Streaming replies and STT, amplitude lip-sync, one place for speaking

Latency: replies are spoken sentence-by-sentence off the desk API's NDJSON
endpoint, so the wait is time-to-first-sentence rather than the whole model
call, and Deepgram's live websocket transcribes while you're still talking
instead of uploading the WAV afterwards. Both fall back invisibly — a stream
that fails before anything was said drops to converse(), and a socket that
never opens just means the old one-shot path.

Speaking lived in four near-copies in the controller (a reply, a holding line,
a streamed sentence, a dialogue scene) that had already drifted: one didn't arm
barge-in, another skipped the follow-up rule. It's now speech.Speaker plus an
Utterance describing the policy differences, with collaborators injected so the
whole of it tests without Qt or audio.

The mouth follows the audio rather than a timer: tts.level_of reduces each PCM
frame to a 0..1 loudness on a sqrt curve (speech sits well below peak, and a
linear map leaves the mouth barely open during normal talking) and that indexes
the talking frames, which the sprite script now draws as an openness ramp.
Offline pyttsx3 has no waveform, so stale levels hand control back to the timed
loop instead of freezing the mouth mid-syllable.

Also: the pet starts where you left it (ignoring positions on monitors that are
no longer connected, since restoring those faithfully is how it ends up
somewhere unreachable), and `python -m bolt_pet --doctor` is a preflight that
says what to do about each problem rather than only what's wrong.

tests/test_pipeline_smoke.py breaks the pure-logic rule on purpose. Every unit
test passed all week while notifications sat unspoken for minutes, the pet said
things twice and [laughing] got read aloud — each an interaction between two
individually-correct units. It drives whole turns against a real HTTP server on
a loopback port, faking only the mic and the speakers. It found a NameError in
the paint path that would have fired on every repaint while talking.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

.env.example

@@ -103,6 +103,9 @@ ELEVENLABS_VOICE_ID=
 #PET_CLICK_THROUGH=false
 #PET_EDGE_SNAP=true
 #PET_SNAP_MARGIN=48
+# Start where you last dragged it. A position on a monitor that's no longer
+# connected is ignored, so unplugging a screen can't hide the pet off-desktop.
+#PET_REMEMBER_POSITION=true
 
 # ── Barge-in (optional) — interrupt the pet mid-sentence ────────────────────
 # BARGE_IN_MODE decides what counts as an interruption:
@@ -138,6 +141,16 @@ ELEVENLABS_VOICE_ID=
 # ── Streaming TTS (optional) — starts talking on the first chunk ────────────
 #TTS_STREAMING=true
 
+# ── Latency: streaming the reply and the transcript ─────────────────────────
+# STREAMING_REPLIES speaks each sentence as the server generates it, instead of
+# waiting out the whole model call before the first word. STT_STREAMING sends
+# mic frames to Deepgram as you talk, so the transcript is ready the moment you
+# stop. Both fall back to the old path automatically if anything goes wrong.
+# VAD_SILENCE_END_SEC is the other half: it is dead air on every single turn,
+# so 0.8-1.0 feels markedly snappier than the 1.2 default.
+#STREAMING_REPLIES=true
+#STT_STREAMING=true
+
 # ── Screen context (optional) ───────────────────────────────────────────────
 # Sends the focused window's title along with what you said, so "what's this
 # error?" has a referent. Text only — no screenshots leave the machine.

CLAUDE.md

@@ -35,6 +35,10 @@ all suppressed while it's napping (quiet hours / fullscreen DND).
 ./run.sh          # macOS/Linux
 run.bat           # Windows
 
+# Preflight: is this install actually going to work? (config, mic, keys, sprites…)
+python -m bolt_pet --doctor          # shallow: no network, no mic
+python -m bolt_pet --doctor --deep   # contacts the server and opens the microphone
+
 # Run tests (no pytest config file — tests self-insert repo root via sys.path).
 # QT_QPA_PLATFORM=offscreen avoids a QApplication segfault on headless/no-display hosts.
 QT_QPA_PLATFORM=offscreen .venv/bin/pytest tests/
@@ -239,7 +243,18 @@ logs a missing-config message and exits its thread instead of starting.
 - **`quiet.py`** — quiet-hours spec parsing (`23:00-08:00`, wraps midnight,
   comma-separated). Napping suppresses *proactive* noise and wandering only;
   wake word / click / push-to-talk still work.
-- **`notifications.py`** — Linux/D-Bus notification bridge: tails
+- **`notifications.py`** — Linux/D-Bus notification bridge. Two latency/loss bugs
+  fixed 2026-08-02, both in `controller._maybe_heartbeat`: draining was wired to the
+  60s heartbeat interval rather than the ~1.2s wake tick, and a heartbeat that landed
+  mid-conversation stamped its own clock *before* checking — burning the slot and
+  waiting another full interval, repeatedly, which is how a notification could go
+  unspoken for five or ten minutes. Draining now runs on every tick while IDLE, and the
+  heartbeat clock only advances when the heartbeat actually runs. Separately, the rate
+  limit used to *drop* notifications inside its window (a second text a minute later was
+  silently lost); the filter still gates at queue time but the limit is gone — a burst
+  is **batched into one turn** instead, same single round trip, no lost messages, capped
+  by `_MAX_PENDING_NOTIFICATIONS`.
+- **`notifications.py` internals** — tails
   `dbus-monitor`, parses Notify calls (pure `iter_notifications()`), filters
   and rate-limits them (`NotificationGate`), and the controller forwards
   survivors through `converse()`. Off by default — each one is a round trip.
@@ -275,6 +290,81 @@ logs a missing-config message and exits its thread instead of starting.
 - **`hotkey.py`** — global push-to-talk via `pynput`; soft-fails with a logged
   reason (Wayland, missing package, macOS permissions) since the wake word is
   the primary trigger.
+- **`audio/stt_stream.py`** — streaming speech-to-text. The one-shot path waits for
+  the utterance to end, uploads the whole WAV, then waits again; that second wait is
+  dead time that grows with how long you spoke. Deepgram's live websocket removes it:
+  `record_utterance(on_frame=...)` hands each captured frame to a
+  `StreamingTranscriber`, so by the time the VAD decides you stopped the transcript is
+  essentially already there. Three deliberate limits: `open()` returning **None is an
+  ordinary outcome** (no websocket-client, no network, no key) because the full audio is
+  still buffered and `controller._transcribe` just falls back; the **local VAD still
+  decides when you stopped** rather than Deepgram's endpointing, since barge-in,
+  follow-up listening and the grace period are all built on it and coupling them to the
+  network is not a first-pass change; and the socket is **per-utterance**, because
+  holding one open across an idle pet bills for silence and dies on the first blip.
+  Off: `STT_STREAMING=false`.
+- **Streamed replies** — `server_client.converse_stream()` reads NDJSON from the desk
+  API's `/desk/converse_stream` and speaks each sentence as it arrives
+  (`controller._speak_stream_chunk`), so the wait is time-to-first-sentence instead of
+  the whole model call. `Reply.spoken` marks a reply whose sentences were already said:
+  the text is still carried, because the follow-up rule needs to see whether it ended on
+  a question, but it must not be read out again. A stream that fails *before* anything
+  was spoken falls back to `converse()` invisibly; one that fails after ends the turn
+  quietly rather than repeating the first half. Off: `STREAMING_REPLIES=false`.
+- **`speech.py`** — everything the pet says, and the policy differences between
+  kinds of saying. There were four near-copies of this in `controller.py` (a
+  reply, a holding line, a streamed sentence, a dialogue scene), each repeating
+  the same dance — transition state, show the bubble, maybe record history,
+  reset barge-in, call TTS, reset barge-in *again*, resume state, decide
+  whether to keep the mic open — and they had already drifted apart: one forgot
+  to arm barge-in, another skipped the follow-up rule. Now the dance is
+  `Speaker.say()` and the differences are data on a frozen `Utterance`
+  (`record`, `resume`, `hold_talking`, `follow_up`, `interruptible`), built by
+  the four classmethods `reply`/`holding`/`stream_chunk`/`scene`. Notable
+  policies: a holding line is **not** recorded (it's filler; the transcript
+  should keep the answer) and **not** interruptible (cutting off "give me a
+  sec" strands the tool already running), and it resumes the state it
+  interrupted rather than dropping to IDLE, because the turn isn't over. A
+  streamed chunk stays TALKING so the sprite doesn't flicker between sentences.
+  Collaborators are injected, so all of this is tested without Qt, audio or a
+  real state machine. Two subtleties that are bugs waiting to happen: the
+  barge-in detector is read through a **callable**, not held (it's built after
+  the Speaker — it needs the mic stream — and swapped when the mode changes; two
+  copies drifting apart is invisible until the wake model starts hearing the
+  pet), and `last_detail` is captured *before* the post-playback reset, since
+  reading it after means every interruption reports zeroed counters.
+  `follow_up_decision()` is the mic-open rule as a pure function.
+- **Lip-sync** — the mouth is driven by the audio, not a timer.
+  `audio/tts.level_of(frame)` reduces a PCM frame to 0..1 loudness on a **sqrt
+  curve** (speech sits well below peak most of the time, so a linear map leaves
+  the mouth barely open during normal talking), `envelope()` does the same for a
+  whole clip, and playback calls the `on_level` hook per chunk. That travels
+  `Speaker` → `controller.mouth` (a Signal) → `PetWindow.set_mouth`, and
+  `_mouth_frame()` indexes the talking frames directly — which works because
+  `generate_bolt_sprites.py` draws them as an **openness ramp** (closed first,
+  widest last) rather than an arbitrary loop. Levels going stale
+  (`_MOUTH_STALE_SECONDS`) hands control back to the ordinary animation, so
+  offline TTS — which has no envelope — degrades to the timed loop instead of
+  freezing the mouth mid-syllable.
+- **`window_state.py`** — where the pet was left, so it starts there.
+  `~/.cache/bolt-pet/window.json`, atomic write on drag-end, every function
+  swallows its own errors (a corrupt state file must mean the default corner,
+  never a pet that won't start). `is_visible_on()` re-validates against the
+  *current* screen layout on load, because the common case for a stale position
+  is exactly the dangerous one: the pet was last on a monitor that is now
+  unplugged, and restoring it faithfully puts it somewhere unreachable. Takes
+  plain rectangles rather than importing Qt. Off: `PET_REMEMBER_POSITION=false`.
+- **`doctor.py`** — `python -m bolt_pet --doctor`, a preflight, written after a
+  week of debugging things one command would have shown: a venv whose python
+  was a zero-byte file, an OCR engine never installed, a wrong proxy header.
+  Twelve independent checks, each reporting ok / warn (degraded but working) /
+  fail, and each saying **what to do about it** — `screen reading: warn` is
+  useless alone, `apt install tesseract-ocr` is the whole point. Nothing raises:
+  a doctor that crashes on a broken install is diagnosing the wrong patient, so
+  `run()` catches per-check and a failed check becomes a FAIL row rather than a
+  traceback. Shallow by default (no network, no mic) since it's the first thing
+  you reach for when the network is what's broken; `--deep` actually contacts
+  the server and opens the microphone.
 - **`speech_text.py`** — sanitizes server replies before they're heard/shown.
   `for_speech()` (called inside `tts.speak()`, so every path to the speakers is
   covered) strips markdown, emoji, URLs and stray symbols the voice would read
@@ -284,7 +374,7 @@ logs a missing-config message and exits its thread instead of starting.
   whether a reply leaves the pet waiting on an answer: it tests the *spoken*
   form (so a '?' inside a stripped code block or URL doesn't count) and only a
   trailing one counts, since a question asked in passing isn't awaiting a
-  reply. `controller._should_follow_up` uses it to keep listening without the
+  reply. `speech.follow_up_decision` uses it to keep listening without the
   wake word, capped by `FOLLOW_UP_MAX_TURNS` so a server that ends every reply
   with a question can't loop forever off mic noise. Pure string logic, no
   Qt/audio imports.
@@ -414,6 +504,20 @@ HTTP chunk reassembly by `chunks_to_int16`, emote motion by
 `screen_context.context_for` / `is_fullscreen_active`, otherwise they shell
 out to xprop on a headless box.
 
+**`test_pipeline_smoke.py` is the exception, deliberately.** Unit tests inject
+a fake at the seam they care about, and a run of production bugs — notifications
+sitting unspoken for minutes, the pet saying things twice, `[laughing]` read
+aloud, a device command arriving as prose — got through with every one of them
+green, because each was an *interaction* between two individually-correct
+units. So that file stands up a real threaded HTTP server on a loopback port,
+speaks the desk protocol at it, and drives whole turns through the real
+`server_client` (including NDJSON streaming), the real controller and the real
+state machine, faking only the mic stream and the speakers. When a bug crosses
+a module boundary, add the case there; when it lives inside one module, the
+pure-function pattern above is still the cheaper test. It is also worth
+mutation-checking a new case — break the source line it is meant to catch and
+confirm it actually goes red.
+
 ## Security notes
 
 The server can relay a shell command back to this machine to execute as the

README.md

@@ -140,6 +140,33 @@ near misses — frames that scored just under the threshold — and the slider
 takes effect immediately, mid-listen. Set the threshold just below the peak
 you can hit reliably, then write it into `.env` as `WAKE_WORD_THRESHOLD`.
 
+## Something not working?
+
+```bash
+python -m bolt_pet --doctor
+```
+
+Checks the things that make the pet look broken in ways that don't point at
+themselves — missing server config (the controller exits its thread at startup,
+so the pet appears alive and simply never answers), no input device, no OCR
+engine behind `petctl read`, a wake model that isn't where `.env` says, a
+silence timeout long enough to feel like lag. Each line says what to do about
+it, not just what's wrong. It doesn't touch the network or open the microphone
+unless you add `--deep`, so it's safe to run when the network is the suspect.
+
+## Little things
+
+The pet **remembers where you left it** — drag it somewhere deliberate and
+that's where it starts next time. If that position is on a monitor you've since
+unplugged it goes back to the default corner rather than restoring itself
+somewhere off-screen. `PET_REMEMBER_POSITION=false` to always start in the
+corner.
+
+Its **mouth moves with the actual audio** rather than flapping on a timer: the
+PCM going to the speakers is reduced to a loudness per frame and that picks the
+talking sprite, so the pet shuts up when the voice pauses. Offline `pyttsx3`
+playback has no waveform to follow, so it falls back to the timed loop.
+
 ## Project layout
 
 ```
@@ -148,18 +175,23 @@ bolt_pet/
   state.py            PetState enum + a small transition-checked state machine
   server_client.py    /desk/converse, /desk/tool_result, /desk/report_status
   controller.py        the pipeline: wake word -> STT -> server -> TTS, on a QThread
-  speech_text.py        strips markdown/emoji/URLs so the voice never says "asterisk"
+  speech.py             what the pet says and how each kind of saying behaves
+  speech_text.py         strips markdown/emoji/URLs so the voice never says "asterisk"
   pet_actions.py         petctl move/emote/say/wander/nap parsing
   screen_context.py       active-window title + fullscreen detection
   quiet.py                 quiet-hours schedule
   notifications.py          desktop notification bridge (Linux/D-Bus)
   history.py                 rolling conversation transcript
   hotkey.py                   global push-to-talk (pynput, optional)
+  window_state.py              remembers where you left the pet
+  doctor.py                     `--doctor` preflight: is this install going to work?
   audio/
     mic.py             input stream + energy-based VAD utterance capture
     wake_word.py        openWakeWord thunderbolt.onnx detection (see above)
-    stt.py               Deepgram
+    stt.py               Deepgram (one-shot)
-    tts.py                ElevenLabs streaming PCM, offline pyttsx3 fallback
+    stt_stream.py         Deepgram live websocket — transcribes while you speak
+    tts.py                 ElevenLabs streaming PCM, offline pyttsx3 fallback,
+                            plus the loudness envelope that drives the mouth
     barge_in.py            "you started talking" detector, to cut playback short
   ui/
     app.py              wires QApplication + window + tray + controller thread together
@@ -172,8 +204,10 @@ bolt_pet/
 scripts/
   slice_spritesheet.py  cuts a grid sprite sheet into the per-frame convention
 tests/                  pure-logic unit tests (state machine, wake-phrase
-                         matching, HTTP client against mocks) — nothing here
+                         matching, HTTP client against mocks) plus one
-                         needs real audio hardware or a display
+                         end-to-end smoke test that drives whole turns against
+                         a real local HTTP server — nothing here needs real
+                         audio hardware or a display
 ```
 
