def do_intents_to_graph(intents, slot_names, targets):
sentences, replacements = ini_jsgf.split_rules(intents)
# Load slot values
for slot_name in slot_names:
slot_path = slots_dir / slot_name
assert slot_path.is_file(), f"Missing slot file at {slot_path}"
# Parse each non-empty line as a JSGF sentence
slot_values = []
with open(slot_path, "r") as slot_file:
for line in slot_file:
line = line.strip()
if line:
sentence = jsgf.Sentence.parse(line)
slot_values.append(sentence)
# Replace $slot with sentences
replacements[f"${slot_name}"] = slot_values
if profile.get("intent.replace_numbers", True):
# Replace numbers in parsed sentences
for intent_sentences in sentences.values():
for sentence in intent_sentences:
jsgf.walk_expression(sentence, number_transform,
replacements)
# Convert to directed graph
graph = intents_to_graph(intents, replacements)
# Write graph to JSON file
json_graph = graph_to_json(graph)
with open(targets[0], "w") as graph_file:
json.dump(json_graph, graph_file)
def do_intents_to_graph(sentences, slot_names, replacements, targets):
# Replace actual numbers
if profile.get("intent.replace_numbers", True):
# Replace numbers in parsed sentences
for intent_sentences in sentences.values():
for sentence in intent_sentences:
jsgf.walk_expression(sentence, number_transform, replacements)
# Determine whether word casing has to be fixed
transform = None
if word_casing == "upper":
transform = str.upper
elif word_casing == "lower":
transform = str.lower
if transform:
def fix_case(word):
if isinstance(word, jsgf.Word):
word.text = transform(word.text)
return word
# Fix casing
for intent_sentences in sentences.values():
for sentence in intent_sentences:
jsgf.walk_expression(sentence, fix_case, replacements)
# Convert to directed graph
graph = intents_to_graph(sentences, replacements)
# Write graph to JSON file
json_graph = graph_to_json(graph)
with open(targets[0], "w") as graph_file:
json.dump(json_graph, graph_file)
async def handle_nlu_train(
self, train: NluTrain, site_id: str = "default"
) -> typing.AsyncIterable[
typing.Union[typing.Tuple[NluTrainSuccess, TopicArgs], NluError]
]:
"""Re-trains NLU system"""
try:
# Load gzipped graph pickle
_LOGGER.debug("Loading %s", train.graph_path)
with gzip.GzipFile(train.graph_path, mode="rb") as graph_gzip:
intent_graph = nx.readwrite.gpickle.read_gpickle(graph_gzip)
# Get JSON intent graph
json_graph = rhasspynlu.graph_to_json(intent_graph)
if self.nlu_train_url:
# Remote NLU server
_LOGGER.debug(self.nlu_train_url)
async with self.http_session.post(
self.nlu_train_url, json=json_graph, ssl=self.ssl_context
) as response:
# No data expected in response
response.raise_for_status()
elif self.nlu_train_command:
# Local NLU training command
_LOGGER.debug(self.nlu_train_command)
proc = await asyncio.create_subprocess_exec(
*self.nlu_train_command,
stdin=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
output, error = await proc.communicate(json.dumps(json_graph).encode())
if output:
_LOGGER.debug(output.decode())
if error:
_LOGGER.debug(error.decode())
else:
_LOGGER.warning("Can't train NLU system. No train URL or command.")
# Report success
yield (NluTrainSuccess(id=train.id), {"site_id": site_id})
except Exception as e:
_LOGGER.exception("handle_nlu_train")
yield NluError(
error=str(e),
context=f"url='{self.nlu_train_url}', command='{self.nlu_train_command}'",
site_id=site_id,
session_id=train.id,
)
def do_intents_to_graph(sentences, slot_names, replacements, targets):
# Replace actual numbers
if profile.get("intent.replace_numbers", True):
# Replace numbers in parsed sentences
for intent_sentences in sentences.values():
for sentence in intent_sentences:
jsgf.walk_expression(sentence, number_transform,
replacements)
if word_transform:
# Fix casing
for intent_sentences in sentences.values():
for sentence in intent_sentences:
jsgf.walk_expression(sentence, fix_word_case, replacements)
# Convert to directed graph
graph = intents_to_graph(sentences, replacements)
# Write graph to JSON file
json_graph = graph_to_json(graph)
with open(targets[0], "w") as graph_file:
json.dump(json_graph, graph_file)