ovos-backend-manager/ovos_backend_manager/datasets.py
2023-04-08 00:50:24 +01:00

517 lines
19 KiB
Python

import json
import os
import time
from base64 import b64encode
from pywebio.input import actions, file_upload, input_group, textarea, select
from pywebio.output import put_text, put_code, use_scope, put_markdown, popup, put_image, put_file, put_html, \
put_buttons, put_table
from ovos_backend_manager.configuration import CONFIGURATION, DB
def _render_ww(rec_id):
def on_tag(bt):
data["tag"] = bt
DB.update_ww_recording(rec_id, tag=bt)
_render_ww(rec_id)
with use_scope("main_view", clear=True):
data = DB.get_ww_recording(rec_id)
data["tag"] = data.get("tag") or "untagged"
# TODO - get binary_data directly
if os.path.isfile(data["path"]):
content = open(data["path"], 'rb').read()
html = f"""
<audio controls src="data:audio/x-wav;base64,{b64encode(content).decode('ascii')}" />
"""
put_table([
['metadata', put_code(json.dumps(data, indent=4), "json")],
['playback', put_html(html)],
['file', put_file(data["path"].split("/")[-1], content, 'Download')],
['classify', put_buttons(["wake_word", "speech", "noise", "silence"],
onclick=on_tag)]
])
else:
put_table([
['metadata', put_code(json.dumps(data, indent=4), "json")],
['playback', put_markdown("**WARNING** - file not found")],
['file', put_markdown("**WARNING** - file not found")],
['classify', put_buttons(["untagged", "wake_word", "speech", "noise", "silence"],
onclick=on_tag)]
])
def ww_select(back_handler=None, uuid=None, ww=None):
buttons = []
if not len(DB.list_ww_recordings()):
with use_scope("main_view", clear=True):
put_text("No wake words uploaded yet!")
datasets_menu(back_handler=back_handler)
return
for m in DB.list_ww_recordings():
if uuid is not None and m["uuid"] != uuid:
continue
if ww is not None and m["transcription"] != ww:
continue
name = f"{m['recording_id']}-{m['transcription']}"
buttons.append({'label': name, 'value': m['recording_id']})
if len(buttons) == 0:
with use_scope("main_view", clear=True):
put_text("No wake words uploaded from this device yet!")
ww_menu(back_handler=back_handler)
return
elif back_handler:
buttons.insert(0, {'label': '<- Go Back', 'value': "main"})
rec_id = actions(label="Select a WakeWord recording", buttons=buttons)
if rec_id == "main": # go back
ww_menu(back_handler=back_handler)
else:
_render_ww(rec_id)
ww_select(back_handler=back_handler, ww=ww, uuid=uuid)
def utt_select(back_handler=None, uuid=None, utt=None):
buttons = []
if not len(DB.list_stt_recordings()):
with use_scope("main_view", clear=True):
put_text("No utterances uploaded yet!")
datasets_menu(back_handler=back_handler)
return
for m in DB.list_stt_recordings():
if uuid is not None and m["uuid"] != uuid:
continue
if utt is not None and m["transcription"] != utt:
continue
name = f"{m['recording_id']}-{m['transcription']}"
buttons.append({'label': name, 'value': m['recording_id']})
if len(buttons) == 0:
with use_scope("main_view", clear=True):
put_text("No utterances uploaded from this device yet!")
utt_menu(back_handler=back_handler)
return
else:
if back_handler:
buttons.insert(0, {'label': '<- Go Back', 'value': "main"})
rec_id = actions(label="Select a Utterance recording",
buttons=buttons)
if rec_id == "main":
utt_menu(back_handler=back_handler)
return
with use_scope("main_view", clear=True):
# opt is recording_id
data = DB.get_stt_recording(rec_id)
put_code(json.dumps(data, indent=4), "json")
# TODO - get binary data from api
if os.path.isfile(data["path"]):
content = open(data["path"], 'rb').read()
html = f"""<audio controls src="data:audio/x-wav;base64,{b64encode(content).decode('ascii')}" />"""
put_html(html)
put_file(data["path"].split("/")[-1], content, 'Download Audio')
else:
put_markdown("**WARNING** - audio file not found")
utt_select(back_handler=back_handler, uuid=uuid, utt=utt)
def device_select(back_handler=None, ww=True):
devices = {device["uuid"]: f"{device['name']}@{device['device_location']}"
for device in DB.list_devices()}
buttons = [{'label': "All Devices", 'value': "all"},
{'label': "Unknown Devices", 'value': "AnonDevice"}] + \
[{'label': d, 'value': uuid} for uuid, d in devices.items()]
if back_handler:
buttons.insert(0, {'label': '<- Go Back', 'value': "main"})
if devices:
uuid = actions(label="What device would you like to inspect?",
buttons=buttons)
if uuid == "main":
datasets_menu(back_handler=back_handler)
return
else:
if uuid == "all":
uuid = None
if ww:
ww_select(uuid=uuid, back_handler=back_handler)
else:
utt_select(uuid=uuid, back_handler=back_handler)
else:
with use_scope("main_view", clear=True):
put_text("No devices paired yet!")
