> IntentCacheValue | None:
"""Get value for cache or None."""
if key not in self.cache:
return None
# Move the key to the end to show it was recently used
self.cache.move_to_end(key)
return self.cache[key]
def put(self, key: IntentCacheKey, value: IntentCacheValue) -> None:
"""Put a value in the cache, evicting the least recently used item if necessary."""
if key in self.cache:
# Update value and mark as recently used
self.cache.move_to_end(key)
elif len(self.cache) >= self.capacity:
# Evict the oldest item
self.cache.popitem(last=False)
self.cache[key] = value
def clear(self) -> None:
"""Clear the cache."""
self.cache.clear()
async def async_setup_default_agent(
hass: HomeAssistant,
entity_component: EntityComponent[ConversationEntity],
config_intents: dict[str, Any],
) -> None:
"""Set up entity registry listener for the default agent."""
agent = DefaultAgent(hass, config_intents)
await entity_component.async_add_entities([agent])
await get_agent_manager(hass).async_setup_default_agent(agent)
@callback
def async_entity_state_listener(event: Event[EventStateChangedData]) -> None:
"""Set expose flag on new entities."""
async_should_expose(hass, DOMAIN, event.data["entity_id"])
@callback
def async_hass_started(hass: HomeAssistant) -> None:
"""Set expose flag on all entities."""
for state in hass.states.async_all():
async_should_expose(hass, DOMAIN, state.entity_id)
async_track_state_added_domain(hass, MATCH_ALL, async_entity_state_listener)
ha_start.async_at_started(hass, async_hass_started)
class DefaultAgent(ConversationEntity):
"""Default agent for conversation agent."""
_attr_name = "Home Assistant"
_attr_supported_features = ConversationEntityFeature.CONTROL
def __init__(self, hass: HomeAssistant, config_intents: dict[str, Any]) -> None:
"""Initialize the default agent."""
self.hass = hass
self._lang_intents: dict[str, LanguageIntents | object] = {}
self._load_intents_lock = asyncio.Lock()
# intent -> [sentences]
self._config_intents: dict[str, Any] = config_intents
# Sentences that will trigger a callback (skipping intent recognition)
self._triggers_details: list[TriggerDetails] = []
self._trigger_intents: Intents | None = None
# Slot lists for entities, areas, etc.
self._slot_lists: dict[str, SlotList] | None = None
self._unsub_clear_slot_list: list[Callable[[], None]] | None = None
# Used to filter slot lists before intent matching
self._exposed_names_trie: Trie | None = None
self._unexposed_names_trie: Trie | None = None
# LRU cache to avoid unnecessary intent matching
self._intent_cache = IntentCache(capacity=128)
# Shared configuration for fuzzy matching
self.fuzzy_matching = True
self._fuzzy_config: FuzzyConfig | None = None
@property
def supported_languages(self) -> list[str]:
"""Return a list of supported languages."""
return get_languages()
@callback
def _filter_entity_registry_changes(
self, event_data: er.EventEntityRegistryUpdatedData
) -> bool:
"""Filter entity registry changed events."""
return event_data["action"] == "update" and any(
field in event_data["changes"] for field in _ENTITY_REGISTRY_UPDATE_FIELDS
)
@callback
def _filter_state_changes(self, event_data: EventStateChangedData) -> bool:
"""Filter state changed events."""
return not event_data["old_state"] or not event_data["new_state"]
@callback
def _listen_clear_slot_list(self) -> None:
"""Listen for changes that can invalidate slot list."""
assert self._unsub_clear_slot_list is None
self._unsub_clear_slot_list = [
self.hass.bus.async_listen(
ar.EVENT_AREA_REGISTRY_UPDATED,
self._async_clear_slot_list,
),
self.hass.bus.async_listen(
fr.EVENT_FLOOR_REGISTRY_UPDATED,
self._async_clear_slot_list,
),
self.hass.bus.async_listen(
er.EVENT_ENTITY_REGISTRY_UPDATED,
self._async_clear_slot_list,
event_filter=self._filter_entity_registry_changes,
),
self.hass.bus.async_listen(
EVENT_STATE_CHANGED,
self._async_clear_slot_list,
event_filter=self._filter_state_changes,
),
async_listen_entity_updates(self.hass, DOMAIN, self._async_clear_slot_list),
]
async def async_recognize_intent(
self, user_input: ConversationInput, strict_intents_only: bool = False
) -> RecognizeResult | None:
"""Recognize intent from user input."""
language = user_input.language or self.hass.config.language
lang_intents = await self.async_get_or_load_intents(language)
if lang_intents is None:
# No intents loaded
_LOGGER.warning("No intents were loaded for language: %s", language)
return None
slot_lists = await self._make_slot_lists()
intent_context = self._make_intent_context(user_input)
if self._exposed_names_trie is not None:
# Filter by input string
text = remove_punctuation(user_input.text).strip().lower()
slot_lists["name"] = TextSlotList(
name="name",
values=[result[2] for result in self._exposed_names_trie.find(text)],
)
start = time.monotonic()
result = await self.hass.async_add_executor_job(
self._recognize,
user_input,
lang_intents,
slot_lists,
intent_context,
language,
strict_intents_only,
)
_LOGGER.debug(
"Recognize done in %.2f seconds",
time.monotonic() - start,
)
return result
async def async_debug_recognize(
self, user_input: ConversationInput
) -> dict[str, Any] | None:
"""Debug recognize from user input."""
