bobmaster/matrix-stt-bot
1
0
Fork 0
Code Issues 1 Pull requests Projects Releases 3 Packages Wiki Activity

Compare commits

base: bobmaster:main
Branches Tags
bobmaster:main
bobmaster:v1.1.1
bobmaster:v1.1.0
bobmaster:v1.0.0
..
compare: bobmaster:v1.0.0
Branches Tags
bobmaster:main
bobmaster:v1.1.1
bobmaster:v1.1.0
bobmaster:v1.0.0

No commits in common. "main" and "v1.0.0" have entirely different histories.
main ... v1.0.0

5 changed files with 181 additions and 236 deletions
Show stats Expand all files Collapse all files

5
README.md
View file

@ -98,7 +98,4 @@ docker compose up -d
## Thanks
1. https://github.com/guillaumekln/faster-whisper
2. https://github.com/poljar/matrix-nio
3. https://github.com/8go/matrix-commander
<a href="https://jb.gg/OpenSourceSupport" target="_blank">
<img src="https://resources.jetbrains.com/storage/products/company/brand/logos/jb_beam.png" alt="JetBrains Logo (Main) logo." width="200" height="200">
</a>
3. https://github.com/8go/matrix-commander

299
bot.py
View file

@ -1,5 +1,4 @@
import os
import signal
import sys
import traceback
from typing import Union, Optional
@ -7,27 +6,12 @@ import aiofiles
import asyncio
import uuid
import json
from nio import (
AsyncClient,
AsyncClientConfig,
InviteMemberEvent,
JoinError,
KeyVerificationCancel,
KeyVerificationEvent,
DownloadError,
KeyVerificationKey,
KeyVerificationMac,
KeyVerificationStart,
LocalProtocolError,
LoginResponse,
MatrixRoom,
MegolmEvent,
RoomMessageAudio,
RoomEncryptedAudio,
ToDeviceError,
crypto,
EncryptionError,
)
from nio import (AsyncClient, AsyncClientConfig, InviteMemberEvent, JoinError,
KeyVerificationCancel, KeyVerificationEvent, DownloadError,
KeyVerificationKey, KeyVerificationMac, KeyVerificationStart,
LocalProtocolError, LoginResponse, MatrixRoom, MegolmEvent,
RoomMessageAudio, RoomEncryptedAudio, ToDeviceError, crypto,
EncryptionError)
from nio.store.database import SqliteStore
from faster_whisper import WhisperModel
@ -56,11 +40,12 @@ class Bot:
num_workers: int = 1,
download_root: str = "models",
):
if homeserver is None or user_id is None or device_id is None:
if (homeserver is None or user_id is None
or device_id is None):
logger.warning("homeserver && user_id && device_id is required")
sys.exit(1)
if password is None and access_token is None:
if (password is None and access_token is None):
logger.warning("password or access_toekn is required")
sys.exit(1)
@ -102,57 +87,73 @@ class Bot:
# initialize AsyncClient object
self.store_path = os.getcwd()
self.config = AsyncClientConfig(
store=SqliteStore,
store_name="db",
store_sync_tokens=True,
encryption_enabled=True,
)
self.client = AsyncClient(
homeserver=self.homeserver,
user=self.user_id,
device_id=self.device_id,
config=self.config,
store_path=self.store_path,
)
self.config = AsyncClientConfig(store=SqliteStore,
store_name="db",
store_sync_tokens=True,
encryption_enabled=True,
)
self.client = AsyncClient(homeserver=self.homeserver, user=self.user_id, device_id=self.device_id,
config=self.config, store_path=self.store_path,)
if self.access_token is not None:
self.client.access_token = self.access_token
# setup event callbacks
self.client.add_event_callback(
self.message_callback,
(
RoomMessageAudio,
RoomEncryptedAudio,
),
)
self.client.add_event_callback(self.decryption_failure, (MegolmEvent,))
self.client.add_event_callback(self.invite_callback, (InviteMemberEvent,))
self.message_callback, (RoomMessageAudio, RoomEncryptedAudio, ))
self.client.add_event_callback(
self.decryption_failure, (MegolmEvent, ))
self.client.add_event_callback(
self.invite_callback, (InviteMemberEvent, ))
self.client.add_to_device_callback(
self.to_device_callback, (KeyVerificationEvent,)
)
self.to_device_callback, (KeyVerificationEvent, ))
# intialize whisper model
if not os.path.exists(os.path.join(self.download_root, "model.txt")):