 ## Security notes
@@ -197,7 +231,6 @@ notifications. No screenshots or images are ever sent.
 
 ## Known limitations / not-yet-done
 
-- Pet screen position isn't persisted across restarts.
 - Wandering is a straight walk to a random point — no Shimeji-style physics,
   wall-climbing or falling.
 - Push-to-talk and the notification bridge are platform-limited: the hotkey

bolt_pet/__main__.py

@@ -1,8 +1,15 @@
-"""Entry point: python -m bolt_pet"""
+"""Entry point: python -m bolt_pet [--doctor [--deep]]"""
 
 import sys
 
-from .ui.app import run
-
 if __name__ == "__main__":
+    if "--doctor" in sys.argv:
+        # Imported inside the branch, not at module scope: the doctor exists to
+        # diagnose installs where importing the UI would itself blow up.
+        from .doctor import main as doctor
+
+        sys.exit(doctor(sys.argv[1:]))
+
+    from .ui.app import run
+
     sys.exit(run())

bolt_pet/audio/mic.py

@@ -37,6 +37,20 @@ def rms(frame: np.ndarray) -> float:
     return float(np.sqrt(np.mean(frame.astype(np.float64) ** 2)))
 
 
+def _emit(on_frame, frame) -> None:
+    """Hand a frame to a listener without letting it break the capture.
+
+    Streaming STT is an optimisation riding along with recording; if its
+    socket dies mid-utterance the recording must carry on untouched, because
+    the one-shot fallback is about to need the full buffer."""
+    if on_frame is None:
+        return
+    try:
+        on_frame(frame)
+    except Exception:
+        pass
+
+
 def record_utterance(
     stream: AudioStream,
     should_continue=lambda: True,
@@ -47,6 +61,7 @@ def record_utterance(
     grace_s: float = None,
     frame_len: int = config.FRAME_LEN,
     sample_rate: int = config.SAMPLE_RATE,
+    on_frame=None,
 ) -> Optional[np.ndarray]:
     """Capture one utterance from *stream*: wait for speech to start, stop
     after trailing silence. Returns None if nothing usable was heard.
@@ -55,6 +70,11 @@ def record_utterance(
     (e.g. the pet window was closed) without needing threading primitives
     baked into this function.
 
+    *on_frame*, if given, is called with each captured frame while speech is
+    in progress — that is how streaming STT transcribes as you talk rather
+    than after (see audio/stt_stream.py). It is fire-and-forget: this function
+    still returns the full buffer, so a failed stream costs nothing.
+
     *grace_s* is how long to wait for speech to *begin* before giving up.
     The controller stretches it for follow-up questions, where you're being
     asked something and need a moment to think rather than having just said
@@ -83,10 +103,12 @@ def record_utterance(
             if frame_rms >= rms_threshold:
                 started = True
                 frames.append(frame)
+                _emit(on_frame, frame)
             elif waited > grace_frames:
                 return None  # woke it up but said nothing
             continue
         frames.append(frame)
+        _emit(on_frame, frame)
         if frame_rms < rms_threshold:
             silence_frames += 1
             if silence_frames >= silence_limit:

bolt_pet/audio/stt_stream.py

@@ -0,0 +1,181 @@
+"""Streaming speech-to-text — transcribing *while* you talk, not after.
+
+The one-shot path (`stt.transcribe`) waits for the utterance to finish, then
+uploads the whole WAV and waits again. That second wait is dead time between
+you stopping and the pet reacting, and it grows with the length of what you
+said — a thirty-second question costs noticeably more than a five-second one.
+
+Deepgram's live endpoint removes it: frames go up as they are captured, so by
+the time the VAD decides you have stopped, the transcript is essentially
+already there. Same model, same account, same accuracy — the difference is
+purely when the work happens.
+
+Design constraints that shaped this:
+
+- **Failure must be invisible.** No websocket, no network, a mid-utterance
+  disconnect — all of it falls back to the one-shot path, which still has the
+  full audio buffered. Streaming is an optimisation, never a dependency, so
+  `open()` returning None is an ordinary outcome rather than an error.
+- **The VAD still decides when you stopped.** Deepgram has its own endpointing
+  and using it would save more, but it would also move a decision the rest of
+  the pipeline is built around (barge-in, follow-up listening, the grace
+  period) into a remote service. Not worth coupling those to the network on
+  the first pass.
+- **The socket is per-utterance.** Holding one open across an idle pet would
+  bill for silence and drop on the first network blip; opening one takes
+  ~100ms, which is already inside the time it takes a person to start talking.
+
+The websocket client is injectable, so the whole protocol — send frames, read
+`is_final` transcripts, close, take the result — is tested without a network.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import threading
+from typing import Callable, Optional
+
+import numpy as np
+
+from .. import config
+
+logger = logging.getLogger("bolt_pet.stt_stream")
+
+_ENDPOINT = (
+    "wss://api.deepgram.com/v1/listen"
+    "?encoding=linear16&channels=1&sample_rate={rate}&model={model}"
+    "&language=en&smart_format=true&interim_results=false"
+)
+
+
+def available() -> bool:
+    """Whether streaming STT can even be attempted in this install."""
+    if not config.STT_STREAMING or not config.DEEPGRAM_API_KEY:
+        return False
+    try:
+        import websocket  # noqa: F401  (websocket-client)
+        return True
+    except Exception:
+        return False
+
+
+class StreamingTranscriber:
+    """One utterance's worth of live transcription.
+
+    Usage mirrors how the capture loop already works — feed frames as they
+    arrive, then ask what was said:
+
+        session = StreamingTranscriber.open()
+        ...
+        session.feed(frame)      # per mic frame, non-blocking
+        text = session.finish()  # after the VAD says you stopped
+    """
+
+    def __init__(self, socket, *, sample_rate: int = None):
+        self._socket = socket
+        self._sample_rate = sample_rate or config.SAMPLE_RATE
+        self._transcript: list[str] = []
+        self._lock = threading.Lock()
+        self._closed = False
+        self._reader = threading.Thread(
+            target=self._read_loop, name="stt-stream-reader", daemon=True)
+        self._reader.start()
+
+    # -- lifecycle ----------------------------------------------------------
+
+    @classmethod
+    def open(cls, *, connect: Optional[Callable] = None,
+             sample_rate: int = None) -> Optional["StreamingTranscriber"]:
+        """Connect, or return None if streaming isn't possible right now.
+
+        None is a normal outcome, not a failure: the caller keeps the audio and
+        falls back to the one-shot upload."""
+        if connect is None and not available():
+            return None
+        rate = sample_rate or config.SAMPLE_RATE
+        try:
+            if connect is not None:
+                socket = connect()
+            else:
+                import websocket
+
+                socket = websocket.create_connection(
+                    _ENDPOINT.format(rate=rate, model=config.DEEPGRAM_MODEL),
+                    header={"Authorization": f"Token {config.DEEPGRAM_API_KEY}"},
+                    timeout=10,
+                )
+            return cls(socket, sample_rate=rate)
+        except Exception as exc:
+            logger.info("Streaming STT unavailable (%s) — using the one-shot path.", exc)
+            return None
+
+    def feed(self, frame: np.ndarray) -> None:
+        """Send one captured frame. Never raises — a dead socket just means the
+        fallback will do the work."""
+        if self._closed:
+            return
+        try:
+            self._socket.send_binary(np.asarray(frame, dtype=np.int16).tobytes())
+        except Exception:
+            logger.debug("Streaming STT send failed; abandoning the stream", exc_info=True)
+            self._closed = True
+
+    def finish(self, timeout: float = 3.0) -> str:
+        """Close the stream and return whatever was transcribed.
+
+        Deepgram flushes its final results after the close frame, so this waits
+        briefly for the reader — bounded, because a hung socket must not hold
+        up the reply."""
+        if not self._closed:
+            try:
+                self._socket.send(json.dumps({"type": "CloseStream"}))
+            except Exception:
+                pass
+        self._closed = True
+        self._reader.join(timeout=timeout)
+        try:
+            self._socket.close()
+        except Exception:
+            pass
+        with self._lock:
+            return " ".join(part for part in self._transcript if part).strip()
+
+    # -- the reader ---------------------------------------------------------
+
+    def _read_loop(self) -> None:
+        while not self._closed:
+            try:
+                message = self._socket.recv()
+            except Exception:
+                break
+            if not message:
+                break
+            text, is_final = self._parse(message)
+            if text and is_final:
+                with self._lock:
+                    self._transcript.append(text)
+
+    @staticmethod
+    def _parse(message) -> tuple[str, bool]:
+        """Pull (text, is_final) out of a Deepgram results frame.
+
+        Tolerant on purpose: anything unrecognised is ignored rather than
+        raising on the reader thread, where an exception would silently kill
+        transcription for the rest of the utterance."""
+        try:
+            if isinstance(message, bytes):
+                message = message.decode("utf-8", "ignore")
+            data = json.loads(message)
+        except (TypeError, ValueError):
+            return "", False
+        if not isinstance(data, dict):
+            return "", False
+        alternatives = (
+            ((data.get("channel") or {}).get("alternatives") or [])
+            if data.get("type") in (None, "Results") else []
+        )
+        if not alternatives:
+            return "", False
+        text = str((alternatives[0] or {}).get("transcript") or "").strip()
+        return text, bool(data.get("is_final") or data.get("speech_final"))

bolt_pet/audio/tts.py

@@ -14,6 +14,7 @@ fallback has no such concept and always sounds like itself.
 
 from __future__ import annotations
 
+import time
 from typing import Iterable, Iterator, Optional
 
 import numpy as np
@@ -162,32 +163,81 @@ def synthesize_dialogue(
     return pcm, config.TTS_SAMPLE_RATE
 
 
-def play_pcm(pcm: np.ndarray, sample_rate: int, blocking: bool = True, should_stop=None) -> bool:
+# ── how loud is it right now ────────────────────────────────────────────────
+# The PCM is already decoded here on its way to the speakers, so the amplitude
+# envelope is free — and it is exactly what a mouth needs to move in time with
+# speech. Throwing it away and animating the mouth on a timer instead is why
+# most talking sprites look dubbed.
+
+# int16 RMS that counts as "mouth fully open". Speech peaks around 8-12k;
+# 6000 keeps normal talking in the upper half of the range without clipping
+# every syllable to wide-open.
+_LOUD_RMS = 6000.0
+
+
+def level_of(frame: np.ndarray) -> float:
+    """0..1 loudness for one chunk of PCM.
+
+    Square-rooted because perceived loudness is not linear in amplitude — a
+    linear mapping leaves the mouth barely moving through ordinary speech."""
+    if frame is None or len(frame) == 0:
+        return 0.0
+    rms = float(np.sqrt(np.mean(np.square(frame.astype(np.float32)))))
+    return float(min(1.0, (rms / _LOUD_RMS) ** 0.5))
+
+
+def envelope(pcm: np.ndarray, sample_rate: int, fps: int = 30) -> list:
+    """Per-frame loudness for a whole clip, for playback that isn't streamed."""
+    if pcm is None or len(pcm) == 0:
+        return []
+    window = max(1, int(sample_rate / max(1, fps)))
+    return [level_of(pcm[start:start + window]) for start in range(0, len(pcm), window)]
+
+
+def play_pcm(pcm: np.ndarray, sample_rate: int, blocking: bool = True, should_stop=None,
+             on_level=None) -> bool:
     """Play a whole clip. Returns True if it finished, False if *should_stop*
     (barge-in) cut it short. *should_stop* is polled while audio plays — each
     poll consumes one mic frame, which is what paces this loop."""
     import sounddevice as sd
 
+    levels = envelope(pcm, sample_rate) if on_level is not None else []
+    started = time.monotonic()
     sd.play(pcm, samplerate=sample_rate, device=config.SPEAKER_DEVICE)
     if not blocking:
         return True
-    if should_stop is None:
+    if should_stop is None and on_level is None:
         sd.wait()
         return True
     while True:
+        if levels:
+            # Indexed by elapsed time rather than by chunk, because this path
+            # hands the whole clip to the device at once and never sees it
+            # again — wall clock is the only position we have.
+            index = int((time.monotonic() - started) * 30)
+            if index < len(levels):
+                try:
+                    on_level(levels[index])
+                except Exception:
+                    levels = []
         try:
             if not sd.get_stream().active:
                 break
         except Exception:
             break  # stream already torn down — playback is over
-        if should_stop():
+        if should_stop is not None and should_stop():
             sd.stop()
             return False
     return True
 
 
-def play_stream(chunks: Iterable[np.ndarray], sample_rate: int, should_stop=None) -> bool:
+def play_stream(chunks: Iterable[np.ndarray], sample_rate: int, should_stop=None,
-    """Play int16 chunks as they arrive. Returns False if interrupted."""
+                on_level=None) -> bool:
+    """Play int16 chunks as they arrive. Returns False if interrupted.
+
+    *on_level* receives each chunk's loudness (0..1) just before it is written,
+    which is what drives the mouth: the sprite is animated by the same audio
+    the speakers are getting, not by a guess about how long a word takes."""
     import sounddevice as sd
 
     with sd.OutputStream(
@@ -199,6 +249,11 @@ def play_stream(chunks: Iterable[np.ndarray], sample_rate: int, should_stop=None
                 # now, not at the end of the buffered chunk.
                 out.abort()
                 return False
+            if on_level is not None:
+                try:
+                    on_level(level_of(chunk))
+                except Exception:
+                    on_level = None  # a broken listener must not stop playback
             out.write(chunk)
     return True
 
@@ -213,7 +268,8 @@ def speak_offline(text: str) -> None:
     engine.runAndWait()
 
 
-def speak(text: str, on_error=None, should_stop=None, voice_id: Optional[str] = None) -> bool:
+def speak(text: str, on_error=None, should_stop=None, voice_id: Optional[str] = None,
+          on_level=None) -> bool:
     """Speak *text*, preferring streaming ElevenLabs, then whole-clip
     ElevenLabs, then offline TTS. *on_error*, if given, is called with the
     exception when ElevenLabs fails (useful for logging) — a fallback still
@@ -233,13 +289,14 @@ def speak(text: str, on_error=None, should_stop=None, voice_id: Optional[str] =
                 stream_pcm(text, voice_id=voice_id),
                 config.TTS_SAMPLE_RATE,
                 should_stop=should_stop,
+                on_level=on_level,
             )
         except TtsError as exc:
             if on_error is not None:
                 on_error(exc)
     try:
         pcm, sample_rate = synthesize_pcm(text, voice_id=voice_id)
-        return play_pcm(pcm, sample_rate, should_stop=should_stop)
+        return play_pcm(pcm, sample_rate, should_stop=should_stop, on_level=on_level)
     except TtsError as exc:
         if on_error is not None:
             on_error(exc)

bolt_pet/config.py

@@ -58,6 +58,11 @@ WAKE_CHECK_INTERVAL_SECONDS = float(os.environ.get("WAKE_CHECK_INTERVAL_SECONDS"
 