result_dict: dict[str, Any] | None = None
if trigger_result := await self.async_recognize_sentence_trigger(user_input):
result_dict = {
# Matched a user-defined sentence trigger.
# We can't provide the response here without executing the
# trigger.
"match": True,
"source": "trigger",
"sentence_template": trigger_result.sentence_template or "",
}
elif intent_result := await self.async_recognize_intent(user_input):
successful_match = not intent_result.unmatched_entities
result_dict = {
# Name of the matching intent (or the closest)
"intent": {
"name": intent_result.intent.name,
},
# Slot values that would be received by the intent
"slots": { # direct access to values
entity_key: entity.text or entity.value
for entity_key, entity in intent_result.entities.items()
},
# Extra slot details, such as the originally matched text
"details": {
entity_key: {
"name": entity.name,
"value": entity.value,
"text": entity.text,
}
for entity_key, entity in intent_result.entities.items()
},
# Entities/areas/etc. that would be targeted
"targets": {},
# True if match was successful
"match": successful_match,
# Text of the sentence template that matched (or was closest)
"sentence_template": "",
# When match is incomplete, this will contain the best slot guesses
"unmatched_slots": _get_unmatched_slots(intent_result),
# True if match was not exact
"fuzzy_match": False,
}
if successful_match:
result_dict["targets"] = {
state.entity_id: {"matched": is_matched}
for state, is_matched in _get_debug_targets(
self.hass, intent_result
)
}
if intent_result.intent_sentence is not None:
result_dict["sentence_template"] = intent_result.intent_sentence.text
if intent_result.intent_metadata:
# Inspect metadata to determine if this matched a custom sentence
if intent_result.intent_metadata.get(METADATA_CUSTOM_SENTENCE):
result_dict["source"] = "custom"
result_dict["file"] = intent_result.intent_metadata.get(
METADATA_CUSTOM_FILE
)
else:
result_dict["source"] = "builtin"
result_dict["fuzzy_match"] = intent_result.intent_metadata.get(
METADATA_FUZZY_MATCH, False
)
return result_dict
async def _async_handle_message(
self,
user_input: ConversationInput,
chat_log: ChatLog,
) -> ConversationResult:
"""Handle a message."""
response: intent.IntentResponse | None = None
# Check if a trigger matched
if trigger_result := await self.async_recognize_sentence_trigger(user_input):
# Process callbacks and get response
response_text = await self._handle_trigger_result(
trigger_result, user_input
)
# Convert to conversation result
response = intent.IntentResponse(
language=user_input.language or self.hass.config.language
)
response.async_set_speech(response_text)
if response is None:
# Match intents
intent_result = await self.async_recognize_intent(user_input)
response = await self._async_process_intent_result(
intent_result, user_input
)
speech: str = response.speech.get("plain", {}).get("speech", "")
chat_log.async_add_assistant_content_without_tools(
AssistantContent(
agent_id=user_input.agent_id,
content=speech,
)
)
return ConversationResult(
response=response, conversation_id=chat_log.conversation_id
)
async def _async_process_intent_result(
self,
result: RecognizeResult | None,
user_input: ConversationInput,
) -> intent.IntentResponse:
"""Process user input with intents."""
language = user_input.language or self.hass.config.language
# Intent match or failure
lang_intents = await self.async_get_or_load_intents(language)
if result is None:
# Intent was not recognized
_LOGGER.debug("No intent was matched for '%s'", user_input.text)
return _make_error_result(
language,
intent.IntentResponseErrorCode.NO_INTENT_MATCH,
self._get_error_text(ErrorKey.NO_INTENT, lang_intents),
)
if result.unmatched_entities:
# Intent was recognized, but not entity/area names, etc.
_LOGGER.debug(
"Recognized intent '%s' for template '%s' but had unmatched: %s",
result.intent.name,
(
result.intent_sentence.text
if result.intent_sentence is not None
else ""
),
result.unmatched_entities_list,
)
error_response_type, error_response_args = _get_unmatched_response(result)
return _make_error_result(
language,
intent.IntentResponseErrorCode.NO_VALID_TARGETS,
self._get_error_text(
error_response_type, lang_intents, **error_response_args
),
)
# Will never happen because result will be None when no intents are
# loaded in async_recognize.
assert lang_intents is not None
# Slot values to pass to the intent
slots: dict[str, Any] = {
entity.name: {
"value": entity.value,
"text": entity.text or entity.value,
}
for entity in result.entities_list
}
satellite_id = user_input.satellite_id
device_id = user_input.device_id
satellite_area, device_id = self._get_satellite_area_and_device(
satellite_id, device_id
)
if satellite_area is not None:
slots["preferred_area_id"] = {"value": satellite_area.id}
async_conversation_trace_append(
ConversationTraceEventType.TOOL_CALL,
{
"intent_name": result.intent.name,
"slots": {
entity.name: entity.value or entity.text
for entity in result.entities_list
},
},
)
try:
intent_response = await intent.async_handle(
self.hass,
DOMAIN,
result.intent.name,
slots,
user_input.text,
user_input.context,
language,
assistant=DOMAIN,
device_id=device_id,
satellite_id=satellite_id,
conversation_agent_id=user_input.agent_id,
)
except intent.MatchFailedError as match_error:
# Intent was valid, but no entities matched the constraints.
error_response_type, error_response_args = _get_match_error_response(
self.hass, match_error
)
return _make_error_result(
language,
intent.IntentResponseErrorCode.NO_VALID_TARGETS,
self._get_error_text(
error_response_type, lang_intents, **error_response_args
),
)
except intent.IntentHandleError as err:
# Intent was valid and entities matched constraints, but an error
# occurred during handling.
_LOGGER.exception("Intent handling error")
return _make_error_result(
language,
intent.IntentResponseErrorCode.FAILED_TO_HANDLE,
self._get_error_text(
err.response_key or ErrorKey.HANDLE_ERROR, lang_intents
),
)
except intent.IntentUnexpectedError:
_LOGGER.exception("Unexpected intent error")
return _make_error_result(
language,
intent.IntentResponseErrorCode.UNKNOWN,
self._get_error_text(ErrorKey.HANDLE_ERROR, lang_intents),
)
if (
(not intent_response.speech)
and (intent_response.intent is not None)
and (response_key := result.response)
):
# Use response template, if available
response_template_str = lang_intents.intent_responses.get(
result.intent.name, {}
).get(response_key)
if response_template_str:
response_template = template.Template(response_template_str, self.hass)
speech = await self._build_speech(
language, response_template, intent_response, result
)
intent_response.async_set_speech(speech)
return intent_response
def _recognize(
self,
user_input: ConversationInput,
lang_intents: LanguageIntents,
slot_lists: dict[str, SlotList],
intent_context: dict[str, Any] | None,
language: str,
strict_intents_only: bool,
) -> RecognizeResult | None:
"""Search intents for a match to user input."""