if ww:
ww_menu(back_handler=back_handler)
else:
utt_menu(back_handler=back_handler)
def ww_opts(back_handler=None, uuid=None):
wws = list(set([ww["transcription"] for ww in DB.list_ww_recordings()]))
buttons = [{'label': "All Wake Words", 'value': "all"}] + \
[{'label': ww, 'value': ww} for ww in wws]
if back_handler:
buttons.insert(0, {'label': '<- Go Back', 'value': "main"})
if wws:
ww = actions(label="What wake word would you like to inspect?", buttons=buttons)
if ww == "main":
datasets_menu(back_handler=back_handler)
return
if ww == "all":
ww = None
ww_select(ww=ww, back_handler=back_handler, uuid=uuid)
else:
with use_scope("main_view", clear=True):
put_text("No wake words uploaded yet!")
ww_menu(back_handler=back_handler)
def utt_opts(back_handler=None, uuid=None):
utts = list(set([ww["transcription"] for ww in DB.list_stt_recordings()]))
buttons = [{'label': "All Utterances", 'value': "all"}] + \
[{'label': ww, 'value': ww} for ww in utts]
if back_handler:
buttons.insert(0, {'label': '<- Go Back', 'value': "main"})
if utts:
utt = actions(label="What utterance would you like to inspect?", buttons=buttons)
if utt == "main":
datasets_menu(back_handler=back_handler)
return
if utt == "all":
utt = None
utt_select(utt=utt, back_handler=back_handler, uuid=uuid)
else:
with use_scope("main_view", clear=True):
put_text("No recordings uploaded yet!")
utt_menu(back_handler=back_handler)
def _render_ww_tagger(selected_idx, selected_wws, untagged_only=False):
def on_tag(tag):
nonlocal selected_idx, selected_wws
if all((ww.get("tag") != "untagged" for ww in selected_wws)):
if untagged_only:
popup("No more wake words to tag!")
return
if tag == "Skip ->":
selected_idx += 1
if selected_idx >= len(selected_wws):
selected_idx = 0
if untagged_only and selected_wws[selected_idx]["tag"] != "untagged":
return on_tag(tag) # recurse
elif selected_idx is not None:
db_id = selected_wws[selected_idx]["recording_id"]
DB.update_ww_recording(db_id, tag=tag)
_render_ww_tagger(selected_idx, selected_wws, untagged_only=untagged_only)
def on_gender(tag):
nonlocal selected_idx, selected_wws
if selected_idx is not None:
db_id = selected_wws[selected_idx]["recording_id"]
DB.update_ww_recording(db_id, speaker_type=tag)
_render_ww_tagger(selected_idx, selected_wws, untagged_only=untagged_only)
with use_scope("main_view", clear=True):
content = open(selected_wws[selected_idx]["path"], 'rb').read()
html = f"""
<audio controls src="data:audio/x-wav;base64,{b64encode(content).decode('ascii')}" />
"""
put_table([
['wake word', selected_wws[selected_idx]["transcription"]],
['metadata', put_code(
json.dumps(selected_wws[selected_idx], indent=4), "json")],
['playback', put_html(html)],
['speaker type', put_buttons(["male", "female", "children"],
onclick=on_gender)],
['tag', put_buttons(["wake_word", "speech", "noise", "silence", "Skip ->"],
onclick=on_tag)],
])
def ww_tagger(back_handler=None, selected_wws=None, selected_idx=None, untagged_only=True):
with use_scope("logo", clear=True):
img = open(f'{os.path.dirname(__file__)}/res/wakewords.png', 'rb').read()
put_image(img)
def get_next_untagged():
nonlocal selected_idx
if untagged_only:
for idx, ww in enumerate(selected_wws):
if ww.get("tag", "untagged") == "untagged":
selected_idx = idx
break
else:
selected_idx = 0
if not selected_wws:
wws = set([w["transcription"] for w in DB.list_ww_recordings()
if os.path.isfile(w["path"])])
if not len(wws):
with use_scope("main_view", clear=True):
put_text("No wake words uploaded yet!")