# that means we have not downloaded the model yet
logger.info("Model downloading......")
self.model_size_or_path = self.model_size
with open(os.path.join(self.download_root, "model.txt"), "w") as f:
f.write(self.model_size)
f.close()
else:
# model exists
f = open(os.path.join(self.download_root, "model.txt"), "r")
old_model = f.read()
if old_model != self.model_size:
# model size changed
self.model_size_or_path = self.model_size
f.close()
with open(os.path.join(self.download_root, "model.txt"), "w") as f:
f.write(self.model_size)
f.close()
else:
# use existed model
self.model_size_or_path = self.download_root
self.model = WhisperModel(
model_size_or_path=self.model_size,
model_size_or_path=self.model_size_or_path,
device=self.device,
compute_type=self.compute_type,
cpu_threads=self.cpu_threads,
num_workers=self.num_workers,
download_root=self.download_root,
)
download_root=self.download_root,)
async def close(self, task: asyncio.Task = None) -> None:
def __del__(self):
try:
loop = asyncio.get_running_loop()
except RuntimeError as e:
loop = asyncio.new_event_loop()
asyncio.set_event_loop(loop)
loop.run_until_complete(self._close())
async def _close(self):
await self.client.close()
task.cancel()
logger.info("Bot closed!")
logger.info("Bot stopped!")
# message_callback event
async def message_callback(
self, room: MatrixRoom, event: Union[RoomMessageAudio, RoomEncryptedAudio]
) -> None:
# message_callback event
async def message_callback(self, room: MatrixRoom,
event: Union[RoomMessageAudio, RoomEncryptedAudio]) -> None:
if self.room_id is None:
room_id = room.room_id
else:
@ -199,8 +200,12 @@ class Bot:
await f.write(
crypto.attachments.decrypt_attachment(
media_data,
event.source["content"]["file"]["key"]["k"],
event.source["content"]["file"]["hashes"]["sha256"],
event.source["content"]["file"]["key"][
"k"
],
event.source["content"]["file"]["hashes"][
"sha256"
],
event.source["content"]["file"]["iv"],
)
)
@ -233,8 +238,8 @@ class Bot:
return
logger.error(
f"Failed to decrypt message: {event.event_id} from {event.sender} in {room.room_id}\n"
+ "Please make sure the bot current session is verified"
f"Failed to decrypt message: {event.event_id} from {event.sender} in {room.room_id}\n" +
"Please make sure the bot current session is verified"
)
# invite_callback event
@ -250,8 +255,7 @@ class Bot:
if type(result) == JoinError:
logger.error(
f"Error joining room {room.room_id} (attempt %d): %s",
attempt,
result.message,
attempt, result.message,
)
else:
break
@ -273,11 +277,11 @@ class Bot:
try:
client = self.client
logger.debug(
f"Device Event of type {type(event)} received in " "to_device_cb()."
)
f"Device Event of type {type(event)} received in "
"to_device_cb().")
if isinstance(event, KeyVerificationStart): # first step
"""first step: receive KeyVerificationStart
""" first step: receive KeyVerificationStart
KeyVerificationStart(
source={'content':
{'method': 'm.sas.v1',
@ -307,14 +311,13 @@ class Bot:
"""
if "emoji" not in event.short_authentication_string:
estr = (
"Other device does not support emoji verification "
f"{event.short_authentication_string}. Aborting."
)
estr = ("Other device does not support emoji verification "
f"{event.short_authentication_string}. Aborting.")
print(estr)
logger.info(estr)
return
resp = await client.accept_key_verification(event.transaction_id)
resp = await client.accept_key_verification(
event.transaction_id)
if isinstance(resp, ToDeviceError):
estr = f"accept_key_verification() failed with {resp}"
print(estr)
@ -330,7 +333,7 @@ class Bot:
logger.info(estr)
elif isinstance(event, KeyVerificationCancel): # anytime
"""at any time: receive KeyVerificationCancel
""" at any time: receive KeyVerificationCancel
KeyVerificationCancel(source={
'content': {'code': 'm.mismatched_sas',
'reason': 'Mismatched authentication string',