skip_exposed_match = False
# Try cache first
cache_key = IntentCacheKey(
text=user_input.text,
language=language,
satellite_id=user_input.satellite_id,
)
cache_value = self._intent_cache.get(cache_key)
if cache_value is not None:
if (cache_value.result is not None) and (
cache_value.stage == IntentMatchingStage.EXPOSED_ENTITIES_ONLY
):
_LOGGER.debug("Got cached result for exposed entities")
return cache_value.result
# Continue with matching, but we know we won't succeed for exposed
# entities only.
skip_exposed_match = True
if not skip_exposed_match:
start_time = time.monotonic()
strict_result = self._recognize_strict(
user_input, lang_intents, slot_lists, intent_context, language
)
_LOGGER.debug(
"Checked exposed entities in %s second(s)",
time.monotonic() - start_time,
)
# Update cache
self._intent_cache.put(
cache_key,
IntentCacheValue(
result=strict_result,
stage=IntentMatchingStage.EXPOSED_ENTITIES_ONLY,
),
)
if strict_result is not None:
# Successful strict match with exposed entities
return strict_result
if strict_intents_only:
# Don't try matching against all entities or doing a fuzzy match
return None
# Use fuzzy matching
skip_fuzzy_match = False
if cache_value is not None:
if (cache_value.result is not None) and (
cache_value.stage == IntentMatchingStage.FUZZY
):
_LOGGER.debug("Got cached result for fuzzy match")
return cache_value.result
# Continue with matching, but we know we won't succeed for fuzzy
# match.
skip_fuzzy_match = True
if (not skip_fuzzy_match) and self.fuzzy_matching:
start_time = time.monotonic()
fuzzy_result = self._recognize_fuzzy(lang_intents, user_input)
# Update cache
self._intent_cache.put(
cache_key,
IntentCacheValue(result=fuzzy_result, stage=IntentMatchingStage.FUZZY),
)
_LOGGER.debug(
"Did fuzzy match in %s second(s)", time.monotonic() - start_time
)
if fuzzy_result is not None:
return fuzzy_result
# Try again with all entities (including unexposed)
skip_unexposed_entities_match = False
if cache_value is not None:
if (cache_value.result is not None) and (
cache_value.stage == IntentMatchingStage.UNEXPOSED_ENTITIES
):
_LOGGER.debug("Got cached result for all entities")
return cache_value.result
# Continue with matching, but we know we won't succeed for all
# entities.
skip_unexposed_entities_match = True
if not skip_unexposed_entities_match:
unexposed_entities_slot_lists = {
**slot_lists,
"name": self._get_unexposed_entity_names(user_input.text),
}
start_time = time.monotonic()
strict_result = self._recognize_strict(
user_input,
lang_intents,
unexposed_entities_slot_lists,
intent_context,
language,
)
_LOGGER.debug(
"Checked all entities in %s second(s)", time.monotonic() - start_time
)
# Update cache
self._intent_cache.put(
cache_key,
IntentCacheValue(
result=strict_result, stage=IntentMatchingStage.UNEXPOSED_ENTITIES
),
)
if strict_result is not None:
# Not a successful match, but useful for an error message.
# This should fail the intent handling phase (async_match_targets).
return strict_result
# Check unknown names
skip_unknown_names = False
if cache_value is not None:
if (cache_value.result is not None) and (
cache_value.stage == IntentMatchingStage.UNKNOWN_NAMES
):
_LOGGER.debug("Got cached result for unknown names")
return cache_value.result
skip_unknown_names = True
maybe_result: RecognizeResult | None = None
if not skip_unknown_names:
start_time = time.monotonic()
maybe_result = self._recognize_unknown_names(
lang_intents, user_input, slot_lists, intent_context
)
# Update cache
self._intent_cache.put(
cache_key,
IntentCacheValue(
result=maybe_result, stage=IntentMatchingStage.UNKNOWN_NAMES
),
)
_LOGGER.debug(
"Did unknown names match in %s second(s)", time.monotonic() - start_time
)
return maybe_result
def _recognize_fuzzy(
self, lang_intents: LanguageIntents, user_input: ConversationInput
) -> RecognizeResult | None:
"""Return fuzzy recognition from hassil."""
if lang_intents.fuzzy_matcher is None:
return None
fuzzy_result = lang_intents.fuzzy_matcher.match(user_input.text)
if fuzzy_result is None:
return None
response = "default"
if lang_intents.fuzzy_responses:
domain = "" # no domain
if "name" in fuzzy_result.slots:
domain = fuzzy_result.name_domain
elif "domain" in fuzzy_result.slots:
domain = fuzzy_result.slots["domain"].value
slot_combo = tuple(sorted(fuzzy_result.slots))
if (
intent_responses := lang_intents.fuzzy_responses.get(
fuzzy_result.intent_name
)
) and (combo_responses := intent_responses.get(slot_combo)):
response = combo_responses.get(domain, response)
entities = [
MatchEntity(name=slot_name, value=slot_value.value, text=slot_value.text)
for slot_name, slot_value in fuzzy_result.slots.items()
]
return RecognizeResult(
intent=Intent(name=fuzzy_result.intent_name),
intent_data=IntentData(sentence_texts=[]),
intent_metadata={METADATA_FUZZY_MATCH: True},
entities={entity.name: entity for entity in entities},
entities_list=entities,
response=response,
)
def _recognize_unknown_names(
self,
lang_intents: LanguageIntents,
user_input: ConversationInput,
slot_lists: dict[str, SlotList],
intent_context: dict[str, Any] | None,
) -> RecognizeResult | None:
"""Return result with unknown names for an error message."""