 DEEPGRAM_API_KEY = os.environ.get("DEEPGRAM_API_KEY", "")
 DEEPGRAM_MODEL = os.environ.get("DEEPGRAM_MODEL", "nova-3")
+# Transcribe *while* you talk instead of uploading the finished clip: frames go
+# up as they are captured, so the transcript is ready the moment the VAD says
+# you stopped. Needs `websocket-client`; falls back to the one-shot upload
+# whenever it can't connect, so turning it on can only help.
+STT_STREAMING = os.environ.get("STT_STREAMING", "true").lower() in ("1", "true", "yes", "on")
 
 # ── TTS (ElevenLabs, requested as raw PCM so playback needs no external
 # player binary — cross-platform via sounddevice instead of shelling out to
@@ -168,6 +173,13 @@ BARGE_IN_FRAMES = int(os.environ.get("BARGE_IN_FRAMES", "4"))  # consecutive lou
 # waking the pet from idle (useful if Bolt's own voice trips the model).
 BARGE_IN_WAKE_THRESHOLD = float(os.environ.get("BARGE_IN_WAKE_THRESHOLD") or 0) or None
 
+# ── streaming the reply ─────────────────────────────────────────────────────
+# Speak each sentence as the server produces it, instead of waiting out the
+# whole model call before the first word. Falls back to the ordinary
+# request/response path automatically if the server has no streaming endpoint
+# or the stream fails before anything has been spoken.
+STREAMING_REPLIES = os.environ.get("STREAMING_REPLIES", "true").lower() in ("1", "true", "yes", "on")
+
 # ── streaming TTS ───────────────────────────────────────────────────────────
 # ElevenLabs' /stream endpoint + chunked playback: the pet starts talking
 # after the first PCM chunk instead of after the whole clip is synthesized.
@@ -307,6 +319,10 @@ PET_WANDER_MARGIN = int(os.environ.get("PET_WANDER_MARGIN", "20"))  # keep off s
 PET_SHAPED_INPUT = os.environ.get("PET_SHAPED_INPUT", "true").lower() in ("1", "true", "yes", "on")
 PET_CLICK_THROUGH = os.environ.get("PET_CLICK_THROUGH", "false").lower() in ("1", "true", "yes", "on")
 # Snap flush to a screen edge when dropped/parked within this many pixels of it.
+# Start where it was left rather than in the bottom-right corner. Ignored when
+# PET_START_X/Y pin it explicitly, and a saved position on a monitor that is no
+# longer plugged in is discarded rather than hiding the pet offscreen.
+PET_REMEMBER_POSITION = os.environ.get("PET_REMEMBER_POSITION", "true").lower() in ("1", "true", "yes", "on")
 PET_EDGE_SNAP = os.environ.get("PET_EDGE_SNAP", "true").lower() in ("1", "true", "yes", "on")
 PET_SNAP_MARGIN = int(os.environ.get("PET_SNAP_MARGIN", "48"))
 

bolt_pet/controller.py

@@ -20,15 +20,20 @@ from typing import Optional
 from PySide6.QtCore import QObject, Signal
 
 from . import (
-    config, dialogue as dialogue_mod, file_delivery, file_ops,
+    config, dialogue as dialogue_mod, speech, file_delivery, file_ops,
     history as history_mod, monitors as monitors_mod, notifications,
     pet_actions, quiet, screen_context, screen_text, self_restart,
     server_client, speech_text, updater,
 )
 from . import __version__
-from .audio import barge_in, mic, stt, tts, wake_word
+from .audio import barge_in, mic, stt, stt_stream, tts, wake_word
 from .state import PetState, PetStateMachine
 
+# A burst while the pet is busy is batched into one turn, but the queue still
+# needs a ceiling — a notification storm must not become an unbounded backlog
+# that gets read out minutes later.
+_MAX_PENDING_NOTIFICATIONS = 12
+
 # How often to re-check whether the pet should be napping. The fullscreen
 # probe shells out to xprop, so this deliberately isn't every heartbeat tick.
 _NAP_CHECK_INTERVAL_SECONDS = 10.0
@@ -41,6 +46,7 @@ class PetController(QObject):
     action = Signal(dict)  # parsed petctl action for the UI to perform
     napping = Signal(bool)  # quiet hours / fullscreen do-not-disturb
     voice_changed = Signal(str)  # name of the server-picked voice ("" = default)
+    mouth = Signal(float)  # 0..1 speech loudness, for lip-sync while talking
     restart_requested = Signal(str)  # version we just updated to
     finished = Signal()
 
@@ -77,6 +83,17 @@ class PetController(QObject):
         self._voice_name = ""
 
         self._barge_in: Optional[barge_in.BargeInDetector] = None
+        # Everything the pet says goes through here; see speech.py for why the
+        # four hand-rolled copies of this became one.
+        self._speaker = speech.Speaker(
+            state=self._state, tts=tts,
+            history=lambda text: self.history.add(history_mod.PET, text, time.time()),
+            on_said=self.said.emit, on_log=self.log.emit,
+            barge_in=lambda: self._barge_in,
+            detail_of=self._barge_in_detail,
+            voice_id=lambda: self._voice_id,
+            on_level=self.mouth.emit,
+        )
         self._napping = False
         self._nap_forced: Optional[bool] = None  # petctl nap on/off overrides the schedule
         self._last_nap_check = 0.0
@@ -249,21 +266,28 @@ class PetController(QObject):
             self._follow_ups = 0
 
         self._state.transition(PetState.LISTENING)
+        # Transcribe while they talk rather than after: frames go to Deepgram
+        # as they are captured, so the text is ready the moment the VAD says
+        # they stopped. None here just means the one-shot path will do it.
+        streamed = stt_stream.StreamingTranscriber.open()
         pcm = mic.record_utterance(
             self._stream,
             should_continue=self._should_continue,
             # Answering a question deserves longer than saying the wake word
             # on purpose does — you were just asked something.
             grace_s=config.FOLLOW_UP_GRACE_SECONDS if following_up else None,
+            on_frame=streamed.feed if streamed is not None else None,
         )
         if pcm is None:
+            if streamed is not None:
+                streamed.finish()
             self._follow_ups = 0  # silence ends the chain
             self._state.transition(PetState.IDLE)
             return
 
         self._state.transition(PetState.THINKING)
         try:
-            text = stt.transcribe(pcm)
+            text = self._transcribe(pcm, streamed)
         except stt.SttError as exc:
             self.log.emit(f"STT failed: {exc}")
             self._state.transition(PetState.ERROR)
@@ -278,9 +302,7 @@ class PetController(QObject):
         try:
             # What's focused right now rides along, so "what's this error?"
             # has a referent without you having to describe the window.
-            reply = server_client.converse(
+            reply = self._ask_server(self._with_context(text))
-                self._with_context(text), on_command=self._handle_command
-            )
         except server_client.ServerError as exc:
             self.log.emit(f"Server error: {exc}")
             self._state.transition(PetState.ERROR)
@@ -289,10 +311,34 @@ class PetController(QObject):
 
         self._check_deliveries()
         self._apply_voice(reply)
-        self._speak(reply.text)
+        if reply.spoken:
+            # Streamed: every sentence was spoken and logged as it arrived.
+            # The text still matters — a reply ending on a question should keep
+            # the mic open — but saying it again would repeat the whole answer.
+            self._after_speaking(reply.text, completed=True)
+        else:
+            self._speak(reply.text)
         self._state.transition(PetState.IDLE)
         self._maybe_self_restart()
 
+    def _transcribe(self, pcm, streamed) -> str:
+        """The transcript, from the live stream if it produced one.
+
+        The fallback is not a rare path to be tolerated — it is the safety net
+        that lets streaming be switched on at all. Whatever happened to the
+        socket, the full audio is still buffered here, so a failed stream costs
+        one ordinary upload and nothing else."""
+        if streamed is not None:
+            try:
+                text = streamed.finish()
+            except Exception:
+                self.log.emit("Streaming STT failed — falling back.")
+                text = ""
+            if text:
+                self.log.emit("(transcribed while you spoke)")
+                return text
+        return stt.transcribe(pcm)
+
     def _with_context(self, text: str) -> str:
         """Everything the server gets alongside what you actually said: the
         focused window title, and a one-line note about the screen layout so
@@ -492,6 +538,26 @@ class PetController(QObject):
             self._speak(reply.text)
         self._state.force(PetState.IDLE)
 
+    def _ask_server(self, text: str):
+        """One turn with the server, streamed when possible.
+
+        Streaming speaks each sentence as it is generated, so the wait is
+        time-to-first-sentence rather than the whole model call. It falls back
+        to the ordinary request/response path when the server has no streaming
+        endpoint, or when a stream dies *before* anything was spoken — after
+        that, retrying would say the first half twice."""
+        if config.STREAMING_REPLIES:
+            try:
+                return server_client.converse_stream(
+                    text, on_say=self._speak_stream_chunk,
+                    on_command=self._handle_command,
+                )
+            except server_client.ServerError as exc:
+                self.log.emit(f"Streaming unavailable ({exc}) — using the plain path.")
+        return server_client.converse(
+            text, on_command=self._handle_command, on_say=self._speak_holding,
+        )
+
     def _play_dialogue(self, scene: dict) -> str:
         """Play a `dialoguectl` scene and report back up the relay.
 
@@ -525,20 +591,8 @@ class PetController(QObject):
             # scene or fix the voice, and try again inside the same turn.
             return f"[dialogue] couldn't synthesize it: {exc}"
 
-        resume = self._state.state
+        completed = self._speaker.say_pcm(
-        self._state.transition(PetState.TALKING)
+            speech.Utterance.scene(text), pcm=pcm, sample_rate=sample_rate)
-        self.said.emit(speech_text.for_display(text))
-        self.history.add(history_mod.PET, text, time.time())
-
-        should_stop = None
-        if self._barge_in is not None:
-            self._barge_in.reset()
-            should_stop = self._barge_in.check
-        completed = tts.play_pcm(pcm, sample_rate, should_stop=should_stop)
-        if self._barge_in is not None:
-            self._barge_in.reset()  # the pet's own voices are in the wake window
-        if resume in (PetState.THINKING, PetState.IDLE):
-            self._state.transition(resume)
 
         if not completed:
             return dialogue_mod.describe(scene) + " (interrupted — they talked over it)"
@@ -569,32 +623,30 @@ class PetController(QObject):
             self.reset_voice()
 
     def _speak(self, text: str) -> None:
-        self._state.transition(PetState.TALKING)
+        """The answer: transcript, bubble, and the follow-up rule."""
-        # Bubble gets the markdown stripped but emoji kept (it can't render
+        completed = self._speaker.say(speech.Utterance.reply(text))
-        # **bold** but draws emoji fine); tts.speak() does its own, stricter
+        self._after_speaking(text, completed=completed,
-        # sanitizing for the voice.
+                             detail=self._speaker.last_detail)
-        self.said.emit(speech_text.for_display(text))
-        self.log.emit(f"Bolt: {text}")
-        self.history.add(history_mod.PET, text, time.time())
 
-        should_stop = None
+    def _speak_holding(self, text: str) -> None:
-        if self._barge_in is not None:
+        """"Give me a sec" while a tool runs — filler, so no transcript, and it
-            self._barge_in.reset()
+        resumes the state it interrupted because the turn isn't over."""
-            should_stop = self._barge_in.check
+        self._speaker.say(speech.Utterance.holding(text))
-        completed = tts.speak(
+
-            text,
+    def _speak_stream_chunk(self, text: str) -> None:
-            on_error=lambda exc: self.log.emit(f"TTS failed: {exc}"),
+        """One sentence of a streamed answer, spoken the moment it arrives."""
-            should_stop=should_stop,
+        completed = self._speaker.say(speech.Utterance.stream_chunk(text))
-            voice_id=self._voice_id or None,
+        if not completed:
-        )
+            self.log.emit("Interrupted — listening.")
-        # Read the scoring history *before* resetting, or the log reports the
+            self._follow_ups = 0
-        # blank counters instead of what actually fired.
+            self._talk_now.set()
-        detail = self._barge_in_detail()
+
-        if self._barge_in is not None:
+    def _after_speaking(self, text: str, *, completed: bool, detail: str = "") -> None:
-            # Playback fed the pet's own voice into the wake model's rolling
+        """What happens once an answer has been said, however it was said.
-            # window. Clear it before the idle listener starts scoring again,
+
-            # or Bolt's last sentence is still in there being re-scored.
+        Shared by the plain and streamed paths: a streamed reply is spoken
-            self._barge_in.reset()
+        sentence by sentence, but it still has to obey the same rules about
+        keeping the mic open when it ended on a question."""
         if not completed:
             # You talked over it — take that as the start of the next turn
             # rather than making you say the wake word again.
@@ -614,20 +666,22 @@ class PetController(QObject):
     def _should_follow_up(self, text: str) -> bool:
         """Whether *text* leaves the pet waiting on an answer.
 
-        Muted is excluded because mute means "don't listen to me" — an
+        The rule itself — question, cap, off switch — is
-        automatic turn would walk straight past it. Napping isn't: quiet
+        `speech.follow_up_decision`, because it is a rule with an off-by-one in
-        hours suppress the pet *starting* something, and a question is only
+        it and deserves a test that needs no audio. What stays here is the part
-        ever asked in reply to you."""
+        that is genuinely the controller's: mute, and the log line. Muted is
-        if not config.FOLLOW_UP_LISTEN or self._muted:
+        excluded because mute means "don't listen to me" and an automatic turn
+        would walk straight past it. Napping isn't: quiet hours suppress the
+        pet *starting* something, and a question is only ever asked in reply to
+        you."""
+        if self._muted:
             return False
-        if not speech_text.is_question(text):
+        keep, why = speech.follow_up_decision(text, completed=True, follow_ups=self._follow_ups)
-            return False
+        if not keep and "cap" in why:
-        # Only worth mentioning the cap on a reply that would otherwise have
+            # Only worth mentioning the cap on a reply that would otherwise have
-        # kept listening, or it fires on every statement the pet makes.
+            # kept listening, or it fires on every statement the pet makes.
-        if config.FOLLOW_UP_MAX_TURNS > 0 and self._follow_ups >= config.FOLLOW_UP_MAX_TURNS:
             self.log.emit("Follow-up limit reached — say the wake word to keep going.")
-            return False
+        return keep
-        return True
 
     def _barge_in_detail(self) -> str:
         """Why the interruption fired, for the log. How far into playback it
@@ -685,33 +739,54 @@ class PetController(QObject):
 
     def _queue_notification(self, notification: notifications.Notification) -> None:
         """Called on the watcher thread — just queue it; forwarding happens on
-        the pipeline thread where it can't collide with a live conversation."""
+        the pipeline thread where it can't collide with a live conversation.
-        if not self._notification_gate.should_forward(notification, time.monotonic()):
+
+        Only the *filter* applies here. The rate limit used to as well, which
+        meant a second message arriving inside the window was silently thrown
+        away — with NOTIFICATION_MIN_INTERVAL_SECONDS=60, two texts a minute
+        apart and you only ever heard about one of them. Losing a message from
+        a person to save a round trip is the wrong trade; they are batched at
+        the far end instead, which costs the same one round trip and keeps
+        them all."""
+        if not self._notification_gate.matches(notification):
             return
         with self._notification_lock:
+            if len(self._pending_notifications) >= _MAX_PENDING_NOTIFICATIONS:
+                self._pending_notifications.pop(0)  # bound it; oldest goes first
             self._pending_notifications.append(notification)
 
     def _drain_notifications(self) -> None:
+        """Forward everything waiting as ONE turn.
+
+        Batching is what makes it safe to keep every notification: five that
+        arrived while the pet was mid-conversation become one message and one
+        round trip, instead of five separate interruptions queued up to fire
+        back to back."""
         with self._notification_lock:
             pending, self._pending_notifications = self._pending_notifications, []
+        if not pending or not self._running or self._napping:
+            return
         for notification in pending:
-            if not self._running or self._napping:
-                return
             self.log.emit(f"Notification: {notification.as_text()}")
             self.history.add(history_mod.SYSTEM, notification.as_text(), time.time())
-            try:
+
-                reply = server_client.converse(
+        if len(pending) == 1:
-                    f"[desktop notification] {notification.as_text()}",
+            message = f"[desktop notification] {pending[0].as_text()}"
-                    on_command=self._handle_command,
+        else:
-                )
+            lines = "\n".join(f"- {n.as_text()}" for n in pending)
-            except server_client.ServerError as exc:
+            message = f"[{len(pending)} desktop notifications]\n{lines}"
-                self.log.emit(f"Couldn't forward notification: {exc}")
+        try:
-                return
+            reply = server_client.converse(
-            self._check_deliveries()
+                message, on_command=self._handle_command, on_say=self._speak_holding,
-            self._apply_voice(reply)
+            )
-            if reply.text.strip():
+        except server_client.ServerError as exc:
-                self._speak(reply.text)
+            self.log.emit(f"Couldn't forward notification: {exc}")
-                self._state.transition(PetState.IDLE)
+            return
+        self._check_deliveries()
+        self._apply_voice(reply)
+        if reply.text.strip() and not reply.spoken:
+            self._speak(reply.text)
+        self._state.transition(PetState.IDLE)
 