@ -347,15 +350,13 @@ class Bot:
# client.cancel_key_verification(tx_id, reject=False)
# here. The SAS flow is already cancelled.
# We only need to inform the user.
estr = (
f"Verification has been cancelled by {event.sender} "
f'for reason "{event.reason}".'
)
estr = (f"Verification has been cancelled by {event.sender} "
f"for reason \"{event.reason}\".")
print(estr)
logger.info(estr)
elif isinstance(event, KeyVerificationKey): # second step
"""Second step is to receive KeyVerificationKey
""" Second step is to receive KeyVerificationKey
KeyVerificationKey(
source={'content': {
'key': 'SomeCryptoKey',
@ -380,44 +381,42 @@ class Bot:
# automatic match, so we use y
yn = "y"
if yn.lower() == "y":
estr = (
"Match! The verification for this " "device will be accepted."
)
estr = ("Match! The verification for this "
"device will be accepted.")
print(estr)
logger.info(estr)
resp = await client.confirm_short_auth_string(event.transaction_id)
resp = await client.confirm_short_auth_string(
event.transaction_id)
if isinstance(resp, ToDeviceError):
estr = "confirm_short_auth_string() " f"failed with {resp}"
estr = ("confirm_short_auth_string() "
f"failed with {resp}")
print(estr)
logger.info(estr)
elif yn.lower() == "n": # no, don't match, reject
estr = (
"No match! Device will NOT be verified "
"by rejecting verification."
)
estr = ("No match! Device will NOT be verified "
"by rejecting verification.")
print(estr)
logger.info(estr)
resp = await client.cancel_key_verification(
event.transaction_id, reject=True
)
event.transaction_id, reject=True)
if isinstance(resp, ToDeviceError):
estr = f"cancel_key_verification failed with {resp}"
estr = (f"cancel_key_verification failed with {resp}")
print(estr)
logger.info(estr)
else: # C or anything for cancel
estr = "Cancelled by user! Verification will be " "cancelled."
estr = ("Cancelled by user! Verification will be "
"cancelled.")
print(estr)
logger.info(estr)
resp = await client.cancel_key_verification(
event.transaction_id, reject=False
)
event.transaction_id, reject=False)
if isinstance(resp, ToDeviceError):
estr = f"cancel_key_verification failed with {resp}"
estr = (f"cancel_key_verification failed with {resp}")
print(estr)
logger.info(estr)
elif isinstance(event, KeyVerificationMac): # third step
"""Third step is to receive KeyVerificationMac
""" Third step is to receive KeyVerificationMac
KeyVerificationMac(
source={'content': {
'mac': {'ed25519:DEVICEIDXY': 'SomeKey1',
@ -437,11 +436,9 @@ class Bot:
todevice_msg = sas.get_mac()
except LocalProtocolError as e:
# e.g. it might have been cancelled by ourselves
estr = (
f"Cancelled or protocol error: Reason: {e}.\n"
f"Verification with {event.sender} not concluded. "
"Try again?"
)
estr = (f"Cancelled or protocol error: Reason: {e}.\n"
f"Verification with {event.sender} not concluded. "
"Try again?")
print(estr)
logger.info(estr)
else:
@ -450,31 +447,25 @@ class Bot:
estr = f"to_device failed with {resp}"
print(estr)
logger.info(estr)
estr = (
f"sas.we_started_it = {sas.we_started_it}\n"
f"sas.sas_accepted = {sas.sas_accepted}\n"
f"sas.canceled = {sas.canceled}\n"
f"sas.timed_out = {sas.timed_out}\n"
f"sas.verified = {sas.verified}\n"
f"sas.verified_devices = {sas.verified_devices}\n"
)
estr = (f"sas.we_started_it = {sas.we_started_it}\n"
f"sas.sas_accepted = {sas.sas_accepted}\n"
f"sas.canceled = {sas.canceled}\n"
f"sas.timed_out = {sas.timed_out}\n"
f"sas.verified = {sas.verified}\n"
f"sas.verified_devices = {sas.verified_devices}\n")
print(estr)
logger.info(estr)
estr = (
"Emoji verification was successful!\n"
"Initiate another Emoji verification from "
"another device or room if desired. "
"Or if done verifying, hit Control-C to stop the "
"bot in order to restart it as a service or to "
"run it in the background."
)
estr = ("Emoji verification was successful!\n"