maybe_result: RecognizeResult | None = None
best_num_matched_entities = 0
best_num_unmatched_entities = 0
best_num_unmatched_ranges = 0
for result in recognize_all(
user_input.text,
lang_intents.intents,
slot_lists=slot_lists,
intent_context=intent_context,
allow_unmatched_entities=True,
):
if result.text_chunks_matched < 1:
# Skip results that don't match any literal text
continue
# Don't count missing entities that couldn't be filled from context
num_matched_entities = 0
for matched_entity in result.entities_list:
if matched_entity.name not in result.unmatched_entities:
num_matched_entities += 1
num_unmatched_entities = 0
num_unmatched_ranges = 0
for unmatched_entity in result.unmatched_entities_list:
if isinstance(unmatched_entity, UnmatchedTextEntity):
if unmatched_entity.text != MISSING_ENTITY:
num_unmatched_entities += 1
elif isinstance(unmatched_entity, UnmatchedRangeEntity):
num_unmatched_ranges += 1
num_unmatched_entities += 1
else:
num_unmatched_entities += 1
if (
(maybe_result is None) # first result
or (
# More literal text matched
result.text_chunks_matched > maybe_result.text_chunks_matched
)
or (
# More entities matched
num_matched_entities > best_num_matched_entities
)
or (
# Fewer unmatched entities
(num_matched_entities == best_num_matched_entities)
and (num_unmatched_entities < best_num_unmatched_entities)
)
or (
# Prefer unmatched ranges
(num_matched_entities == best_num_matched_entities)
and (num_unmatched_entities == best_num_unmatched_entities)
and (num_unmatched_ranges > best_num_unmatched_ranges)
)
or (
# Prefer match failures with entities
(result.text_chunks_matched == maybe_result.text_chunks_matched)
and (num_unmatched_entities == best_num_unmatched_entities)
and (num_unmatched_ranges == best_num_unmatched_ranges)
and (
("name" in result.entities)
or ("name" in result.unmatched_entities)
)
)
):
maybe_result = result
best_num_matched_entities = num_matched_entities
best_num_unmatched_entities = num_unmatched_entities
best_num_unmatched_ranges = num_unmatched_ranges
return maybe_result
def _get_unexposed_entity_names(self, text: str) -> TextSlotList:
"""Get filtered slot list with unexposed entity names in Home Assistant."""
if self._unexposed_names_trie is None:
# Build trie
self._unexposed_names_trie = Trie()
for name_tuple in self._get_entity_name_tuples(exposed=False):
self._unexposed_names_trie.insert(
name_tuple[0].lower(),
TextSlotValue.from_tuple(name_tuple, allow_template=False),
)
# Build filtered slot list
text_lower = text.strip().lower()
return TextSlotList(
name="name",
values=[
result[2] for result in self._unexposed_names_trie.find(text_lower)
],
)
def _get_entity_name_tuples(
self, exposed: bool
) -> Iterable[tuple[str, str, dict[str, Any]]]:
"""Yield (input name, output name, context) tuples for entities."""
entity_registry = er.async_get(self.hass)
for state in self.hass.states.async_all():
entity_exposed = async_should_expose(self.hass, DOMAIN, state.entity_id)
if exposed and (not entity_exposed):
# Required exposed, entity is not
continue
if (not exposed) and entity_exposed:
# Required not exposed, entity is
continue
# Checked against "requires_context" and "excludes_context" in hassil
context = {"domain": state.domain}
if state.attributes:
# Include some attributes
for attr in _DEFAULT_EXPOSED_ATTRIBUTES:
if attr not in state.attributes:
continue
context[attr] = state.attributes[attr]
if (
entity := entity_registry.async_get(state.entity_id)
) and entity.aliases:
for alias in entity.aliases:
alias = alias.strip()
if not alias:
continue
yield (alias, alias, context)
# Default name
yield (state.name, state.name, context)
def _recognize_strict(
self,
user_input: ConversationInput,
lang_intents: LanguageIntents,
slot_lists: dict[str, SlotList],
intent_context: dict[str, Any] | None,
language: str,
) -> RecognizeResult | None:
"""Search intents for a strict match to user input."""
return recognize_best(
user_input.text,
lang_intents.intents,
slot_lists=slot_lists,
intent_context=intent_context,
language=language,
best_metadata_key=METADATA_CUSTOM_SENTENCE,
best_slot_name="name",
)
async def _build_speech(
self,
language: str,
response_template: template.Template,
intent_response: intent.IntentResponse,
recognize_result: RecognizeResult,
) -> str:
# Get first matched or unmatched state.
# This is available in the response template as "state".
state1: State | None = None
if intent_response.matched_states:
state1 = intent_response.matched_states[0]
elif intent_response.unmatched_states:
state1 = intent_response.unmatched_states[0]
# Render response template
speech_slots = {
entity_name: entity_value.text or entity_value.value
for entity_name, entity_value in recognize_result.entities.items()
}
speech_slots.update(intent_response.speech_slots)
speech = response_template.async_render(
{
# Slots from intent recognizer and response
"slots": speech_slots,
# First matched or unmatched state
"state": (
template.TemplateState(self.hass, state1)
if state1 is not None
else None
),
"query": {
# Entity states that matched the query (e.g, "on")
"matched": [
template.TemplateState(self.hass, state)
for state in intent_response.matched_states
],
# Entity states that did not match the query
"unmatched": [
template.TemplateState(self.hass, state)
for state in intent_response.unmatched_states
],
},
}
)
# Normalize whitespace
if speech is not None:
speech = str(speech)
speech = " ".join(speech.strip().split())
return speech
async def async_reload(self, language: str | None = None) -> None:
"""Clear cached intents for a language."""
if language is None:
self._lang_intents.clear()
_LOGGER.debug("Cleared intents for all languages")
else:
self._lang_intents.pop(language, None)
_LOGGER.debug("Cleared intents for language: %s", language)
# Intents have changed, so we must clear the cache
self._intent_cache.clear()
async def async_prepare(self, language: str | None = None) -> None:
"""Load intents for a language."""