     # ── file delivery ────────────────────────────────────────────────────
 
@@ -784,20 +859,36 @@ class PetController(QObject):
     # ── heartbeat ────────────────────────────────────────────────────────
 
     def _maybe_heartbeat(self) -> None:
+        """Called from the wake listener's tick (~every WAKE_CHECK_INTERVAL_SECONDS).
+
+        Two different cadences live here, and conflating them was costing
+        minutes. A *notification* is an event that already happened — it should
+        go out as soon as the pet is free, which is the next tick. The
+        *heartbeat* is a poll, and polling the server every 1.2s would be
+        absurd, so it stays on its own interval."""
         self._refresh_nap_state()
         self._maybe_update()
         if self._update_pending:
             return  # on the way out — don't start a conversation now
+
+        # Notifications: every tick, not every heartbeat.
+        if self._state.state == PetState.IDLE and not self._napping:
+            self._drain_notifications()
+
         now = time.monotonic()
         if now - self._last_heartbeat < config.HEARTBEAT_INTERVAL_SECONDS:
             return
-        self._last_heartbeat = now
         if self._state.state != PetState.IDLE:
+            # Mid-conversation. Do NOT stamp the clock — an earlier version
+            # did, so a heartbeat that landed while the pet was talking burned
+            # its slot and waited another full interval. With follow-up
+            # listening that could repeat for several cycles, which is why a
+            # notification could sit unspoken for five or ten minutes.
             return
         if self._napping:
             return  # quiet hours: still answers when spoken to, just doesn't start
+        self._last_heartbeat = now
         self._check_deliveries()
-        self._drain_notifications()
         if self._state.state != PetState.IDLE:
             return
         try:

bolt_pet/dialogue.py

@@ -224,4 +224,12 @@ def describe(action: dict, *, played: bool = True) -> str:
     return (
         f"[dialogue] played {len(lines)} line{'s' if len(lines) != 1 else ''} "
         f"in {len(voices)} voice{'s' if len(voices) != 1 else ''}: {', '.join(voices)}{note}"
+        # The scene was spoken out loud before this result got back to the
+        # server, and the model has no other way to know that. Without saying
+        # so it writes a final reply summarising what the user just heard, and
+        # the pet says the same thing twice in a row (observed 2026-07-31).
+        # An instruction delivered here, at the moment it applies, lands far
+        # better than a rule buried in a long system prompt.
+        "\nThe user HEARD this already. Do not repeat, summarise or narrate it "
+        "in your reply — answer with at most one short line, or nothing new."
     )

bolt_pet/doctor.py

@@ -0,0 +1,272 @@
+"""`python -m bolt_pet --doctor` — is this install actually going to work?
+
+Written after a week of debugging things a preflight would have shown in one
+command: a missing port publish, a wrong reverse-proxy header, a venv whose
+python was a zero-byte file, an OCR engine that was never installed. Every one
+of those presented as "the pet is being weird" and took a conversation to find.
+
+Each check is independent and reports one of three things — ok, a warning
+(works, but degraded), or a failure (this will not do what you expect) — and
+says *what to do about it* rather than just what is wrong. Nothing here raises:
+a doctor that crashes on a broken install is diagnosing the wrong patient.
+
+Deliberately does not open the mic or call a paid API by default. It checks
+that the door is unlocked, not that the room is furnished; `--deep` is there
+when you want it to actually knock.
+"""
+
+from __future__ import annotations
+
+import importlib
+import shutil
+import sys
+from dataclasses import dataclass
+from typing import Callable, Optional
+
+from . import config
+
+OK, WARN, FAIL = "ok", "warn", "fail"
+
+_MARKS = {OK: "  ok  ", WARN: " warn ", FAIL: " FAIL "}
+
+
+@dataclass
+class Check:
+    name: str
+    status: str
+    detail: str = ""
+    fix: str = ""
+
+    def line(self) -> str:
+        text = f"[{_MARKS[self.status]}] {self.name:22} {self.detail}"
+        if self.fix and self.status != OK:
+            text += f"\n{'':32}→ {self.fix}"
+        return text
+
+
+def _module(name: str) -> bool:
+    try:
+        importlib.import_module(name)
+        return True
+    except Exception:
+        return False
+
+
+# ── the checks ──────────────────────────────────────────────────────────────
+
+def check_config() -> Check:
+    missing = config.missing_config()
+    if missing:
+        return Check("server config", FAIL, f"missing {', '.join(missing)}",
+                     "set them in .env — without these the controller exits at startup")
+    return Check("server config", OK, f"{config.SERVER_URL} as {config.SESSION_ID}")
+
+
+def check_server(deep: bool = False) -> Check:
+    if not config.SERVER_URL:
+        return Check("server", FAIL, "no BOLT_SERVER_URL",
+                     "cp .env.example .env and set BOLT_SERVER_URL to your Bolt server")
+    if not deep:
+        return Check("server", OK, f"{config.SERVER_URL} (not contacted; --deep to try)")
+    try:
+        from . import server_client
+
+        health = server_client.check_health(timeout=8)
+        return Check("server", OK, f"reachable — {health}")
+    except Exception as exc:
+        return Check("server", FAIL, f"unreachable: {exc}",
+                     "check the URL, the key, and that the container is up")
+
+
+def check_streaming_endpoint(deep: bool = False) -> Check:
+    """The streamed-reply endpoint is newer than some deployed servers."""
+    if not config.STREAMING_REPLIES:
+        return Check("streamed replies", WARN, "disabled (STREAMING_REPLIES=false)")
+    if not deep:
+        return Check("streamed replies", OK, "enabled (endpoint not probed)")
+    try:
+        import requests
+
+        response = requests.post(
+            f"{config.SERVER_URL}/desk/converse_stream",
+            json={"session_id": config.SESSION_ID, "text": ""},
+            headers={"X-Desk-Api-Key": config.API_KEY}, timeout=8, stream=True,
+        )
+        if response.status_code == 404:
+            return Check("streamed replies", WARN, "server has no /desk/converse_stream",
+                         "update the server, or set STREAMING_REPLIES=false to skip the probe")
+        return Check("streamed replies", OK, f"endpoint answered {response.status_code}")
+    except Exception as exc:
+        return Check("streamed replies", WARN, f"probe failed: {exc}")
+
+
+def check_microphone(deep: bool = False) -> Check:
+    if not _module("sounddevice"):
+        return Check("microphone", FAIL, "sounddevice is not installed",
+                     "pip install -r requirements.txt")
+    try:
+        import sounddevice as sd
+
+        devices = [d for d in sd.query_devices() if d.get("max_input_channels", 0) > 0]
+        if not devices:
+            return Check("microphone", FAIL, "no input devices",
+                         "on Linux check PipeWire/PulseAudio is running as your user")
+        chosen = config.MIC_DEVICE or "system default"
+        if not deep:
+            return Check("microphone", OK, f"{len(devices)} input device(s), using {chosen}")
+        with sd.InputStream(samplerate=config.SAMPLE_RATE, channels=1, dtype="int16",
+                            device=config.MIC_DEVICE, blocksize=config.FRAME_LEN):
+            pass
+        return Check("microphone", OK, f"opened at {config.SAMPLE_RATE} Hz ({chosen})")
+    except Exception as exc:
+        return Check("microphone", FAIL, f"could not open: {exc}",
+                     "don't run the pet as root — PortAudio can't reach your PipeWire socket")
+
+
+def check_wake_model() -> Check:
+    from pathlib import Path
+
+    path = Path(config.WAKE_MODEL_PATH)
+    if not path.exists():
+        return Check("wake word", FAIL, f"{path.name} is missing",
+                     "it ships in the project root; check WAKE_MODEL_FILE")
+    if not _module("openwakeword"):
+        return Check("wake word", FAIL, "openwakeword is not installed",
+                     "pip install -r requirements.txt")
+    return Check("wake word", OK, f"{path.name}, threshold {config.WAKE_WORD_THRESHOLD}")
+
+
+def check_stt() -> Check:
+    if not config.DEEPGRAM_API_KEY:
+        return Check("speech-to-text", FAIL, "no DEEPGRAM_API_KEY",
+                     "set it in .env — nothing you say can be transcribed without it")
+    if config.STT_STREAMING and not _module("websocket"):
+        return Check("speech-to-text", WARN, "streaming on, but websocket-client is missing",
+                     "pip install websocket-client — it falls back to one-shot uploads")
+    mode = "streaming" if config.STT_STREAMING else "one-shot"
+    return Check("speech-to-text", OK, f"Deepgram {config.DEEPGRAM_MODEL}, {mode}")
+
+
+def check_tts() -> Check:
+    if not (config.ELEVENLABS_API_KEY and config.ELEVENLABS_VOICE_ID):
+        if _module("pyttsx3"):
+            return Check("text-to-speech", WARN, "no ElevenLabs key/voice — offline voice only",
+                         "set ELEVENLABS_API_KEY and ELEVENLABS_VOICE_ID for the real voice")
+        return Check("text-to-speech", FAIL, "no ElevenLabs config and no pyttsx3 fallback")
+    return Check("text-to-speech", OK,
+                 f"ElevenLabs {config.ELEVENLABS_MODEL_ID}, voice …{config.ELEVENLABS_VOICE_ID[-6:]}")
+
+
+def check_dialogue() -> Check:
+    if not config.DIALOGUE:
+        return Check("multi-voice scenes", WARN, "disabled (DIALOGUE=false)")
+    from . import dialogue
+
+    cast = dialogue.parse_voice_map(config.DIALOGUE_VOICES)
+    if not cast:
+        return Check("multi-voice scenes", WARN, "no cast configured — only 'self' works",
+                     'set DIALOGUE_VOICES=narrator:<id>,villain:<id>')
+    return Check("multi-voice scenes", OK, f"{len(cast)} voice(s): {', '.join(sorted(cast))}")
+
+
+def check_screen_text() -> Check:
+    if not config.SCREEN_TEXT:
+        return Check("screen reading", WARN, "disabled (SCREEN_TEXT=false)")
+    if not _module("mss"):
+        return Check("screen reading", WARN, "mss is not installed — petctl read will decline",
+                     "pip install mss (and note it cannot capture on Wayland)")
+    engine = "pytesseract" if _module("pytesseract") else (
+        "rapidocr" if _module("rapidocr_onnxruntime") else "")
+    if not engine:
+        return Check("screen reading", WARN, "no OCR engine",
+                     "pip install pytesseract && apt install tesseract-ocr, "
+                     "or pip install rapidocr-onnxruntime")
+    if engine == "pytesseract" and not shutil.which("tesseract"):
+        return Check("screen reading", WARN, "pytesseract is installed but tesseract is not",
+                     "apt install tesseract-ocr")
+    return Check("screen reading", OK, f"mss + {engine}")
+
+
+def check_hotkey() -> Check:
+    if not _module("pynput"):
+        return Check("push-to-talk", WARN, "pynput is not installed",
+                     "the wake word still works; pip install pynput for the hotkey")
+    import os
+
+    if os.environ.get("WAYLAND_DISPLAY") and not os.environ.get("DISPLAY"):
+        return Check("push-to-talk", WARN, "Wayland session — global hotkeys usually blocked",
+                     "use the wake word, or click the pet")
+    return Check("push-to-talk", OK, config.PUSH_TO_TALK_HOTKEY)
+
+
+def check_sprites() -> Check:
+    from pathlib import Path
+
+    root = Path(__file__).resolve().parent / "assets" / "sprites"
+    if not root.exists():
+        return Check("sprites", WARN, "no art — the placeholder blob will be drawn",
+                     "python scripts/generate_bolt_sprites.py")
+    counts = {d.name: len(list(d.glob("*.png"))) for d in sorted(root.iterdir()) if d.is_dir()}
+    empty = [name for name, count in counts.items() if count == 0]
+    if empty:
+        return Check("sprites", WARN, f"no frames for: {', '.join(empty)}",
+                     "python scripts/generate_bolt_sprites.py")
+    return Check("sprites", OK, ", ".join(f"{n} {c}" for n, c in counts.items()))
+
+
+def check_latency() -> Check:
+    """The setting most likely to make it feel slow, and the least obvious."""
+    silence = config.SILENCE_END_SEC
+    if silence >= 1.5:
+        return Check("turn latency", WARN, f"VAD_SILENCE_END_SEC={silence:g}s of dead air per turn",
+                     "0.8-1.0 feels markedly snappier; it is pure wait before anything starts")
+    return Check("turn latency", OK, f"silence timeout {silence:g}s, "
+                 f"streaming {'on' if config.STREAMING_REPLIES else 'off'}")
+
+
+CHECKS: tuple[tuple[str, Callable], ...] = (
+    ("config", check_config),
+    ("server", check_server),
+    ("stream", check_streaming_endpoint),
+    ("mic", check_microphone),
+    ("wake", check_wake_model),
+    ("stt", check_stt),
+    ("tts", check_tts),
+    ("dialogue", check_dialogue),
+    ("screen", check_screen_text),
+    ("hotkey", check_hotkey),
+    ("sprites", check_sprites),
+    ("latency", check_latency),
+)
+
+
+def run(deep: bool = False) -> list[Check]:
+    results = []
+    for _name, check in CHECKS:
+        try:
+            try:
+                results.append(check(deep))
+            except TypeError:
+                results.append(check())
+        except Exception as exc:  # a broken check must not hide the others
+            results.append(Check(_name, FAIL, f"the check itself failed: {exc}"))
+    return results
+
+
+def main(argv: Optional[list] = None) -> int:
+    argv = list(argv if argv is not None else sys.argv[1:])
+    deep = "--deep" in argv
+    print(f"Bolt pet preflight{' (deep: contacting the server and opening the mic)' if deep else ''}\n")
+    results = run(deep=deep)
+    for check in results:
+        print(check.line())
+    failures = [c for c in results if c.status == FAIL]
+    warnings = [c for c in results if c.status == WARN]
+    print()
+    if failures:
+        print(f"{len(failures)} problem(s) will stop this working. Fix those first.")
+    elif warnings:
+        print(f"Ready. {len(warnings)} thing(s) degraded but working.")
+    else:
+        print("Everything checks out.")
+    return 1 if failures else 0

bolt_pet/server_client.py

@@ -19,6 +19,7 @@ Kept dependency-free beyond `requests` so it's easy to unit test with mocks.
 
 from __future__ import annotations
 
+import json
 import subprocess
 from pathlib import Path
 from typing import Callable, NamedTuple, Optional
@@ -43,6 +44,10 @@ class Reply(NamedTuple):
     text: str
     voice_id: str = ""
     voice_name: str = ""
+    # True when the sentences were already spoken as they streamed in. The
+    # text is still carried — the follow-up rule needs to see whether the
+    # answer ended on a question — it just must not be read out again.
+    spoken: bool = False
 
 
 def _headers() -> dict:
@@ -90,12 +95,20 @@ def converse(
     text: str,
     on_command: Callable[[str], str] = run_local_command,
     timeout: float = 120.0,
+    on_say: Optional[Callable[[str], None]] = None,
 ) -> Reply:
     """Send one turn of conversation to the desk API, relaying any commands
     the server sends back until it produces a final reply.
 