"Initiate another Emoji verification from "
"another device or room if desired. "
"Or if done verifying, hit Control-C to stop the "
"bot in order to restart it as a service or to "
"run it in the background.")
print(estr)
logger.info(estr)
else:
estr = (
f"Received unexpected event type {type(event)}. "
f"Event is {event}. Event will be ignored."
)
estr = (f"Received unexpected event type {type(event)}. "
f"Event is {event}. Event will be ignored.")
print(estr)
logger.info(estr)
except BaseException:
@ -483,6 +474,7 @@ class Bot:
logger.info(estr)
# bot login
async def login(self) -> None:
if self.access_token is not None:
logger.info("Login via access_token")
@ -510,14 +502,14 @@ class Bot:
# import keys
async def import_keys(self):
resp = await self.client.import_keys(
self.import_keys_path, self.import_keys_password
self.import_keys_path,
self.import_keys_password
)
if isinstance(resp, EncryptionError):
logger.error(f"import_keys failed with {resp}")
else:
logger.info(
f"import_keys success, please remove import_keys configuration!!!"
)
f"import_keys success, please remove import_keys configuration!!!")
# whisper function
def transcribe(self, filename: str) -> str:
@ -537,33 +529,30 @@ async def main():
config = json.load(fp)
bot = Bot(
homeserver=config.get("homeserver"),
user_id=config.get("user_id"),
password=config.get("password"),
device_id=config.get("device_id"),
room_id=config.get("room_id"),
access_token=config.get("access_token"),
import_keys_path=config.get("import_keys_path"),
import_keys_password=config.get("import_keys_password"),
model_size=config.get("model_size"),
device=config.get("device"),
compute_type=config.get("compute_type"),
cpu_threads=config.get("cpu_threads"),
num_workers=config.get("num_workers"),
download_root=config.get("download_root"),
homeserver=config.get('homeserver'),
user_id=config.get('user_id'),
password=config.get('password'),
device_id=config.get('device_id'),
room_id=config.get('room_id'),
access_token=config.get('access_token'),
import_keys_path=config.get('import_keys_path'),
import_keys_password=config.get('import_keys_password'),
model_size=config.get('model_size'),
device=config.get('device'),
compute_type=config.get('compute_type'),
cpu_threads=config.get('cpu_threads'),
num_workers=config.get('num_workers'),
download_root=config.get('download_root'),
)
if (
config.get("import_keys_path")
and config.get("import_keys_password") is not None
):
if config.get('import_keys_path') and config.get('import_keys_password') is not None:
need_import_keys = True
else:
bot = Bot(
homeserver=os.environ.get("HOMESERVER"),
user_id=os.environ.get("USER_ID"),
password=os.environ.get("PASSWORD"),
homeserver=os.environ.get('HOMESERVER'),
user_id=os.environ.get('USER_ID'),
password=os.environ.get('PASSWORD'),
device_id=os.environ.get("DEVICE_ID"),
room_id=os.environ.get("ROOM_ID"),
access_token=os.environ.get("ACCESS_TOKEN"),
@ -576,10 +565,7 @@ async def main():
num_workers=os.environ.get("NUM_WORKERS"),
download_root=os.environ.get("DOWNLOAD_ROOT"),
)
if (
os.environ.get("IMPORT_KEYS_PATH")
and os.environ.get("IMPORT_KEYS_PASSWORD") is not None
):
if os.environ.get("IMPORT_KEYS_PATH") and os.environ.get("IMPORT_KEYS_PASSWORD") is not None:
need_import_keys = True
await bot.login()
@ -587,21 +573,8 @@ async def main():
logger.info("start import_keys process, this may take a while...")
await bot.import_keys()
sync_task = asyncio.create_task(bot.sync_forever())
await bot.sync_forever()
# handle signal interrupt
loop = asyncio.get_running_loop()
for signame in (
"SIGINT",
"SIGTERM",
):
loop.add_signal_handler(
getattr(signal, signame), lambda: asyncio.create_task(bot.close(sync_task))
)
await sync_task
if __name__ == "__main__":
if __name__ == '__main__':
logger.info("Bot started!")
asyncio.run(main())