if language is None:
language = self.hass.config.language
lang_intents = await self.async_get_or_load_intents(language)
# No intents loaded
if lang_intents is None:
return
await self._make_slot_lists()
async def async_get_or_load_intents(self, language: str) -> LanguageIntents | None:
"""Load all intents of a language with lock."""
if lang_intents := self._lang_intents.get(language):
return (
None
if lang_intents is ERROR_SENTINEL
else cast(LanguageIntents, lang_intents)
)
async with self._load_intents_lock:
# In case it was loaded now
if lang_intents := self._lang_intents.get(language):
return (
None
if lang_intents is ERROR_SENTINEL
else cast(LanguageIntents, lang_intents)
)
start = time.monotonic()
result = await self.hass.async_add_executor_job(
self._load_intents, language
)
if result is None:
self._lang_intents[language] = ERROR_SENTINEL
else:
self._lang_intents[language] = result
_LOGGER.debug(
"Full intents load completed for language=%s in %.2f seconds",
language,
time.monotonic() - start,
)
return result
def _load_intents(self, language: str) -> LanguageIntents | None:
"""Load all intents for language (run inside executor)."""
intents_dict: dict[str, Any] = {}
supported_langs = set(get_languages())
# Choose a language variant upfront and commit to it for custom
# sentences, etc.
lang_matches = language_util.matches(language, supported_langs)
if not lang_matches:
_LOGGER.warning(
"Unable to find supported language variant for %s", language
)
return None
language_variant = lang_matches[0]
# Load intents for this language variant
lang_variant_intents = get_intents(language_variant, json_load=json_load)
if lang_variant_intents:
# Merge sentences into existing dictionary
# Overriding because source dict is empty
intents_dict = lang_variant_intents
_LOGGER.debug(
"Loaded built-in intents for language=%s (%s)",
language,
language_variant,
)
# Check for custom sentences in <config>/custom_sentences/<language>/
custom_sentences_dir = Path(
self.hass.config.path("custom_sentences", language_variant)
)
if custom_sentences_dir.is_dir():
for custom_sentences_path in custom_sentences_dir.rglob("*.yaml"):
with custom_sentences_path.open(
encoding="utf-8"
) as custom_sentences_file:
# Merge custom sentences
if not isinstance(
custom_sentences_yaml := yaml.safe_load(custom_sentences_file),
dict,
):
_LOGGER.warning(
"Custom sentences file does not match expected format path=%s",
custom_sentences_file.name,
)
continue
# Add metadata so we can identify custom sentences in the debugger
custom_intents_dict = custom_sentences_yaml.get("intents", {})
for intent_dict in custom_intents_dict.values():
intent_data_list = intent_dict.get("data", [])
for intent_data in intent_data_list:
sentence_metadata = intent_data.get("metadata", {})
sentence_metadata[METADATA_CUSTOM_SENTENCE] = True
sentence_metadata[METADATA_CUSTOM_FILE] = str(
custom_sentences_path.relative_to(
custom_sentences_dir.parent
)
)
intent_data["metadata"] = sentence_metadata
merge_dict(intents_dict, custom_sentences_yaml)
_LOGGER.debug(
"Loaded custom sentences language=%s (%s), path=%s",
language,
language_variant,
custom_sentences_path,
)
# Load sentences from HA config for default language only
if self._config_intents and (
self.hass.config.language in (language, language_variant)
):
hass_config_path = self.hass.config.path()
merge_dict(
intents_dict,
{
"intents": {
intent_name: {
"data": [
{
"sentences": sentences,
"metadata": {
METADATA_CUSTOM_SENTENCE: True,
METADATA_CUSTOM_FILE: hass_config_path,
},
}
]
}
for intent_name, sentences in self._config_intents.items()
}
},
)
_LOGGER.debug(
"Loaded intents from configuration.yaml",
)
if not intents_dict:
return None
intents = Intents.from_dict(intents_dict)
# Load responses
responses_dict = intents_dict.get("responses", {})
intent_responses = responses_dict.get("intents", {})
error_responses = responses_dict.get("errors", {})
if not self.fuzzy_matching:
_LOGGER.debug("Fuzzy matching is disabled")
return LanguageIntents(
intents,
intents_dict,
intent_responses,
error_responses,
language_variant,
)
# Load fuzzy
fuzzy_info = get_fuzzy_language(language_variant, json_load=json_load)
if fuzzy_info is None:
_LOGGER.debug(
"Fuzzy matching not available for language: %s", language_variant
)
return LanguageIntents(
intents,
intents_dict,
intent_responses,
error_responses,
language_variant,
)
if self._fuzzy_config is None:
# Load shared config
self._fuzzy_config = get_fuzzy_config(json_load=json_load)
_LOGGER.debug("Loaded shared fuzzy matching config")
assert self._fuzzy_config is not None
fuzzy_matcher: FuzzyNgramMatcher | None = None
fuzzy_responses: FuzzyLanguageResponses | None = None
start_time = time.monotonic()
fuzzy_responses = fuzzy_info.responses
fuzzy_matcher = FuzzyNgramMatcher(
intents=intents,
intent_models={
intent_name: Sqlite3NgramModel(
order=fuzzy_model.order,
words={
word: str(word_id)
for word, word_id in fuzzy_model.words.items()
},
database_path=fuzzy_model.database_path,
)
for intent_name, fuzzy_model in fuzzy_info.ngram_models.items()
},
intent_slot_list_names=self._fuzzy_config.slot_list_names,
slot_combinations={
intent_name: {
combo_key: [
SlotCombinationInfo(
name_domains=(set(name_domains) if name_domains else None)
)
]
for combo_key, name_domains in intent_combos.items()
}
for intent_name, intent_combos in self._fuzzy_config.slot_combinations.items()
},
domain_keywords=fuzzy_info.domain_keywords,
stop_words=fuzzy_info.stop_words,
)
_LOGGER.debug(
"Loaded fuzzy matcher in %s second(s): language=%s, intents=%s",
time.monotonic() - start_time,
language_variant,
sorted(fuzzy_matcher.intent_models.keys()),
)
return LanguageIntents(
intents,
intents_dict,
intent_responses,
error_responses,
language_variant,
fuzzy_matcher=fuzzy_matcher,
fuzzy_responses=fuzzy_responses,
)
@callback
def _async_clear_slot_list(self, event: Event[Any] | None = None) -> None:
"""Clear slot lists when a registry has changed."""