     *on_command* is injectable for tests; defaults to actually running the
     command locally (matching bolt_desk.py's behavior).
+
+    *on_say* is called with a short holding line ("give me a sec") when the
+    server sends one alongside a command. It is the difference between silence
+    and an answer while a tool runs: the model's acknowledgement used to be
+    discarded server-side, so the whole round trip was dead air and the model
+    then repeated itself in the final reply. Optional, so an older pet against
+    a newer server simply stays quiet as before.
     """
     headers = _headers()
     try:
@@ -111,6 +124,13 @@ def converse(
     for _ in range(_MAX_RELAY_HOPS):
         if payload.get("type") != "command":
             break
+        holding = str(payload.get("say") or "").strip()
+        if holding and on_say is not None:
+            # Spoken *before* the command runs — that is the whole point.
+            try:
+                on_say(holding)
+            except Exception:
+                pass  # a failed acknowledgement must not cost the tool call
         output = on_command(str(payload.get("command") or ""))
         try:
             response = requests.post(
@@ -135,6 +155,104 @@ def converse(
     raise ServerError(str(payload.get("error") or "unknown server response"))
 
 
+def converse_stream(
+    text: str,
+    on_say: Callable[[str], None],
+    on_command: Callable[[str], str] = run_local_command,
+    timeout: float = 180.0,
+) -> Reply:
+    """Same turn as converse(), but speaking each sentence as it arrives.
+
+    Without this the pet waits out the *entire* model call before a single
+    word is heard; with it the wait is time-to-first-sentence, which on a
+    multi-sentence answer is most of the difference.
+
+    Falls back by raising ServerError before anything has been spoken — the
+    caller then retries the ordinary path and the user never finds out. Once a
+    sentence *has* been spoken there is no going back, so late failures end the
+    turn with whatever was said rather than repeating it.
+    """
+    spoke_anything = False
+    try:
+        response = requests.post(
+            f"{config.SERVER_URL}/desk/converse_stream",
+            json={"session_id": config.SESSION_ID, "text": text},
+            headers=_headers(), timeout=timeout, stream=True,
+        )
+        response.raise_for_status()
+        for raw in response.iter_lines(decode_unicode=True):
+            if not raw:
+                continue
+            try:
+                event = json.loads(raw)
+            except (TypeError, ValueError):
+                continue
+            kind = str(event.get("type") or "")
+
+            if kind == "say":
+                line = str(event.get("text") or "").strip()
+                if line:
+                    spoke_anything = True
+                    on_say(line)
+            elif kind == "command":
+                # The tool loop is request/response, so the rest of the turn
+                # finishes through the ordinary relay rather than inside the
+                # stream — one protocol for tools, not two.
+                response.close()
+                return _finish_relay(event, on_command, on_say)
+            elif kind == "reply":
+                return Reply(
+                    text=str(event.get("text") or ""),
+                    voice_id=str(event.get("voice_id") or ""),
+                    voice_name=str(event.get("voice_name") or ""),
+                    spoken=bool(event.get("already_spoken")),
+                )
+            elif kind == "error":
+                raise ServerError(str(event.get("error") or "stream failed"))
+    except ServerError:
+        raise
+    except Exception as exc:
+        if spoke_anything:
+            # Half a reply is out loud already; ending quietly beats saying it
+            # all again through the fallback path.
+            return Reply(text="", spoken=True)
+        raise ServerError(f"streaming failed: {exc}") from exc
+    raise ServerError("stream ended without a reply")
+
+
+def _finish_relay(event: dict, on_command, on_say) -> Reply:
+    """Run the tool the stream handed over, then continue the classic relay."""
+    payload = dict(event)
+    headers = _headers()
+    for _ in range(_MAX_RELAY_HOPS):
+        if payload.get("type") != "command":
+            break
+        holding = str(payload.get("say") or "").strip()
+        if holding and on_say is not None:
+            try:
+                on_say(holding)
+            except Exception:
+                pass
+        output = on_command(str(payload.get("command") or ""))
+        try:
+            response = requests.post(
+                f"{config.SERVER_URL}/desk/tool_result",
+                json={"session_id": config.SESSION_ID,
+                      "token": payload.get("token"), "output": output},
+                headers=headers, timeout=180,
+            )
+            payload = response.json()
+        except Exception as exc:
+            raise ServerError(f"couldn't reach the server during tool relay: {exc}") from exc
+    if payload.get("type") == "reply":
+        return Reply(
+            text=str(payload.get("text") or ""),
+            voice_id=str(payload.get("voice_id") or ""),
+            voice_name=str(payload.get("voice_name") or ""),
+        )
+    raise ServerError(str(payload.get("error") or "unknown server response"))
+
+
 def list_outbox_files(timeout: float = 15.0) -> list:
     """Files the server has queued for this session via its deliver_files
     tool (e.g. "send me that report" during a conversation) — each entry has

bolt_pet/speech.py

@@ -0,0 +1,194 @@
+"""Everything the pet says, and the policy differences between kinds of saying.
+
+There used to be four of these in controller.py — a reply, a holding line, a
+streamed sentence, a dialogue scene — each written when its feature was built,
+each repeating the same dance: transition state, show the bubble, maybe record
+history, reset barge-in, call TTS, reset barge-in again, resume state, decide
+whether to keep the mic open. Only the *policy* differed, and the copies had
+already started to drift: one forgot to arm barge-in, another logged a
+different prefix, a third skipped the follow-up rule.
+
+So the dance lives here once, and the differences are data:
+
+    reply        the answer. Transcript, bubble, follow-up rule, ends IDLE.
+    holding      "give me a sec" while a tool runs. No transcript — it is
+                 filler, and the transcript should keep the answer. Resumes
+                 whatever state it interrupted, because the turn isn't over.
+    stream       one sentence of a streamed answer. Transcript and bubble like
+                 a reply, but stays TALKING so the sprite doesn't flicker
+                 between sentences, and the follow-up rule waits for the last.
+    scene        a dialoguectl take. Transcript (the user heard it), resumes
+                 mid-turn like a holding line.
+
+The controller keeps the pipeline; this keeps the rules about talking.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Callable, Optional
+
+from . import config, speech_text
+from .state import PetState
+
+
+@dataclass(frozen=True)
+class Utterance:
+    """One thing to say, and how saying it should behave."""
+
+    text: str
+    record: bool = True         # goes in the transcript, or is it filler?
+    resume: bool = False        # return to the state it interrupted (mid-turn)
+    hold_talking: bool = False  # stay TALKING afterwards (more is coming)
+    follow_up: bool = True      # may leave the mic open if it ends on a question
+    interruptible: bool = True  # arm barge-in for this one
+    log_prefix: str = "Bolt"
+
+    @classmethod
+    def reply(cls, text: str) -> "Utterance":
+        return cls(text)
+
+    @classmethod
+    def holding(cls, text: str) -> "Utterance":
+        # Filler: no transcript, no follow-up, and not interruptible — cutting
+        # off "give me a sec" would strand the tool that is already running.
+        return cls(text, record=False, resume=True, follow_up=False,
+                   interruptible=False, log_prefix="Bolt (holding)")
+
+    @classmethod
+    def stream_chunk(cls, text: str) -> "Utterance":
+        return cls(text, hold_talking=True, follow_up=False)
+
+    @classmethod
+    def scene(cls, text: str) -> "Utterance":
+        return cls(text, resume=True, follow_up=False)
+
+
+class Speaker:
+    """Says things on behalf of the controller.
+
+    Collaborators are passed in rather than reached for, so the whole of
+    speaking is testable without Qt, audio hardware or a state machine: hand it
+    fakes and assert on what came out."""
+
+    def __init__(self, *, state, tts, history=None, on_said=None, on_log=None,
+                 barge_in: Callable[[], object] = lambda: None,
+                 voice_id: Callable[[], str] = lambda: "",
+                 detail_of: Callable[[], str] = lambda: "",
+                 on_level: Optional[Callable[[float], None]] = None):
+        self._state = state
+        self._tts = tts
+        self._history = history
+        self._on_said = on_said or (lambda _text: None)
+        self._on_log = on_log or (lambda _msg: None)
+        # Read through a callable rather than held: the detector is built after
+        # the speaker (it needs the mic stream), replaced when barge-in mode
+        # changes, and swapped by tests. Two copies of it drifting apart is a
+        # bug nobody notices until the wake model starts hearing the pet.
+        self._barge_in_of = barge_in
+        self._voice_id = voice_id
+        # What actually fired, captured BEFORE the post-playback reset. Read it
+        # afterwards and every interruption reports 0.000 at frame 0 — which
+        # looks like hard evidence and is nothing of the sort.
+        self._detail_of = detail_of
+        self.last_detail = ""
+        self._on_level = on_level
+
+    @property
+    def _barge_in(self):
+        try:
+            return self._barge_in_of()
+        except Exception:
+            return None
+
+    def say(self, utterance: Utterance) -> bool:
+        """Speak it. Returns False if it was interrupted.
+
+        The one place that knows the order these steps go in — which is the
+        point, because getting that order wrong is invisible until the wake
+        model starts hearing the pet's own voice."""
+        line = speech_text.for_display(utterance.text)
+        if not line:
+            return True
+
+        resume_state = self._state.state
+        if self._state.state != PetState.TALKING:
+            self._state.transition(PetState.TALKING)
+        self._on_said(line)
+        self._on_log(f"{utterance.log_prefix}: {line}")
+        if utterance.record and self._history is not None:
+            self._history(utterance.text)
+
+        should_stop = None
+        if self._barge_in is not None and utterance.interruptible:
+            self._barge_in.reset()
+            should_stop = self._barge_in.check
+
+        completed = self._tts.speak(
+            utterance.text,
+            on_error=lambda exc: self._on_log(f"TTS failed: {exc}"),
+            should_stop=should_stop,
+            voice_id=self._voice_id() or None,
+            on_level=self._on_level,
+        )
+        self.last_detail = self._detail_of()
+        if self._barge_in is not None:
+            # Playback fed the pet's own voice into the wake model's rolling
+            # window. Clear it before the idle listener scores again, or the
+            # last sentence is still in there being re-heard.
+            self._barge_in.reset()
+        if self._on_level is not None:
+            self._on_level(0.0)  # mouth closed; nothing is playing now
+
+        if utterance.resume and resume_state in (PetState.THINKING, PetState.IDLE):
+            self._state.transition(resume_state)
+        return bool(completed)
+
+
+    def say_pcm(self, utterance: Utterance, *, pcm, sample_rate: int) -> bool:
+        """Speak audio that is already synthesized — a dialoguectl scene.
+
+        Same policy, same barge-in handling, same mouth; only the source of
+        the samples differs. Sharing this is why a scene can be talked over
+        exactly like an ordinary reply."""
+        line = speech_text.for_display(utterance.text)
+        resume_state = self._state.state
+        if self._state.state != PetState.TALKING:
+            self._state.transition(PetState.TALKING)
+        if line:
+            self._on_said(line)
+        self._on_log(f"{utterance.log_prefix}: {line}")
+        if utterance.record and self._history is not None:
+            self._history(utterance.text)
+
+        should_stop = None
+        if self._barge_in is not None and utterance.interruptible:
+            self._barge_in.reset()
+            should_stop = self._barge_in.check
+        completed = self._tts.play_pcm(
+            pcm, sample_rate, should_stop=should_stop, on_level=self._on_level)
+        self.last_detail = self._detail_of()
+        if self._barge_in is not None:
+            self._barge_in.reset()
+        if self._on_level is not None:
+            self._on_level(0.0)
+        if utterance.resume and resume_state in (PetState.THINKING, PetState.IDLE):
+            self._state.transition(resume_state)
+        return bool(completed)
+
+
+def follow_up_decision(text: str, *, completed: bool, follow_ups: int) -> tuple[bool, str]:
+    """Whether to keep listening after speaking, and why.
+
+    Split out as a pure function because it is a *rule* with an off-by-one cap
+    in it, and rules with counters deserve a test that doesn't need audio."""
+    if not completed:
+        return True, "interrupted"
+    if not speech_text.is_question(text):
+        return False, ""
+    cap = config.FOLLOW_UP_MAX_TURNS
+    if not config.FOLLOW_UP_LISTEN:
+        return False, ""
+    if cap > 0 and follow_ups >= cap:
+        return False, f"follow-up cap ({cap}) reached"
+    return True, "question"

bolt_pet/speech_text.py

@@ -97,10 +97,34 @@ def _bullets_to_sentences(text: str) -> str:
     return " ".join(line if line[-1] in ".!?:,;" else line + "." for line in lines)
 
 
+# ElevenLabs v3 delivery tags — "[laughing]", "[whispering]", "[sighs]". They
+# are instructions to the *dialogue* model (see dialogue.py), and only inside a
+# dialoguectl scene. Left in an ordinary reply they reach eleven_flash_v2,
+# which has no idea what they mean and simply reads the word: observed live
+# 2026-07-31, the pet announcing "laughing That one came through clean".
+#
+# Matched narrowly — a bracketed adverb/gerund, or one of the noise words that
+# aren't either — so real bracketed text ("[1]", "[see the docs]") survives.
+_AUDIO_TAG_RE = re.compile(
+    r"\[\s*(?:"
+    r"[a-z]+(?:ly|ing)"
+    r"|laughs?|chuckles?|sighs?|exhales?|inhales?|gulps?|pause|beat"
+    r"|clears throat|shouts?|whispers?|cries|sings?|gasps?|snorts?"
+    r"|sarcastic|excited|curious|nervous|angry|sad|happy|deadpan|flat"
+    r")\s*\]",
+    re.IGNORECASE,
+)
+
+
+def strip_audio_tags(text: str) -> str:
+    """Remove v3 delivery tags from text destined for the ordinary voice."""
+    return _AUDIO_TAG_RE.sub(" ", str(text or ""))
+
+
 def for_speech(text: str) -> str:
     """Plain prose for the TTS engine: no markdown, no emoji, no bare URLs,
     no stray symbols that would be read out character by character."""
-    text = (text or "").strip()
+    text = strip_audio_tags((text or "").strip())
     if not text:
         return ""
     for symbol, spoken in _PRE_SPOKEN_SYMBOLS.items():
@@ -136,8 +160,11 @@ def is_question(text: str) -> bool:
 
 def for_display(text: str) -> str:
     """What the speech bubble shows: markdown syntax removed (the bubble
-    can't render it) but emoji and layout-ish punctuation left alone."""
+    can't render it) but emoji and layout-ish punctuation left alone.
-    text = (text or "").strip()
+
+    Delivery tags go too — the voice no longer says them, so showing them
+    would caption a laugh nobody heard."""
+    text = strip_audio_tags((text or "").strip())
     if not text:
         return ""
     text = _strip_markdown(text, keep_emoji=True)

bolt_pet/ui/app.py

@@ -41,6 +41,8 @@ def run() -> int:
     thread.started.connect(controller.run)
     controller.state_changed.connect(lambda value: window.set_state(PetState(value)))
     controller.said.connect(window.say)
+    # Lip-sync: the loudness of the audio actually going to the speakers.
+    controller.mouth.connect(window.set_mouth)
     controller.log.connect(_log)
     controller.action.connect(window.apply_action)  # petctl move/emote/say/...
     controller.finished.connect(thread.quit)

bolt_pet/ui/pet_window.py

@@ -19,7 +19,7 @@ from PySide6.QtGui import (
 )
 from PySide6.QtWidgets import QApplication, QWidget
 
