Files
dikte/dikte/worker.py
T
GökhanandClaude Opus 5 45b18722be Keep the front where the dictation started on macOS
Recording goes through ffmpeg's avfoundation input, and opening a capture
session there brings the process that did it to the front. ffmpeg is a
child of Dikte with no bundle of its own, so macOS credits the move to
Dikte: the window the user was dictating into goes inactive, its caret
stops, its title bar greys out, and the Cmd+V at the end of the run lands
somewhere other than the document it was meant for. Measured with a
TextEdit document in front:

    press the shortcut          front = TextEdit
    recorder.start returns      front = TextEdit
    89 ms later                 front = Dikte

Nothing about the capture session can be asked not to do this. Starting
ffmpeg in its own session, and clearing __CFBundleIdentifier from its
environment, were both tried and both measured to make no difference, so
it is undone instead: the application in front is noted before the
indicator goes up, and a short watch puts it back the moment Dikte takes
the front. Measured at 99 ms from the moment it is taken, against the
title bar staying grey for the whole dictation before. All three ways in
do it, dictation, agent and meeting, since all three open the same
capture.

The indicator had a share of the same problem and needed AppKit for it
too, so mac_window.py carries both. An NSPanel is hidden by the system
the moment its application stops being the active one, which for a
dictation indicator is immediately, and ordering one to the front brings
its application with it unless it carries the nonactivating style bit.
Neither is reachable through Qt.

The runtime is loaded in _appkit() rather than at import, the way paste.py
loads its frameworks in _macos_api(), so the module imports on a machine
with no AppKit and the tests below run there as well: 52 new tests, none
of them skipped anywhere.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
2026-08-16 17:37:31 +03:00