# Two subscribers can be scheduled at same time
_LOGGER.debug("Clearing slot lists")
if self._unsub_clear_slot_list is None:
return
self._slot_lists = None
self._exposed_names_trie = None
self._unexposed_names_trie = None
for unsub in self._unsub_clear_slot_list:
unsub()
self._unsub_clear_slot_list = None
# Slot lists have changed, so we must clear the cache
self._intent_cache.clear()
async def _make_slot_lists(self) -> dict[str, SlotList]:
"""Create slot lists with areas and entity names/aliases."""
if self._slot_lists is not None:
return self._slot_lists
start = time.monotonic()
# Gather entity names, keeping track of exposed names.
# We try intent recognition with only exposed names first, then all names.
#
# NOTE: We do not pass entity ids in here because multiple entities may
# have the same name. The intent matcher doesn't gather all matching
# values for a list, just the first. So we will need to match by name no
# matter what.
exposed_entity_names = list(self._get_entity_name_tuples(exposed=True))
_LOGGER.debug("Exposed entities: %s", exposed_entity_names)
# Expose all areas.
areas = ar.async_get(self.hass)
area_names = []
for area in areas.async_list_areas():
area_names.append((area.name, area.name))
if not area.aliases:
continue
for alias in area.aliases:
alias = alias.strip()
if not alias:
continue
area_names.append((alias, alias))
# Expose all floors.
floors = fr.async_get(self.hass)
floor_names = []
for floor in floors.async_list_floors():
floor_names.append((floor.name, floor.name))
if not floor.aliases:
continue
for alias in floor.aliases:
alias = alias.strip()
if not alias:
continue
floor_names.append((alias, floor.name))
# Build trie
self._exposed_names_trie = Trie()
name_list = TextSlotList.from_tuples(exposed_entity_names, allow_template=False)
for name_value in name_list.values:
assert isinstance(name_value.text_in, TextChunk)
name_text = remove_punctuation(name_value.text_in.text).strip().lower()
self._exposed_names_trie.insert(name_text, name_value)
self._slot_lists = {
"area": TextSlotList.from_tuples(area_names, allow_template=False),
"name": name_list,
"floor": TextSlotList.from_tuples(floor_names, allow_template=False),
}
# Reload fuzzy matchers with new slot lists
if self.fuzzy_matching:
await self.hass.async_add_executor_job(self._load_fuzzy_matchers)
self._listen_clear_slot_list()
_LOGGER.debug(
"Created slot lists in %.2f seconds",
time.monotonic() - start,
)
return self._slot_lists
def _load_fuzzy_matchers(self) -> None:
"""Reload fuzzy matchers for all loaded languages."""
for lang_intents in self._lang_intents.values():
if (not isinstance(lang_intents, LanguageIntents)) or (
lang_intents.fuzzy_matcher is None
):
continue
lang_matcher = lang_intents.fuzzy_matcher
lang_intents.fuzzy_matcher = FuzzyNgramMatcher(
intents=lang_matcher.intents,
intent_models=lang_matcher.intent_models,
intent_slot_list_names=lang_matcher.intent_slot_list_names,
slot_combinations=lang_matcher.slot_combinations,
domain_keywords=lang_matcher.domain_keywords,
stop_words=lang_matcher.stop_words,
slot_lists=self._slot_lists,
)
def _make_intent_context(
self, user_input: ConversationInput
) -> dict[str, Any] | None:
"""Return intent recognition context for user input."""
satellite_area, _ = self._get_satellite_area_and_device(
user_input.satellite_id, user_input.device_id
)
if satellite_area is None:
return None
return {"area": {"value": satellite_area.name, "text": satellite_area.name}}
def _get_satellite_area_and_device(
self, satellite_id: str | None, device_id: str | None = None
) -> tuple[ar.AreaEntry | None, str | None]:
"""Return area entry and device id."""
hass = self.hass
area_id: str | None = None
if (
satellite_id is not None
and (entity_entry := er.async_get(hass).async_get(satellite_id)) is not None
):
area_id = entity_entry.area_id
device_id = entity_entry.device_id
if (
area_id is None
and device_id is not None
and (device_entry := dr.async_get(hass).async_get(device_id)) is not None
):
area_id = device_entry.area_id
if area_id is None:
return None, device_id
return ar.async_get(hass).async_get_area(area_id), device_id
def _get_error_text(
self,
error_key: ErrorKey | str,
lang_intents: LanguageIntents | None,
**response_args,
) -> str:
"""Get response error text by type."""
if lang_intents is None:
return _DEFAULT_ERROR_TEXT
if isinstance(error_key, ErrorKey):
response_key = error_key.value
else:
response_key = error_key
response_str = (
lang_intents.error_responses.get(response_key) or _DEFAULT_ERROR_TEXT
)
response_template = template.Template(response_str, self.hass)
return response_template.async_render(response_args)
@callback
def update_triggers(self, triggers_details: list[TriggerDetails]) -> None:
"""Update triggers."""
self._triggers_details = triggers_details
# Force rebuild on next use
self._trigger_intents = None
def _rebuild_trigger_intents(self) -> None:
"""Rebuild the HassIL intents object from the current trigger sentences."""
intents_dict = {
"language": self.hass.config.language,
"intents": {
# Use trigger data index as a virtual intent name for HassIL.
# This works because the intents are rebuilt on every
# register/unregister.
str(trigger_id): {"data": [{"sentences": trigger_details.sentences}]}
for trigger_id, trigger_details in enumerate(self._triggers_details)
},
}
trigger_intents = Intents.from_dict(intents_dict)
# Assume slot list references are wildcards
wildcard_names: set[str] = set()
for trigger_intent in trigger_intents.intents.values():
for intent_data in trigger_intent.data:
for sentence in intent_data.sentences:
_collect_list_references(sentence.expression, wildcard_names)
for wildcard_name in wildcard_names:
trigger_intents.slot_lists[wildcard_name] = WildcardSlotList(wildcard_name)
self._trigger_intents = trigger_intents
_LOGGER.debug("Rebuilt trigger intents: %s", intents_dict)
async def async_recognize_sentence_trigger(
self, user_input: ConversationInput
) -> SentenceTriggerResult | None:
"""Try to match sentence against registered trigger sentences.