-from .. import config
+from .. import config, window_state
 from ..monitors import Monitor
 from ..state import PetState
 from .sprite import WALK, SpriteSet
@@ -36,6 +36,11 @@ _NAP_OPACITY = 0.35
 # and the feet skate whenever PET_WANDER_SPEED doesn't happen to match the fps.
 # Eight frames at 13px is a ~104px stride cycle, a bit under the pet's width.
 _WALK_PIXELS_PER_FRAME = 13.0
+# How long a loudness level stays believable. The audio thread sends one per
+# ~30ms while a clip plays; if they stop arriving (offline TTS has no envelope,
+# or playback died) the mouth must not stay frozen mid-syllable, so after this
+# long the ordinary looping animation takes back over.
+_MOUTH_STALE_SECONDS = 0.35
 
 
 def emote_transform(emote: str, progress: float) -> tuple[float, float, float, float]:
@@ -213,6 +218,9 @@ class PetWindow(QWidget):
         self._commanded_move = False  # a petctl move — happens even mid-conversation
         self._next_wander_at = 0.0
         self._bob_offset = 0
+        # Lip-sync: loudness of what is playing right now, and when it arrived.
+        self._mouth_level: Optional[float] = None
+        self._mouth_at = 0.0
         self._bob_phase = 0.0
         self._walking = False
         self._facing = 1  # +1 right, -1 left; the walk art is drawn facing right
@@ -249,6 +257,19 @@ class PetWindow(QWidget):
             y = int(config.PET_START_Y) if config.PET_START_Y else None
         except ValueError:
             x = y = None
+        if x is None and y is None and config.PET_REMEMBER_POSITION:
+            remembered = window_state.load()
+            # Only if it still lands on a screen that exists — the usual reason
+            # a saved position is stale is that the monitor it was on has been
+            # unplugged, and restoring it faithfully would hide the pet.
+            if remembered is not None:
+                rectangles = [
+                    (s.availableGeometry().left(), s.availableGeometry().top(),
+                     s.availableGeometry().right(), s.availableGeometry().bottom())
+                    for s in QApplication.screens()
+                ]
+                if window_state.is_visible_on(*remembered, self.width(), rectangles):
+                    x, y = remembered
         if geo is not None:
             x = geo.right() - self.width() - 40 if x is None else x
             y = geo.bottom() - self.height() - 60 if y is None else y
@@ -688,6 +709,34 @@ class PetWindow(QWidget):
 
     # ── animation ────────────────────────────────────────────────────────
 
+    def set_mouth(self, level: float) -> None:
+        """How loud the pet is *right now* (0..1), straight off the PCM going
+        to the speakers (audio/tts.level_of).
+
+        The talking frames are ordered by mouth openness, so this indexes them
+        directly: the mouth moves with the actual waveform instead of flapping
+        on a timer, which is the difference between a talking sprite and a
+        dubbed one."""
+        self._mouth_level = max(0.0, min(1.0, float(level)))
+        self._mouth_at = time.monotonic()
+        if self._current_state == PetState.TALKING:
+            self.update()
+
+    def _mouth_frame(self, animation) -> Optional[QPixmap]:
+        """The frame matching the current loudness, or None to use the timer.
+
+        Falls back the moment the levels go stale — offline TTS has no
+        envelope, and a mouth frozen mid-syllable is worse than a timed loop."""
+        if self._mouth_level is None or animation is None:
+            return None
+        if time.monotonic() - self._mouth_at > _MOUTH_STALE_SECONDS:
+            return None
+        frames = animation.frames
+        if len(frames) < 2:
+            return None
+        index = int(round(self._mouth_level * (len(frames) - 1)))
+        return frames[max(0, min(len(frames) - 1, index))]
+
     def _advance_frame(self) -> None:
         # While walking the cycle is stepped by _advance_walk from distance
         # travelled; letting this timer also advance it would double-step it
@@ -702,7 +751,12 @@ class PetWindow(QWidget):
         painter.setRenderHint(QPainter.Antialiasing)
         painter.setRenderHint(QPainter.SmoothPixmapTransform)
         key = self._animation_key()
-        pixmap: Optional[QPixmap] = self.sprites.get(key).current()
+        animation = self.sprites.get(key)
+        pixmap: Optional[QPixmap] = None
+        if key == PetState.TALKING:
+            pixmap = self._mouth_frame(animation)
+        if pixmap is None:
+            pixmap = animation.current()
         if key == WALK:
             pixmap = self._oriented(pixmap)
         if pixmap is None:
@@ -759,6 +813,9 @@ class PetWindow(QWidget):
         was_click = not self._dragged
         self._drag_offset = None
         self._press_pos = None
+        if not was_click and config.PET_REMEMBER_POSITION:
+            position = self.geometry().topLeft()
+            window_state.save(position.x(), position.y())
         if was_click:
             self.talk_requested.emit()
         else:

bolt_pet/window_state.py

@@ -0,0 +1,75 @@
+"""Where the pet was left, so it starts there next time.
+
+Listed in the README as a known limitation: drag it somewhere deliberate, and
+the next launch puts it back in the bottom-right corner. For something that
+lives on your desktop all day that is a small daily annoyance, and it is a
+config write on drag-end.
+
+Lives in the cache dir rather than the repo, next to the restart context: it
+is per-machine state about this install, not something that belongs in a git
+diff. Every function swallows its own errors — a corrupt or unwritable state
+file must never stop the pet from starting, it just means the default corner.
+
+Positions are validated against the *current* screen layout on load, because
+the common case for a stale position is exactly the case where it is
+dangerous: the pet was last on a monitor that is now unplugged, and restoring
+it faithfully would put it somewhere you cannot see or reach.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import os
+import tempfile
+from pathlib import Path
+from typing import Optional
+
+logger = logging.getLogger("bolt_pet.window_state")
+
+DEFAULT_PATH = Path.home() / ".cache" / "bolt-pet" / "window.json"
+
+
+def load(path: Optional[Path] = None) -> Optional[tuple[int, int]]:
+    """The saved position, or None if there isn't a usable one."""
+    try:
+        data = json.loads(Path(path or DEFAULT_PATH).read_text(encoding="utf-8"))
+        return int(data["x"]), int(data["y"])
+    except (FileNotFoundError, json.JSONDecodeError, KeyError, TypeError, ValueError, OSError):
+        return None
+
+
+def save(x: int, y: int, path: Optional[Path] = None) -> None:
+    """Remember where it is now. Atomic, so a crash mid-write can't leave a
+    half-file that makes the next start fall back to the corner."""
+    target = Path(path or DEFAULT_PATH)
+    try:
+        target.parent.mkdir(parents=True, exist_ok=True)
+        descriptor, temp_path = tempfile.mkstemp(dir=target.parent, prefix=".window_", suffix=".tmp")
+        try:
+            with os.fdopen(descriptor, "w", encoding="utf-8") as handle:
+                json.dump({"x": int(x), "y": int(y)}, handle)
+            os.replace(temp_path, target)
+        except BaseException:
+            try:
+                os.unlink(temp_path)
+            except OSError:
+                pass
+            raise
+    except Exception:
+        logger.debug("Could not save the window position", exc_info=True)
+
+
+def is_visible_on(x: int, y: int, size: int, rectangles) -> bool:
+    """Whether that position still lands on a screen that exists.
+
+    *rectangles* are (left, top, right, bottom) tuples — the caller's job,
+    because this module has no business importing Qt. Requires a real overlap
+    rather than a touching edge, so a pet saved flush against the boundary of a
+    monitor that has since been unplugged is not counted as reachable."""
+    for left, top, right, bottom in rectangles:
+        overlap_x = min(x + size, right) - max(x, left)
+        overlap_y = min(y + size, bottom) - max(y, top)
+        if overlap_x > size * 0.25 and overlap_y > size * 0.25:
+            return True
+    return False

requirements.txt

@@ -8,6 +8,9 @@ numpy>=1.24
 
 # HTTP client to the Bolt desk API
 requests>=2.31
+# Streaming speech-to-text (audio/stt_stream.py). Optional in practice: without
+# it the pet falls back to uploading the finished clip, exactly as before.
+websocket-client>=1.7
 
 # Wake-word detection (local, offline after first run) — runs the
 # custom-trained thunderbolt.onnx model shipped in this repo, same runtime

scripts/generate_bolt_sprites.py

@@ -663,31 +663,43 @@ def render_frame(p) -> Image.Image:
 
 def frames_for(state: str) -> list[dict]:
     if state == "idle":
+        # Eight frames, all distinct. The old version drove breathing on
+        # sin(2*pi*t) and the tail on sin(4*pi*t), which both cross zero at
+        # i=0 and i=4 — so frame 4 was byte-identical to frame 0 and the loop
+        # was really four frames stored twice.
         out = []
         for i in range(8):
             t = i / 8
-            br = math.sin(t * 2 * math.pi)
+            br = math.sin(t * 2 * math.pi + math.pi / 7)
             out.append(
                 default_pose(
                     breathe=br,
                     head_dy=-0.006 * br,
-                    tail=math.sin(t * 4 * math.pi),
+                    # Three-halves harmonic: never in phase with the breath, so
+                    # no two frames of the cycle can coincide.
+                    tail=math.sin(t * 3 * math.pi + 0.6),
+                    ear_twitch=0.30 if i == 3 else 0.0,   # a flick, once a loop
                     blink=1.0 if i == 6 else 0.0,
                 )
             )
         return out
     if state == "listening":
+        # Six frames of *orienting*, not idling: ears up, head turning toward
+        # whoever is talking, then settling. The old four had frames 0 and 2
+        # differing by 0.09 mean pixels — a two-pose animation wearing four.
         out = []
-        for i in range(4):
+        for i in range(6):
-            t = i / 4
+            t = i / 6
+            lean = math.sin(t * 2 * math.pi + math.pi / 5)
             out.append(
                 default_pose(
                     ear=1.0,
-                    ear_twitch=0.35 * math.sin(t * 2 * math.pi),
+                    ear_twitch=0.45 * math.sin(t * 4 * math.pi),
-                    tilt=-7 + 2.0 * math.sin(t * 2 * math.pi),
+                    tilt=-9 + 4.0 * lean,
+                    look=(0.010 * lean, -0.004),
                     brow=1.0,
-                    tail=0.5 * math.sin(t * 2 * math.pi),
+                    tail=0.7 * math.sin(t * 2 * math.pi + 1.1),
-                    head_dy=-0.008,
+                    head_dy=-0.010 - 0.004 * lean,
                     tag_glow=True,
                     extras="listen",
                     phase=i,
@@ -695,35 +707,51 @@ def frames_for(state: str) -> list[dict]:
             )
         return out
     if state == "thinking":
+        # The old six moved by a mean of ~1.3 pixels — effectively a still
+        # image. Thinking should *look* like thinking: the head tilts, the eyes
+        # travel as if following a thought, and one ear rotates independently.
         out = []
         for i in range(6):
             t = i / 6
+            sway = math.sin(t * 2 * math.pi)
             out.append(
                 default_pose(
-                    ear=0.25,
+                    ear=0.25 + 0.35 * abs(sway),
-                    tilt=6.0,
+                    ear_twitch=0.5 * math.cos(t * 2 * math.pi),
-                    look=(0.022, -0.026),
+                    tilt=4.0 + 7.0 * sway,
-                    brow=0.5,
+                    # Eyes wander a small circle: the cheapest possible read of
-                    breathe=0.4 * math.sin(t * 2 * math.pi),
+                    # "working something out" and the thing most obviously
-                    tail=0.2 * math.sin(t * 2 * math.pi),
+                    # missing before.
+                    look=(0.026 * math.cos(t * 2 * math.pi),
+                          -0.020 + 0.014 * math.sin(t * 2 * math.pi)),
+                    brow=0.5 + 0.4 * abs(sway),
+                    breathe=0.5 * math.sin(t * 2 * math.pi + 0.9),
+                    head_dy=-0.008 * sway,
+                    tail=0.35 * math.sin(t * 3 * math.pi),
+                    blink=1.0 if i == 4 else 0.0,
                     extras="think",
                     phase=i // 2,
                 )
             )
         return out
     if state == "talking":
+        # Ordered by mouth openness — closed at frame 0, widest at the last —
+        # because the window indexes these by the loudness of the audio that is
+        # actually playing (see audio/tts.level_of and PetWindow.set_mouth).
+        # A time-ordered loop cannot be indexed that way, and a mouth that
+        # flaps on a timer is what makes a talking sprite look dubbed.
         out = []
-        for i in range(4):
+        count = 6
-            t = i / 4
+        for i in range(count):
-            open_ = (math.sin(t * 2 * math.pi) + 1) / 2
+            open_ = i / (count - 1)
             out.append(
                 default_pose(
-                    mouth=0.25 + 0.75 * open_,
+                    mouth=0.06 + 0.94 * open_,
                     ear=0.6,
                     head_dy=-0.010 * open_,
                     breathe=open_,
-                    tail=math.sin(t * 2 * math.pi + 1.0),
+                    tail=0.45 * math.sin(i * 0.9),
-                    brow=0.35,
+                    brow=0.35 * open_,
                 )
             )
         return out

tests/test_controller.py

@@ -22,6 +22,14 @@ def no_screen_probes(monkeypatch):
     monkeypatch.setattr(controller_mod.screen_context, "is_fullscreen_active", lambda: False)
 
 
+@pytest.fixture(autouse=True)
+def _no_streaming(monkeypatch):
+    """Most tests drive the plain request/response path; leaving streaming on
+    would have them attempt a real HTTP call, fail, and fall back — passing,
+    slowly, for the wrong reason. The streaming path has its own tests."""
+    monkeypatch.setattr(controller_mod.config, "STREAMING_REPLIES", False)
+
+
 @pytest.fixture
 def ctrl():
     return controller_mod.PetController()
@@ -41,10 +49,10 @@ def test_full_turn_happy_path(monkeypatch, ctrl):
 
     monkeypatch.setattr(controller_mod.mic, "record_utterance", lambda *a, **k: np.zeros(10, dtype=np.int16))
     monkeypatch.setattr(controller_mod.stt, "transcribe", lambda pcm: "what's the weather")
-    monkeypatch.setattr(controller_mod.server_client, "converse", lambda text, on_command=None: controller_mod.server_client.Reply("sunny and 72"))
+    monkeypatch.setattr(controller_mod.server_client, "converse", lambda text, on_command=None, on_say=None: controller_mod.server_client.Reply("sunny and 72"))
     spoken = []
     monkeypatch.setattr(controller_mod.tts, "speak",
-                        lambda text, on_error=None, should_stop=None, voice_id=None: (spoken.append(text), True)[1])
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: (spoken.append(text), True)[1])
 
     ctrl._handle_conversation_turn()
 