datasets_menu(back_handler=back_handler)
return
current_ww = select("Target WW", wws)
selected_wws = [w for w in DB.list_ww_recordings()
if w["transcription"] == current_ww
and os.path.isfile(w["path"])]
selected_idx = 0
else:
selected_idx = selected_idx or 0
current_ww = selected_wws[selected_idx]["transcription"]
if untagged_only:
get_next_untagged()
# add "untagged" tag if needed
for idx, ww in enumerate(selected_wws):
if "tag" not in ww:
selected_wws[idx]["tag"] = "untagged"
if "speaker_type" not in ww:
selected_wws[idx]["speaker_type"] = "untagged"
_render_ww_tagger(selected_idx, selected_wws, untagged_only)
buttons = [
{'label': "Show all recordings" if untagged_only else 'Show untagged only', 'value': "toggle"},
{'label': f'Delete {current_ww} database', 'value': "delete_ww"}
]
if back_handler:
buttons.insert(0, {'label': '<- Go Back', 'value': "main"})
opt = actions(label="What would you like to do?",
buttons=buttons)
if opt == "toggle":
untagged_only = not untagged_only
get_next_untagged()
if opt == "delete_ww":
with use_scope("main_view", clear=True):
put_markdown(f"""Are you sure you want to delete the {current_ww} wake word database?
**this can not be undone**, proceed with caution!
**ALL** {current_ww} recordings will be **lost**""")
opt = actions(label=f"Delete {current_ww} database?",
buttons=[{'label': "yes", 'value': True},
{'label': "no", 'value': False}])
if opt:
for ww in selected_wws:
if os.path.isfile(ww["path"]):
os.remove(ww["path"])
rec_id = current_ww # TODO - rec_id
DB.delete_ww_recording(rec_id)
with use_scope("main_view", clear=True):
put_text(f"{current_ww} database deleted!")
ww_tagger(back_handler=back_handler, untagged_only=untagged_only)
return
if opt == "main":
with use_scope("main_view", clear=True):
datasets_menu(back_handler=back_handler)
return
ww_tagger(back_handler=back_handler,
selected_idx=selected_idx,
selected_wws=selected_wws,
untagged_only=untagged_only)
def ww_menu(back_handler=None):
with use_scope("logo", clear=True):
img = open(f'{os.path.dirname(__file__)}/res/wakewords.png', 'rb').read()
put_image(img)
buttons = [{'label': 'Inspect Devices', 'value': "dev"},
{'label': 'Inspect Wake Words', 'value': "ww"},
{'label': 'Upload Wake Word', 'value': "upload"},
{'label': 'Delete wake words database', 'value': "delete_ww"}]
if back_handler:
buttons.insert(0, {'label': '<- Go Back', 'value': "main"})
opt = actions(label="What would you like to do?",
buttons=buttons)
if opt == "dev":
device_select(back_handler=back_handler, ww=True)
if opt == "ww":
ww_opts(back_handler=back_handler)
if opt == "upload":
with use_scope("main_view", clear=True):
data = input_group("Upload wake word", [
textarea("wake word name", placeholder="hey mycroft", required=True, name="name"),
file_upload("wake word recording", name="file")
])
name = data["name"]
filename = data["file"]["filename"]
mime = data["file"]["mime_type"]
content = data["file"]["content"]
if mime != "audio/x-wav":
popup("invalid format!")
else:
os.makedirs(f"{CONFIGURATION['data_path']}/wakewords", exist_ok=True)
uuid = "AnonDevice" # TODO - allow tagging to a device
meta = {
"transcription": name,
"meta": {
"name": name,
"time": time.time(),
"accountId": "0",
"sessionId": "0",
"model": "uploaded_file",
"engine": "uploaded_file"
},
"uuid": uuid
}
rec = DB.add_ww_recording(byte_data=content, transcription=name, metadata=meta)
with popup("wake word uploaded!"):
put_code(json.dumps(meta, indent=4), "json")
if opt == "delete_ww":
with use_scope("main_view", clear=True):
put_markdown("""Are you sure you want to delete the wake word database?