Calls the registered callbacks if there's a match and returns a sentence
trigger result.
"""
if not self._triggers_details:
# No triggers registered
return None
if self._trigger_intents is None:
# Need to rebuild intents before matching
self._rebuild_trigger_intents()
assert self._trigger_intents is not None
matched_triggers: dict[int, RecognizeResult] = {}
matched_template: str | None = None
for result in recognize_all(user_input.text, self._trigger_intents):
if result.intent_sentence is not None:
matched_template = result.intent_sentence.text
trigger_id = int(result.intent.name)
if trigger_id in matched_triggers:
# Already matched a sentence from this trigger
break
matched_triggers[trigger_id] = result
if not matched_triggers:
# Sentence did not match any trigger sentences
return None
_LOGGER.debug(
"'%s' matched %s trigger(s): %s",
user_input.text,
len(matched_triggers),
list(matched_triggers),
)
return SentenceTriggerResult(
user_input.text, matched_template, matched_triggers
)
async def _handle_trigger_result(
self, result: SentenceTriggerResult, user_input: ConversationInput
) -> str:
"""Run sentence trigger callbacks and return response text."""
# Gather callback responses in parallel
trigger_callbacks = [
self._triggers_details[trigger_id].callback(user_input, trigger_result)
for trigger_id, trigger_result in result.matched_triggers.items()
]
# Use first non-empty result as response.
#
# There may be multiple copies of a trigger running when editing in
# the UI, so it's critical that we filter out empty responses here.
response_text = ""
response_set_by_trigger = False
for trigger_future in asyncio.as_completed(trigger_callbacks):
trigger_response = await trigger_future
if trigger_response is None:
continue
response_text = trigger_response
response_set_by_trigger = True
break
if response_set_by_trigger:
# Response was explicitly set to empty
response_text = response_text or ""
elif not response_text:
# Use translated acknowledgment for pipeline language
language = user_input.language or self.hass.config.language
translations = await translation.async_get_translations(
self.hass, language, DOMAIN, [DOMAIN]
)
response_text = translations.get(
f"component.{DOMAIN}.conversation.agent.done", "Done"
)
return response_text
async def async_handle_sentence_triggers(
self, user_input: ConversationInput
) -> str | None:
"""Try to input sentence against sentence triggers and return response text.
Returns None if no match occurred.
"""
if trigger_result := await self.async_recognize_sentence_trigger(user_input):
return await self._handle_trigger_result(trigger_result, user_input)
return None
async def async_handle_intents(
self,
user_input: ConversationInput,
*,
intent_filter: Callable[[RecognizeResult], bool] | None = None,
) -> intent.IntentResponse | None:
"""Try to match sentence against registered intents and return response.
Only performs strict matching with exposed entities and exact wording.
Returns None if no match or a matching error occurred.
"""
result = await self.async_recognize_intent(user_input, strict_intents_only=True)
if not isinstance(result, RecognizeResult) or (
intent_filter is not None and intent_filter(result)
):
# No error message on failed match
return None
response = await self._async_process_intent_result(result, user_input)
if (
response.response_type == intent.IntentResponseType.ERROR
and response.error_code
not in (
intent.IntentResponseErrorCode.FAILED_TO_HANDLE,
intent.IntentResponseErrorCode.UNKNOWN,
)
):
# We ignore no matching errors
return None
return response
async def async_get_language_scores(self) -> dict[str, LanguageScores]:
"""Get support scores per language."""
return await self.hass.async_add_executor_job(get_language_scores)
def _make_error_result(
language: str,
error_code: intent.IntentResponseErrorCode,
response_text: str,
) -> intent.IntentResponse:
"""Create conversation result with error code and text."""
response = intent.IntentResponse(language=language)
response.async_set_error(error_code, response_text)
return response
def _get_unmatched_response(result: RecognizeResult) -> tuple[ErrorKey, dict[str, Any]]:
"""Get key and template arguments for error when there are unmatched intent entities/slots."""
# Filter out non-text and missing context entities
unmatched_text: dict[str, str] = {
key: entity.text.strip()
for key, entity in result.unmatched_entities.items()
if isinstance(entity, UnmatchedTextEntity) and entity.text != MISSING_ENTITY
}
if unmatched_area := unmatched_text.get("area"):
# area only
return ErrorKey.NO_AREA, {"area": unmatched_area}
if unmatched_floor := unmatched_text.get("floor"):
# floor only
return ErrorKey.NO_FLOOR, {"floor": unmatched_floor}
# Area may still have matched
matched_area: str | None = None
if matched_area_entity := result.entities.get("area"):
matched_area = matched_area_entity.text.strip()
matched_floor: str | None = None
if matched_floor_entity := result.entities.get("floor"):
matched_floor = matched_floor_entity.text.strip()
if unmatched_name := unmatched_text.get("name"):
if matched_area:
# device in area
return ErrorKey.NO_ENTITY_IN_AREA, {
"entity": unmatched_name,
"area": matched_area,
}
if matched_floor:
# device on floor
return ErrorKey.NO_ENTITY_IN_FLOOR, {
"entity": unmatched_name,
"floor": matched_floor,
}
# device only
return ErrorKey.NO_ENTITY, {"entity": unmatched_name}
# Default error
return ErrorKey.NO_INTENT, {}
def _get_match_error_response(
hass: HomeAssistant,
match_error: intent.MatchFailedError,
) -> tuple[ErrorKey, dict[str, Any]]:
"""Return key and template arguments for error when target matching fails."""