@@ -59,7 +67,7 @@ def test_turn_with_nothing_heard_returns_to_idle_without_calling_server(monkeypa
     monkeypatch.setattr(controller_mod.mic, "record_utterance", lambda *a, **k: None)
     called = {"n": 0}
     monkeypatch.setattr(controller_mod.server_client, "converse",
-                        lambda text, on_command=None: called.__setitem__("n", called["n"] + 1))
+                        lambda text, on_command=None, on_say=None: called.__setitem__("n", called["n"] + 1))
 
     ctrl._handle_conversation_turn()
 
@@ -97,7 +105,7 @@ def test_turn_with_server_error_flashes_error_then_idle(monkeypatch, ctrl):
     monkeypatch.setattr(controller_mod.mic, "record_utterance", lambda *a, **k: np.zeros(10, dtype=np.int16))
     monkeypatch.setattr(controller_mod.stt, "transcribe", lambda pcm: "hello")
 
-    def boom(text, on_command=None):
+    def boom(text, on_command=None, on_say=None):
         raise controller_mod.server_client.ServerError("server is down")
     monkeypatch.setattr(controller_mod.server_client, "converse", boom)
 
@@ -150,7 +158,7 @@ def test_heartbeat_speaks_a_pending_announcement_when_idle(monkeypatch, ctrl):
     monkeypatch.setattr(controller_mod.server_client, "report_status", lambda: "don't forget your 3pm")
     spoken = []
     monkeypatch.setattr(controller_mod.tts, "speak",
-                        lambda text, on_error=None, should_stop=None, voice_id=None: (spoken.append(text), True)[1])
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: (spoken.append(text), True)[1])
     said = _capture(ctrl.said)
 
     ctrl._maybe_heartbeat()

tests/test_controller_features.py

@@ -29,6 +29,14 @@ def no_screen_probes(monkeypatch):
     monkeypatch.setattr(controller_mod.screen_context, "context_for", lambda text: text)
 
 
+@pytest.fixture(autouse=True)
+def _no_streaming(monkeypatch):
+    """Most tests drive the plain request/response path; leaving streaming on
+    would have them attempt a real HTTP call, fail, and fall back — passing,
+    slowly, for the wrong reason. The streaming path has its own tests."""
+    monkeypatch.setattr(controller_mod.config, "STREAMING_REPLIES", False)
+
+
 @pytest.fixture
 def ctrl():
     return controller_mod.PetController()
@@ -160,7 +168,7 @@ def test_the_interrupt_log_reports_what_fired_not_the_reset_counters(monkeypatch
     ctrl._barge_in = detector
     logs = _capture(ctrl.log)
 
-    def interrupted_playback(text, on_error=None, should_stop=None, voice_id=None):
+    def interrupted_playback(text, on_error=None, should_stop=None, voice_id=None, on_level=None):
         # What really happens: frames get scored during playback, then one
         # clears the threshold and playback aborts.
         detector._frames, detector._peak, detector._last = 7, 0.81, 0.81
@@ -180,7 +188,7 @@ def test_the_interrupt_log_reports_what_fired_not_the_reset_counters(monkeypatch
 def test_interrupted_playback_queues_an_immediate_next_turn(monkeypatch, ctrl):
     logs = _capture(ctrl.log)
     monkeypatch.setattr(controller_mod.tts, "speak",
-                        lambda text, on_error=None, should_stop=None, voice_id=None: False)  # interrupted
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: False)  # interrupted
 
     ctrl._speak("a very long explanation")
 
@@ -190,7 +198,7 @@ def test_interrupted_playback_queues_an_immediate_next_turn(monkeypatch, ctrl):
 
 def test_uninterrupted_playback_does_not_queue_a_turn(monkeypatch, ctrl):
     monkeypatch.setattr(controller_mod.tts, "speak",
-                        lambda text, on_error=None, should_stop=None, voice_id=None: True)
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: True)
     ctrl._speak("short answer")
     assert not ctrl._talk_now.is_set()
 
@@ -202,7 +210,7 @@ def spoke(monkeypatch):
     """Playback that always completes, so only the follow-up rule decides
     whether another turn is queued."""
     monkeypatch.setattr(controller_mod.tts, "speak",
-                        lambda text, on_error=None, should_stop=None, voice_id=None: True)
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: True)
 
 
 def test_a_reply_ending_in_a_question_keeps_listening(spoke, ctrl):
@@ -290,7 +298,7 @@ def test_follow_up_can_be_turned_off(spoke, monkeypatch, ctrl):
 
 def test_being_interrupted_restarts_the_chain(monkeypatch, ctrl):
     monkeypatch.setattr(controller_mod.tts, "speak",
-                        lambda text, on_error=None, should_stop=None, voice_id=None: False)
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: False)
     ctrl._follow_ups = 3
 
     ctrl._speak("a very long explanation")
@@ -301,7 +309,7 @@ def test_being_interrupted_restarts_the_chain(monkeypatch, ctrl):
 
 def test_speech_is_recorded_in_the_history(monkeypatch, ctrl):
     monkeypatch.setattr(controller_mod.tts, "speak",
-                        lambda text, on_error=None, should_stop=None, voice_id=None: True)
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: True)
     ctrl._speak("**bold** reply")
     assert ctrl.history.last().text == "**bold** reply"  # raw, for copy/paste
 
@@ -315,10 +323,10 @@ def test_the_active_window_rides_along_with_the_utterance(monkeypatch, ctrl):
     monkeypatch.setattr(controller_mod.screen_context, "context_for",
                         lambda text: f"{text}\n\n[on screen right now: app.py]")
     monkeypatch.setattr(controller_mod.tts, "speak",
-                        lambda text, on_error=None, should_stop=None, voice_id=None: True)
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: True)
     sent = []
     monkeypatch.setattr(controller_mod.server_client, "converse",
-                        lambda text, on_command=None: sent.append(text) or Reply("that's a KeyError"))
+                        lambda text, on_command=None, on_say=None: sent.append(text) or Reply("that's a KeyError"))
 
     ctrl._handle_conversation_turn()
 
@@ -371,10 +379,10 @@ def test_napping_still_answers_when_spoken_to(monkeypatch, ctrl):
                         lambda *a, **k: np.zeros(10, dtype=np.int16))
     monkeypatch.setattr(controller_mod.stt, "transcribe", lambda pcm: "you awake?")
     monkeypatch.setattr(controller_mod.server_client, "converse",
-                        lambda text, on_command=None: Reply("always"))
+                        lambda text, on_command=None, on_say=None: Reply("always"))
     spoken = []
     monkeypatch.setattr(controller_mod.tts, "speak",
-                        lambda text, on_error=None, should_stop=None, voice_id=None: spoken.append(text) or True)
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: spoken.append(text) or True)
     ctrl.set_napping(True)
 
     ctrl._handle_conversation_turn()
@@ -389,9 +397,9 @@ def test_notifications_are_forwarded_and_spoken(monkeypatch, ctrl):
     ctrl._notification_gate = controller_mod.notifications.NotificationGate("", 0)
     sent, spoken = [], []
     monkeypatch.setattr(controller_mod.server_client, "converse",
-                        lambda text, on_command=None: sent.append(text) or Reply("your build is green"))
+                        lambda text, on_command=None, on_say=None: sent.append(text) or Reply("your build is green"))
     monkeypatch.setattr(controller_mod.tts, "speak",
-                        lambda text, on_error=None, should_stop=None, voice_id=None: spoken.append(text) or True)
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: spoken.append(text) or True)
 
     ctrl._queue_notification(Notification(app="CI", summary="Build finished", body=""))
     ctrl._drain_notifications()
@@ -411,7 +419,7 @@ def test_notifications_are_not_forwarded_while_napping(monkeypatch, ctrl):
     ctrl._notification_gate = controller_mod.notifications.NotificationGate("", 0)
     called = {"n": 0}
     monkeypatch.setattr(controller_mod.server_client, "converse",
-                        lambda text, on_command=None: called.__setitem__("n", called["n"] + 1))
+                        lambda text, on_command=None, on_say=None: called.__setitem__("n", called["n"] + 1))
     ctrl.set_napping(True)
 
     ctrl._queue_notification(Notification(app="CI", summary="Build finished", body=""))
@@ -507,9 +515,9 @@ def test_check_deliveries_runs_after_a_conversation_turn(monkeypatch, ctrl, tmp_
                         lambda *a, **k: np.zeros(10, dtype=np.int16))
     monkeypatch.setattr(controller_mod.stt, "transcribe", lambda pcm: "send me that file")
     monkeypatch.setattr(controller_mod.server_client, "converse",
-                        lambda text, on_command=None: Reply("it's on the way"))
+                        lambda text, on_command=None, on_say=None: Reply("it's on the way"))
     monkeypatch.setattr(controller_mod.tts, "speak",
-                        lambda text, on_error=None, should_stop=None, voice_id=None: True)
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: True)
     monkeypatch.setattr(controller_mod.config, "DELIVERED_FILES_DIR", tmp_path)
     monkeypatch.setattr(controller_mod.server_client, "list_outbox_files",
                         lambda: [{"id": "abc", "name": "notes.txt", "size": 2}])
@@ -531,10 +539,10 @@ def _voice_turn(monkeypatch, ctrl, reply, said="talk like a pirate"):
                         lambda *a, **k: np.zeros(10, dtype=np.int16))
     monkeypatch.setattr(controller_mod.stt, "transcribe", lambda pcm: said)
     monkeypatch.setattr(controller_mod.server_client, "converse",
-                        lambda text, on_command=None: reply)
+                        lambda text, on_command=None, on_say=None: reply)
     monkeypatch.setattr(
         controller_mod.tts, "speak",
-        lambda text, on_error=None, should_stop=None, voice_id=None:
+        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None:
             voices.append(voice_id) or True,
     )
     ctrl._handle_conversation_turn()
@@ -625,7 +633,7 @@ def test_dialoguectl_never_reaches_the_shell(monkeypatch, ctrl):
     monkeypatch.setattr(controller_mod.tts, "synthesize_dialogue",
                         lambda inputs, model_id=None, stability=None: (np.zeros(4, dtype=np.int16), 24000))
     monkeypatch.setattr(controller_mod.tts, "play_pcm",
-                        lambda pcm, rate, should_stop=None: True)
+                        lambda pcm, rate, should_stop=None, on_level=None: True)
 
     output = ctrl._handle_command(_dialogue_command({"voice": "self", "text": "[cheerfully] hi"}))
 
@@ -638,7 +646,7 @@ def test_a_scene_shows_in_the_bubble_with_the_delivery_tags_stripped(monkeypatch
     monkeypatch.setattr(controller_mod.config, "DIALOGUE_VOICES", "narrator:9BWtsMINqrJLrRacOk9x")
     monkeypatch.setattr(controller_mod.tts, "synthesize_dialogue",
                         lambda inputs, model_id=None, stability=None: (np.zeros(4, dtype=np.int16), 24000))
-    monkeypatch.setattr(controller_mod.tts, "play_pcm", lambda pcm, rate, should_stop=None: True)
+    monkeypatch.setattr(controller_mod.tts, "play_pcm", lambda pcm, rate, should_stop=None, on_level=None: True)
     said = _capture(ctrl.said)
 
     ctrl._handle_command(_dialogue_command(
@@ -656,7 +664,7 @@ def test_a_mid_turn_scene_returns_to_thinking_not_idle(monkeypatch, ctrl):
     monkeypatch.setattr(controller_mod.config, "ELEVENLABS_VOICE_ID", "aaorr6ZHIL88gEexu7dC")
     monkeypatch.setattr(controller_mod.tts, "synthesize_dialogue",
                         lambda inputs, model_id=None, stability=None: (np.zeros(4, dtype=np.int16), 24000))
-    monkeypatch.setattr(controller_mod.tts, "play_pcm", lambda pcm, rate, should_stop=None: True)
+    monkeypatch.setattr(controller_mod.tts, "play_pcm", lambda pcm, rate, should_stop=None, on_level=None: True)
     ctrl._state.transition(PetState.LISTENING)
     ctrl._state.transition(PetState.THINKING)
     states = _capture(ctrl.state_changed)
@@ -676,7 +684,7 @@ def test_the_scene_uses_a_voice_the_server_picked_with_speak_as(monkeypatch, ctr
         return np.zeros(4, dtype=np.int16), 24000
 
     monkeypatch.setattr(controller_mod.tts, "synthesize_dialogue", capture)
-    monkeypatch.setattr(controller_mod.tts, "play_pcm", lambda pcm, rate, should_stop=None: True)
+    monkeypatch.setattr(controller_mod.tts, "play_pcm", lambda pcm, rate, should_stop=None, on_level=None: True)
     ctrl._apply_voice(controller_mod.server_client.Reply("ok", "PICKEDvoice123456789", "Terence"))
 
     ctrl._handle_command(_dialogue_command({"voice": "self", "text": "hi"}))
@@ -718,7 +726,7 @@ def test_talking_over_a_scene_is_reported_up_the_relay(monkeypatch, ctrl):
     monkeypatch.setattr(controller_mod.tts, "synthesize_dialogue",
                         lambda inputs, model_id=None, stability=None: (np.zeros(4, dtype=np.int16), 24000))
     monkeypatch.setattr(controller_mod.tts, "play_pcm",
-                        lambda pcm, rate, should_stop=None: False)  # barge-in
+                        lambda pcm, rate, should_stop=None, on_level=None: False)  # barge-in
 
     output = ctrl._handle_command(_dialogue_command({"voice": "self", "text": "hi"}))
 