205 lines
7.8 KiB
Python

"""The dictation chain: transcribe → clean up → clipboard → paste.
The same chain also carries the other thing a dictation can be. Asked to, it
hands the transcript to Claude Code instead of pasting it, and pastes back
whatever came of it: an answer to a question, or a sentence saying what was
done.
"""
import os
import shutil
import sys
import threading
import time
import traceback
from PyQt6.QtCore import QObject, pyqtSignal
from . import api
from . import assistant
from . import audio
from . import cleanup
from . import config as cfg
from . import i18n
from . import paste
from . import vad
from .i18n import t
CHUNK_SECONDS = audio.CHUNK_FRAMES / audio.RATE
# A dictation and a command to the agent run side by side and can finish at the
# same moment. Pasting is not one step but three that must not interleave: read
# what is on the clipboard, put ours there, press the key. Two runs doing that
# at once would paste one answer and restore the other's clipboard over it.
_paste_lock = threading.Lock()
class Pipeline(QObject):
stage = pyqtSignal(str) # human-readable progress line
finished = pyqtSignal(str, str, str) # raw transcript, final text, warning
failed = pyqtSignal(str)
cancelled = pyqtSignal()
def __init__(self, conf, parent=None):
super().__init__(parent)
self.conf = conf
self._thread = None
self._stop = threading.Event()
@property
def busy(self):
return self._thread is not None and self._thread.is_alive()
def run(self, wav_path, duration, rms_values=(), ask=False, paste=None,
focus=None):
"""`paste` overrides the setting for this one run, which is what a
dictation asked for from a terminal wants: the text comes back down the
socket, and pasting it into whatever had focus is nobody's intention.
`focus` is the application that was in front when the recording began,
as a process id, and is where the paste is meant to land."""
if self.busy:
return
self._stop.clear()
self._thread = threading.Thread(
target=self._work,
args=(wav_path, duration, list(rms_values), ask, paste, focus),
daemon=True,
)
self._thread.start()
def cancel(self):
"""Give up on a job already under way.
Only the Claude call can honour this, and it is the only one long enough
to be worth interrupting: a transcription is over in seconds, a command
that went looking through the web is not.
"""
self._stop.set()
def _work(self, wav_path, duration, rms_values, ask, paste_override=None,
focus=None):
conf = self.conf
started = time.monotonic()
raw = ""
# Room tone only: don't spend an API call, and don't invite a
# hallucinated sentence back.
if conf["skip_silent"]:
stats = vad.analyse(rms_values, CHUNK_SECONDS, conf["speech_margin_db"])
if vad.is_silent(stats, conf["silence_db"], conf["speech_margin_db"],
conf["min_voiced_seconds"]):
self._discard(wav_path)
self.failed.emit(
t("No speech detected ({level} dB)", level=round(stats["speech_db"]))
)
return
try:
self.stage.emit(t("Transcribing…"))
target = conf.transcribe_target()
raw = api.transcribe(
target,
wav_path,
language=conf["language"],
prompt=conf["transcribe_prompt"],
)
if conf["filter_hallucinations"] and vad.looks_like_hallucination(raw, duration):
self._discard(wav_path)
self.failed.emit(t("Discarded a stock phrase: “{text}”", text=raw[:60]))
return
text = raw
warning = ""
# Claude reads through “eee” and “hani” without help, so a dictation
# on its way there is normally sent as it was heard, one API call and
# a second or two lighter.
if (conf["assistant_cleanup"] if ask else conf["cleanup_enabled"]):
self.stage.emit(t("Cleaning up…"))
try:
text = cleanup.run(raw, conf, conf.cleanup_prompt())
except api.ApiError as exc:
# Keep the transcript, but never let the failure pass unseen:
# a rejected key would otherwise look like working dictation.
text = raw
warning = str(exc)
print(f"dikte: cleanup failed: {exc}", file=sys.stderr)
question = ""
if ask:
question = text
self.stage.emit(t("Asking {name}…", name=i18n.name(
assistant.display_name(conf), "dative")))
text, denied = assistant.ask(
question, conf,
on_stage=self.stage.emit,
should_stop=self._stop.is_set,
)
warning = "\n".join(x for x in (warning, denied) if x)
wants_paste = (conf["assistant_paste"] if ask else conf["auto_paste"])
if paste_override is not None:
wants_paste = paste_override
with _paste_lock:
previous = (paste.read_clipboard()
if conf["restore_clipboard"] and wants_paste else None)
try:
paste.copy(text)
if wants_paste:
self.stage.emit(t("Pasting…"))
paste.press(conf["paste_shortcut"], focus=focus)
finally:
if previous is not None:
# Let the focused application consume the temporary
# transcription before putting every old clipboard type
# back. This also runs when key injection fails.
time.sleep(0.35)
paste.copy_bytes(previous)
cfg.append_history({
"ts": time.strftime("%Y-%m-%d %H:%M:%S"),
"duration": round(duration, 1),
"elapsed": round(time.monotonic() - started, 1),
"model": target.model,
"cleanup_model": cleanup.model(conf) if conf["cleanup_enabled"] else "",
"cleanup_error": warning,
"mode": "ask" if ask else "",
"question": question,
"assistant_model": conf["assistant_model"] if ask else "",
"raw": raw,
"text": text,
})
try:
cfg.trim_history(conf["history_limit"])
except OSError as exc:
print(f"dikte: could not trim the history: {exc}", file=sys.stderr)
self.finished.emit(raw, text, warning)
except assistant.Cancelled:
self.cancelled.emit()
except (api.ApiError, paste.PasteError, assistant.AssistantError) as exc:
print(f"dikte: {exc}", file=sys.stderr)
self.failed.emit(str(exc))
except Exception as exc: # never fail silently
traceback.print_exc()
self.failed.emit(t("Unexpected error: {error}", error=exc))
finally:
self._discard(wav_path)
def _discard(self, wav_path):
if not os.path.exists(wav_path):
return
if self.conf["keep_audio"]:
try:
cfg.RECORDINGS_DIR.mkdir(parents=True, exist_ok=True)
shutil.move(wav_path, cfg.RECORDINGS_DIR / (time.strftime("%Y%m%d-%H%M%S") + ".wav"))
return
except OSError:
pass
try:
os.unlink(wav_path)
except OSError:
pass