**this can not be undone**, proceed with caution!
**ALL** wake word recordings will be **lost**""")
opt = actions(label="Delete wake words database?",
buttons=[{'label': "yes", 'value': True},
{'label': "no", 'value': False}])
if opt:
for rec in DB.list_ww_recordings():
DB.delete_ww_recording(rec_id=rec["recording_id"])
with use_scope("main_view", clear=True):
put_text("wake word database deleted!")
datasets_menu(back_handler=back_handler)
return
if opt == "main":
with use_scope("main_view", clear=True):
datasets_menu(back_handler=back_handler)
return
ww_menu(back_handler=back_handler)
def utt_menu(back_handler=None):
with use_scope("logo", clear=True):
img = open(f'{os.path.dirname(__file__)}/res/utterances.png', 'rb').read()
put_image(img)
buttons = [{'label': 'Inspect Devices', 'value': "dev"},
{'label': 'Inspect Recordings', 'value': "utt"},
{'label': 'Upload Utterance', 'value': "upload"},
{'label': 'Delete utterances database', 'value': "delete_utt"}]
if back_handler:
buttons.insert(0, {'label': '<- Go Back', 'value': "main"})
opt = actions(label="What would you like to do?",
buttons=buttons)
if opt == "dev":
device_select(back_handler=back_handler, ww=False)
if opt == "utt":
utt_opts(back_handler=back_handler)
if opt == "upload":
with use_scope("main_view", clear=True):
data = input_group("Upload utterance", [
textarea("transcript", placeholder="hello world", required=True, name="utterance"),
file_upload("speech recording", name="file")
])
utterance = data["utterance"]
filename = data["file"]["filename"]
mime = data["file"]["mime_type"]
content = data["file"]["content"]
if mime != "audio/x-wav":
popup("invalid format!")
# if mime in ["application/json"]:
else:
os.makedirs(f"{CONFIGURATION['data_path']}/utterances", exist_ok=True)
uuid = "AnonDevice" # TODO - allow tagging to a device
meta = {
"transcription": utterance,
"uuid": uuid
}
DB.add_stt_recording(content, utterance, meta)
with popup("utterance recording uploaded!"):
put_code(json.dumps(meta, indent=4), "json")
if opt == "delete_utt":
with use_scope("main_view", clear=True):
put_markdown("""Are you sure you want to delete the utterances database?
**this can not be undone**, proceed with caution!
**ALL** utterance recordings will be **lost**""")
opt = actions(label="Delete utterances database?",
buttons=[{'label': "yes", 'value': True},
{'label': "no", 'value': False}])
if opt:
for rec in DB.list_stt_recordings():
DB.delete_stt_recording(rec["recording_id"])
with use_scope("main_view", clear=True):
put_text("utterance database deleted!")
datasets_menu(back_handler=back_handler)
return
if opt == "main":
with use_scope("main_view", clear=True):
datasets_menu(back_handler=back_handler)
return
utt_menu(back_handler=back_handler)
def datasets_menu(back_handler=None):
with use_scope("logo", clear=True):
img = open(f'{os.path.dirname(__file__)}/res/open_dataset.png', 'rb').read()
put_image(img)
buttons = [
{'label': 'Tag Wake Words', 'value': "dataset"},
{'label': 'Manage Wake Words', 'value': "ww"},
{'label': 'Manage Utterance Recordings', 'value': "utt"},
]
if back_handler:
buttons.insert(0, {'label': '<- Go Back', 'value': "main"})
opt = actions(label="What would you like to do?",
buttons=buttons)
if opt == "dataset":
ww_tagger(back_handler=back_handler)
elif opt == "utt":
utt_menu(back_handler=back_handler)
elif opt == "ww":
ww_menu(back_handler=back_handler)
elif opt == "main":
with use_scope("main_view", clear=True):
if back_handler:
back_handler()
return
datasets_menu(back_handler=back_handler)