10
log.py
View file

@ -9,19 +9,17 @@ def getlogger():
# create handlers
warn_handler = logging.StreamHandler()
info_handler = logging.StreamHandler()
error_handler = logging.FileHandler("bot.log", mode="a")
error_handler = logging.FileHandler('bot.log', mode='a')
warn_handler.setLevel(logging.WARNING)
error_handler.setLevel(logging.ERROR)
info_handler.setLevel(logging.INFO)
# create formatters
warn_format = logging.Formatter(
"%(asctime)s - %(funcName)s - %(levelname)s - %(message)s"
)
'%(asctime)s - %(funcName)s - %(levelname)s - %(message)s')
error_format = logging.Formatter(
"%(asctime)s - %(name)s - %(funcName)s - %(levelname)s - %(message)s"
)
info_format = logging.Formatter("%(asctime)s - %(levelname)s - %(message)s")
'%(asctime)s - %(name)s - %(funcName)s - %(levelname)s - %(message)s')
info_format = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s')
# set formatter
warn_handler.setFormatter(warn_format)

52
requirements.txt
View file

@ -1,5 +1,5 @@
aiofiles==23.1.0
aiohttp==3.8.5
aiohttp==3.8.4
aiohttp-socks==0.7.1
aiosignal==1.3.1
async-timeout==4.0.2
@ -7,48 +7,44 @@ atomicwrites==1.4.1
attrs==23.1.0
av==10.0.0
cachetools==4.2.4
certifi==2023.5.7
certifi==2022.12.7
cffi==1.15.1
charset-normalizer==3.2.0
charset-normalizer==3.1.0
coloredlogs==15.0.1
ctranslate2==3.17.0
faster-whisper==0.7.0
filelock==3.12.2
flatbuffers==23.5.26
frozenlist==1.4.0
fsspec==2023.6.0
ctranslate2==3.12.0
faster-whisper @ git+https://github.com/guillaumekln/faster-whisper.git@3adcc12d0f91369446a624e33185c555facc8ed2
filelock==3.12.0
flatbuffers==23.3.3
frozenlist==1.3.3
future==0.18.3
h11==0.14.0
h2==4.1.0
hpack==4.0.0
huggingface-hub==0.16.4
huggingface-hub @ git+https://github.com/huggingface/huggingface_hub.git@a433410ac37f0efaacb92b09684efa3ae46ed3fd
humanfriendly==10.0
hyperframe==6.0.1
idna==3.4
jsonschema==4.18.4
jsonschema-specifications==2023.7.1
jsonschema==4.17.3
Logbook==1.5.3
matrix-nio==0.21.2
matrix-nio[e2e]==0.20.2
mpmath==1.3.0
multidict==6.0.4
numpy==1.25.1
onnxruntime==1.15.1
numpy==1.24.2
onnxruntime==1.14.1
packaging==23.1
peewee==3.16.2
protobuf==4.23.4
peewee==3.16.1
protobuf==4.22.3
pycparser==2.21
pycryptodome==3.18.0
pycryptodome==3.17
pyrsistent==0.19.3
python-olm==3.2.15
python-socks==2.3.0
PyYAML==6.0.1
referencing==0.30.0
requests==2.31.0
rpds-py==0.9.2
sympy==1.12
python-olm==3.1.3
python-socks==2.2.0
PyYAML==6.0
requests==2.28.2
sympy==1.11.1
tokenizers==0.13.3
tqdm==4.65.0
typing_extensions==4.7.1
typing_extensions==4.5.0
unpaddedbase64==2.1.0
urllib3==2.0.4
yarl==1.9.2
urllib3==1.26.15
yarl==1.8.2

51
send_message.py
View file

@ -1,42 +1,23 @@
from nio import AsyncClient
async def send_room_message(
client: AsyncClient,
room_id: str,
reply_message: str,
sender_id: str = "",
reply_to_event_id: str = "",
) -> None:
NORMAL_BODY = content = {
"msgtype": "m.text",
"body": reply_message,
}
if reply_to_event_id == "":
content = NORMAL_BODY
async def send_room_message(client: AsyncClient,
room_id: str,
reply_message: str,
sender_id: str = '',
reply_to_event_id: str = '',
) -> None:
NORMAL_BODY = content = {"msgtype": "m.text", "body": reply_message, }
if reply_to_event_id == '':
content = NORMAL_BODY
else:
body = r"> <" + sender_id + r"> sent an audio file.\n\n" + reply_message
format = r"org.matrix.custom.html"
formatted_body = (
r'<mx-reply><blockquote><a href="https://matrix.to/#/'
+ room_id
+ r"/"
+ reply_to_event_id
+ r'">In reply to</a> <a href="https://matrix.to/#/'
+ sender_id
+ r'">'
+ sender_id
+ r"</a><br>sent an audio file.</blockquote></mx-reply>"
+ reply_message
)
body = r'> <' + sender_id + r'> sent an audio file.\n\n' + reply_message
format = r'org.matrix.custom.html'
formatted_body = r'<mx-reply><blockquote><a href="https://matrix.to/#/' + room_id + r'/' + reply_to_event_id \
+ r'">In reply to</a> <a href="https://matrix.to/#/' + sender_id + r'">' + sender_id \
+ r'</a><br>sent an audio file.</blockquote></mx-reply>' + reply_message
content = {
"msgtype": "m.text",
"body": body,
"format": format,
"formatted_body": formatted_body,
"m.relates_to": {"m.in_reply_to": {"event_id": reply_to_event_id}},
}
content = {"msgtype": "m.text", "body": body, "format": format, "formatted_body": formatted_body,
"m.relates_to": {"m.in_reply_to": {"event_id": reply_to_event_id}}, }
await client.room_send(
room_id,
message_type="m.room.message",