constraints, result = match_error.constraints, match_error.result
reason = result.no_match_reason
if (
reason
in (intent.MatchFailedReason.DEVICE_CLASS, intent.MatchFailedReason.DOMAIN)
) and constraints.device_classes:
device_class = next(iter(constraints.device_classes)) # first device class
if constraints.area_name:
# device_class in area
return ErrorKey.NO_DEVICE_CLASS_IN_AREA, {
"device_class": device_class,
"area": constraints.area_name,
}
# device_class only
return ErrorKey.NO_DEVICE_CLASS, {"device_class": device_class}
if (reason == intent.MatchFailedReason.DOMAIN) and constraints.domains:
domain = next(iter(constraints.domains)) # first domain
if constraints.area_name:
# domain in area
return ErrorKey.NO_DOMAIN_IN_AREA, {
"domain": domain,
"area": constraints.area_name,
}
if constraints.floor_name:
# domain in floor
return ErrorKey.NO_DOMAIN_IN_FLOOR, {
"domain": domain,
"floor": constraints.floor_name,
}
# domain only
return ErrorKey.NO_DOMAIN, {"domain": domain}
if reason == intent.MatchFailedReason.DUPLICATE_NAME:
if constraints.floor_name:
# duplicate on floor
return ErrorKey.DUPLICATE_ENTITIES_IN_FLOOR, {
"entity": result.no_match_name,
"floor": constraints.floor_name,
}
if constraints.area_name:
# duplicate on area
return ErrorKey.DUPLICATE_ENTITIES_IN_AREA, {
"entity": result.no_match_name,
"area": constraints.area_name,
}
return ErrorKey.DUPLICATE_ENTITIES, {"entity": result.no_match_name}
if reason == intent.MatchFailedReason.INVALID_AREA:
# Invalid area name
return ErrorKey.NO_AREA, {"area": result.no_match_name}
if reason == intent.MatchFailedReason.INVALID_FLOOR:
# Invalid floor name
return ErrorKey.NO_FLOOR, {"floor": result.no_match_name}
if reason == intent.MatchFailedReason.FEATURE:
# Feature not supported by entity
return ErrorKey.FEATURE_NOT_SUPPORTED, {}
if reason == intent.MatchFailedReason.STATE:
# Entity is not in correct state
assert constraints.states
state = next(iter(constraints.states))
return ErrorKey.ENTITY_WRONG_STATE, {"state": state}
if reason == intent.MatchFailedReason.ASSISTANT:
# Not exposed
if constraints.name:
if constraints.area_name:
return ErrorKey.NO_ENTITY_IN_AREA_EXPOSED, {
"entity": constraints.name,
"area": constraints.area_name,
}
if constraints.floor_name:
return ErrorKey.NO_ENTITY_IN_FLOOR_EXPOSED, {
"entity": constraints.name,
"floor": constraints.floor_name,
}
return ErrorKey.NO_ENTITY_EXPOSED, {"entity": constraints.name}
if constraints.device_classes:
device_class = next(iter(constraints.device_classes))
if constraints.area_name:
return ErrorKey.NO_DEVICE_CLASS_IN_AREA_EXPOSED, {
"device_class": device_class,
"area": constraints.area_name,
}
if constraints.floor_name:
return ErrorKey.NO_DEVICE_CLASS_IN_FLOOR_EXPOSED, {
"device_class": device_class,
"floor": constraints.floor_name,
}
return ErrorKey.NO_DEVICE_CLASS_EXPOSED, {"device_class": device_class}
if constraints.domains:
domain = next(iter(constraints.domains))
if constraints.area_name:
return ErrorKey.NO_DOMAIN_IN_AREA_EXPOSED, {
"domain": domain,
"area": constraints.area_name,
}
if constraints.floor_name:
return ErrorKey.NO_DOMAIN_IN_FLOOR_EXPOSED, {
"domain": domain,
"floor": constraints.floor_name,
}
return ErrorKey.NO_DOMAIN_EXPOSED, {"domain": domain}
# Default error
return ErrorKey.NO_INTENT, {}
def _collect_list_references(expression: Expression, list_names: set[str]) -> None:
"""Collect list reference names recursively."""
if isinstance(expression, Group):
for item in expression.items:
_collect_list_references(item, list_names)
elif isinstance(expression, ListReference):
# {list}
list_names.add(expression.slot_name)
def _get_debug_targets(
hass: HomeAssistant,
result: RecognizeResult,
) -> Iterable[tuple[State, bool]]:
"""Yield state/is_matched pairs for a hassil recognition."""
entities = result.entities
name: str | None = None
area_name: str | None = None
domains: set[str] | None = None
device_classes: set[str] | None = None
state_names: set[str] | None = None
if "name" in entities:
name = str(entities["name"].value)
if "area" in entities:
area_name = str(entities["area"].value)
if "domain" in entities:
domains = set(cv.ensure_list(entities["domain"].value))
if "device_class" in entities:
device_classes = set(cv.ensure_list(entities["device_class"].value))
if "state" in entities:
# HassGetState only
state_names = set(cv.ensure_list(entities["state"].value))
if (
(name is None)
and (area_name is None)
and (not domains)
and (not device_classes)
and (not state_names)
):
# Avoid "matching" all entities when there is no filter
return
states = intent.async_match_states(
hass,
name=name,
area_name=area_name,
domains=domains,
device_classes=device_classes,
)
for state in states:
# For queries, a target is "matched" based on its state
is_matched = (state_names is None) or (state.state in state_names)
yield state, is_matched
def _get_unmatched_slots(
result: RecognizeResult,
) -> dict[str, str | int | float]:
"""Return a dict of unmatched text/range slot entities."""
unmatched_slots: dict[str, str | int | float] = {}
for entity in result.unmatched_entities_list:
if isinstance(entity, UnmatchedTextEntity):
if entity.text == MISSING_ENTITY:
# Don't report <missing> since these are just missing context
# slots.
continue
unmatched_slots[entity.name] = entity.text
elif isinstance(entity, UnmatchedRangeEntity):
unmatched_slots[entity.name] = entity.value
return unmatched_slots