@@ -788,9 +796,9 @@ def test_coming_back_up_reports_to_the_server_and_speaks_the_reply(monkeypatch,
                                     path=state, now=1000.0)
     sent, spoken = [], []
     monkeypatch.setattr(controller_mod.server_client, "converse",
-                        lambda text, on_command=None: sent.append(text) or Reply("Good, it's up."))
+                        lambda text, on_command=None, on_say=None: sent.append(text) or Reply("Good, it's up."))
     monkeypatch.setattr(controller_mod.tts, "speak",
-                        lambda text, on_error=None, should_stop=None, voice_id=None:
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None:
                             spoken.append(text) or True)
 
     ctrl._report_self_restart()
@@ -805,8 +813,230 @@ def test_an_ordinary_start_reports_nothing(monkeypatch, ctrl, tmp_path):
     monkeypatch.setattr(controller_mod.self_restart, "DEFAULT_STATE_PATH", tmp_path / "none.json")
     called = []
     monkeypatch.setattr(controller_mod.server_client, "converse",
-                        lambda text, on_command=None: called.append(text))
+                        lambda text, on_command=None, on_say=None: called.append(text))
 
     ctrl._report_self_restart()
 
     assert called == []
+
+
+# ── holding line: "give me a sec" while a tool runs ────────────────────────
+# The server used to discard whatever the model wrote alongside a tool call,
+# so the whole round trip was silence — and the model, with no evidence its
+# sentence landed, said it again in the final reply.
+
+def test_a_holding_line_is_spoken_before_the_tool_runs(monkeypatch, ctrl):
+    order = []
+    monkeypatch.setattr(controller_mod.tts, "speak",
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None:
+                            order.append(("spoke", text)) or True)
+    monkeypatch.setattr(controller_mod.server_client, "run_local_command",
+                        lambda cmd: order.append(("ran", cmd)) or "[exit 0]")
+
+    ctrl._speak_holding("Give me a sec.")
+    ctrl._handle_command("df -h /")
+
+    assert order == [("spoke", "Give me a sec."), ("ran", "df -h /")]
+
+
+def test_the_holding_line_shows_in_the_bubble_but_not_the_transcript(monkeypatch, ctrl):
+    """It's filler. The transcript should keep the actual answer."""
+    monkeypatch.setattr(controller_mod.tts, "speak",
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: True)
+    said = _capture(ctrl.said)
+
+    ctrl._speak_holding("I'll check on that — one moment.")
+
+    assert said == ["I'll check on that — one moment."]
+    assert ctrl.history.entries() == []
+
+
+def test_a_holding_line_returns_to_waiting_not_idle(monkeypatch, ctrl):
+    monkeypatch.setattr(controller_mod.tts, "speak",
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: True)
+    ctrl._state.transition(PetState.LISTENING)
+    ctrl._state.transition(PetState.THINKING)
+    states = _capture(ctrl.state_changed)
+
+    ctrl._speak_holding("one sec")
+
+    assert states == ["talking", "thinking"]
+
+
+def test_the_relay_speaks_whatever_the_server_attaches(monkeypatch, ctrl):
+    spoken = []
+    monkeypatch.setattr(controller_mod.tts, "speak",
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None:
+                            spoken.append(text) or True)
+    monkeypatch.setattr(controller_mod.mic, "record_utterance",
+                        lambda *a, **k: np.zeros(10, dtype=np.int16))
+    monkeypatch.setattr(controller_mod.stt, "transcribe", lambda pcm: "what's in that folder?")
+
+    def fake_converse(text, on_command=None, on_say=None):
+        on_say("Let me check that for you.")      # what the server now sends
+        on_command("filectl {\"op\": \"list\", \"path\": \"/tmp\"}")
+        return Reply("It's got three files in it.")
+
+    monkeypatch.setattr(controller_mod.server_client, "converse", fake_converse)
+
+    ctrl._handle_conversation_turn()
+
+    assert spoken == ["Let me check that for you.", "It's got three files in it."]
+
+
+# ── streamed replies ───────────────────────────────────────────────────────
+# Without streaming the pet waits out the entire model call before saying a
+# word. With it, the wait is time-to-first-sentence.
+
+def test_each_sentence_is_spoken_as_it_arrives(monkeypatch, ctrl):
+    monkeypatch.setattr(controller_mod.config, "STREAMING_REPLIES", True)
+    spoken = []
+    monkeypatch.setattr(controller_mod.tts, "speak",
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None:
+                            spoken.append(text) or True)
+
+    def fake_stream(text, on_say, on_command=None, timeout=180.0):
+        on_say("The disk is fine.")
+        on_say("About sixty percent used.")
+        return controller_mod.server_client.Reply(
+            "The disk is fine. About sixty percent used.", spoken=True)
+
+    monkeypatch.setattr(controller_mod.server_client, "converse_stream", fake_stream)
+    monkeypatch.setattr(controller_mod.mic, "record_utterance",
+                        lambda *a, **k: np.zeros(10, dtype=np.int16))
+    monkeypatch.setattr(controller_mod.stt, "transcribe", lambda pcm: "how's the disk?")
+
+    ctrl._handle_conversation_turn()
+
+    assert spoken == ["The disk is fine.", "About sixty percent used."]
+    # ...and the empty final reply is not spoken as an extra blank utterance.
+    assert ctrl.history.entries()[-1].text == "About sixty percent used."
+
+
+def test_a_stream_that_fails_before_speaking_falls_back_silently(monkeypatch, ctrl):
+    monkeypatch.setattr(controller_mod.config, "STREAMING_REPLIES", True)
+    spoken = []
+    monkeypatch.setattr(controller_mod.tts, "speak",
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None:
+                            spoken.append(text) or True)
+
+    def broken_stream(text, on_say, on_command=None, timeout=180.0):
+        raise controller_mod.server_client.ServerError("no streaming endpoint")
+
+    monkeypatch.setattr(controller_mod.server_client, "converse_stream", broken_stream)
+    monkeypatch.setattr(controller_mod.server_client, "converse",
+                        lambda text, on_command=None, on_say=None: Reply("Fell back fine."))
+
+    assert ctrl._ask_server("hello").text == "Fell back fine."
+    assert spoken == []   # nothing was said twice
+
+
+def test_streaming_can_be_switched_off(monkeypatch, ctrl):
+    monkeypatch.setattr(controller_mod.config, "STREAMING_REPLIES", False)
+    called = []
+    monkeypatch.setattr(controller_mod.server_client, "converse_stream",
+                        lambda *a, **k: called.append(1))
+    monkeypatch.setattr(controller_mod.server_client, "converse",
+                        lambda text, on_command=None, on_say=None: Reply("plain path"))
+
+    assert ctrl._ask_server("hi").text == "plain path"
+    assert called == []
+
+
+def test_talking_over_a_streamed_reply_still_interrupts(monkeypatch, ctrl):
+    monkeypatch.setattr(controller_mod.config, "STREAMING_REPLIES", True)
+    monkeypatch.setattr(controller_mod.tts, "speak",
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: False)
+
+    ctrl._speak_stream_chunk("A long explanation you cut short.")
+
+    assert ctrl._talk_now.is_set()
+
+
+# ── notification latency (reported 2026-08-02: 5-10 minutes) ───────────────
+# Draining was wired to the heartbeat's 60s interval, and a heartbeat that
+# landed mid-conversation stamped its clock before noticing — so it burned the
+# slot and waited another full interval. Several of those in a row is minutes.
+
+def test_a_notification_goes_out_on_the_next_tick_not_the_next_heartbeat(monkeypatch, ctrl):
+    monkeypatch.setattr(controller_mod.config, "NOTIFICATION_MIN_INTERVAL_SECONDS", 0)
+    ctrl._notification_gate = controller_mod.notifications.NotificationGate("", 0)
+    sent = []
+    monkeypatch.setattr(controller_mod.server_client, "converse",
+                        lambda text, on_command=None, on_say=None:
+                            sent.append(text) or Reply("noted"))
+    monkeypatch.setattr(controller_mod.tts, "speak",
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: True)
+    monkeypatch.setattr(controller_mod.server_client, "report_status", lambda: None)
+    # A heartbeat has just run, so the next one is a full interval away.
+    ctrl._last_heartbeat = controller_mod.time.monotonic()
+
+    ctrl._queue_notification(Notification(app="Signal", summary="Harry", body="you there?"))
+    ctrl._maybe_heartbeat()
+
+    assert sent and "Harry" in sent[0]
+
+
+def test_a_heartbeat_skipped_mid_conversation_does_not_burn_its_slot(monkeypatch, ctrl):
+    monkeypatch.setattr(controller_mod.server_client, "report_status", lambda: None)
+    ctrl._last_heartbeat = 0.0
+    ctrl._state.transition(PetState.LISTENING)
+    ctrl._state.transition(PetState.THINKING)
+
+    ctrl._maybe_heartbeat()          # due, but the pet is busy
+    assert ctrl._last_heartbeat == 0.0, "the clock must not advance on a skipped tick"
+
+    ctrl._state.transition(PetState.IDLE)
+    ctrl._maybe_heartbeat()          # free now — runs immediately
+    assert ctrl._last_heartbeat > 0.0
+
+
+def test_notifications_wait_while_the_pet_is_mid_turn(monkeypatch, ctrl):
+    """Speaking over the answer they're already getting would be worse than
+    waiting a couple of seconds."""
+    ctrl._notification_gate = controller_mod.notifications.NotificationGate("", 0)
+    sent = []
+    monkeypatch.setattr(controller_mod.server_client, "converse",
+                        lambda text, on_command=None, on_say=None: sent.append(text))
+    ctrl._state.transition(PetState.LISTENING)
+
+    ctrl._queue_notification(Notification(app="CI", summary="Build finished", body=""))
+    ctrl._maybe_heartbeat()
+
+    assert sent == []
+    assert len(ctrl._pending_notifications) == 1   # kept, not dropped
+
+
+def test_a_second_message_inside_the_rate_limit_is_no_longer_lost(monkeypatch, ctrl):
+    """The gate used to drop it. With NOTIFICATION_MIN_INTERVAL_SECONDS=60,
+    two texts a minute apart meant you heard about one of them."""
+    ctrl._notification_gate = controller_mod.notifications.NotificationGate("", 60)
+    sent = []
+    monkeypatch.setattr(controller_mod.server_client, "converse",
+                        lambda text, on_command=None, on_say=None:
+                            sent.append(text) or Reply("ok"))
+    monkeypatch.setattr(controller_mod.tts, "speak",
+                        lambda text, on_error=None, should_stop=None, voice_id=None, on_level=None: True)
+
+    ctrl._queue_notification(Notification(app="Signal", summary="Harry", body="you there?"))
+    ctrl._queue_notification(Notification(app="Signal", summary="Harry", body="it's urgent"))
+    ctrl._drain_notifications()
+
+    assert len(sent) == 1                      # one round trip...
+    assert "you there?" in sent[0] and "it's urgent" in sent[0]   # ...both messages
+    assert "2 desktop notifications" in sent[0]
+
+
+def test_the_filter_still_applies(ctrl):
+    ctrl._notification_gate = controller_mod.notifications.NotificationGate("deploy", 0)
+    ctrl._queue_notification(Notification(app="Chat", summary="lunch?", body=""))
+    assert ctrl._pending_notifications == []
+
+
+def test_a_notification_storm_cannot_become_an_unbounded_backlog(ctrl):
+    ctrl._notification_gate = controller_mod.notifications.NotificationGate("", 0)
+    for index in range(40):
+        ctrl._queue_notification(Notification(app="Spam", summary=f"#{index}", body=""))
+
+    assert len(ctrl._pending_notifications) == controller_mod._MAX_PENDING_NOTIFICATIONS
+    assert "#39" in ctrl._pending_notifications[-1].as_text()   # newest kept

tests/test_dialogue.py

@@ -161,7 +161,18 @@ def test_the_relay_report_names_the_cast():
     action = dialogue.parse('dialoguectl ' + _scene(
         '{"voice": "self", "text": "one"}', '{"voice": "narrator", "text": "two"}',
     ))
-    assert dialogue.describe(action) == "[dialogue] played 2 lines in 2 voices: narrator, self"
+    assert dialogue.describe(action).startswith(
+        "[dialogue] played 2 lines in 2 voices: narrator, self")
+
+
+def test_the_report_says_the_scene_was_already_heard():
+    """Observed 2026-07-31: the scene played, then the final reply summarised
+    it, so the pet said the same thing twice with nothing in between. The
+    model cannot know the audio already happened unless it is told."""
+    action = dialogue.parse('dialoguectl {"lines": [{"text": "hello"}]}')
+    report = dialogue.describe(action)
+    assert "HEARD this already" in report
+    assert "Do not repeat" in report
 
 
 # ── the HTTP request ────────────────────────────────────────────────────────

tests/test_doctor.py

@@ -0,0 +1,115 @@
+"""The preflight. Its one job is to never be the thing that's broken.
+
+A doctor that raises on a broken install diagnoses the wrong patient, so the
+tests that matter here are the ugly-input ones: no config at all, a check that
+throws, a dependency missing. The individual diagnoses are simple enough to
+read; that they *run* on a machine missing everything is the property worth
+pinning down.
+"""
+
+import sys
+from pathlib import Path
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from bolt_pet import config, doctor
+from bolt_pet.doctor import FAIL, OK, WARN
+
+
+def test_every_check_returns_a_verdict_on_a_bare_machine(monkeypatch):
+    """Nothing configured, nothing installed — still a full report."""
+    for name in ("SERVER_URL", "API_KEY", "DEEPGRAM_API_KEY",
+                 "ELEVENLABS_API_KEY", "ELEVENLABS_VOICE_ID"):
+        monkeypatch.setattr(config, name, "")
+    monkeypatch.setattr(doctor, "_module", lambda _n: False)
+
+    results = doctor.run()
+
+    assert len(results) == len(doctor.CHECKS)
+    assert all(c.status in (OK, WARN, FAIL) for c in results)
+    assert all(c.name and c.detail for c in results)
+
+
+def test_a_check_that_raises_does_not_hide_the_others():
+    """One broken probe must not cost you the other eleven diagnoses."""
+    def explode(*_args):
+        raise RuntimeError("boom")
+
+    original = doctor.CHECKS
+    doctor.CHECKS = (("mic", explode),) + original[:2]
+    try:
+        results = doctor.run()
+    finally:
+        doctor.CHECKS = original
+
+    assert len(results) == 3
+    assert results[0].status == FAIL
+    assert "boom" in results[0].detail
+
+
+def test_missing_server_config_is_a_failure_not_a_warning(monkeypatch):
+    """Without it the controller exits its thread at startup — the pet looks
+    alive and simply never answers. That is the worst failure mode there is."""
+    monkeypatch.setattr(config, "missing_config", lambda: ["BOLT_SERVER_URL", "DESK_API_KEY"])
+    check = doctor.check_config()
+    assert check.status == FAIL
+    assert "BOLT_SERVER_URL" in check.detail
+    assert check.fix
+
+
+def test_a_configured_server_passes_without_being_contacted(monkeypatch):
+    """The shallow run must not need the network — it's the first thing you
+    reach for when the network is what's wrong."""
+    monkeypatch.setattr(config, "missing_config", lambda: [])
+    monkeypatch.setattr(config, "SERVER_URL", "http://bolt.local:8000")
+    assert doctor.check_config().status == OK
+    assert doctor.check_server(deep=False).status == OK
+
+
+def test_every_problem_comes_with_something_to_do_about_it(monkeypatch):
+    """"screen reading: warn" is useless on its own; "apt install tesseract-ocr"
+    is the entire point of the tool."""
+    for name in ("SERVER_URL", "API_KEY", "DEEPGRAM_API_KEY"):
+        monkeypatch.setattr(config, name, "")
+    monkeypatch.setattr(doctor, "_module", lambda _n: False)
+
+    for check in doctor.run():
+        if check.status == FAIL:
+            assert check.fix, f"{check.name} says what's wrong but not what to do"
+
+
+def test_the_exit_code_is_nonzero_only_for_real_failures(monkeypatch, capsys):
+    monkeypatch.setattr(doctor, "run", lambda deep=False: [
+        doctor.Check("a", OK, "fine"), doctor.Check("b", WARN, "degraded")])
+    assert doctor.main([]) == 0
+
+    monkeypatch.setattr(doctor, "run", lambda deep=False: [doctor.Check("a", FAIL, "broken")])
+    assert doctor.main([]) == 1
+    assert "Fix those first" in capsys.readouterr().out
+
+
+def test_deep_is_off_unless_asked(monkeypatch):
+    seen = []
+    monkeypatch.setattr(doctor, "run", lambda deep=False: seen.append(deep) or [])
+    doctor.main([])
+    doctor.main(["--deep"])
+    assert seen == [False, True]
+
+
+def test_a_slow_silence_timeout_is_flagged(monkeypatch):
+    """The setting most likely to make it feel sluggish, and the least obvious
+    — it is pure dead air before anything at all starts happening."""
+    monkeypatch.setattr(config, "SILENCE_END_SEC", 2.0)
+    check = doctor.check_latency()
+    assert check.status == WARN
+    assert "2s" in check.detail or "2 " in check.detail
+
+    monkeypatch.setattr(config, "SILENCE_END_SEC", 0.9)
+    assert doctor.check_latency().status == OK
+
+
+def test_a_check_line_renders_the_fix_only_when_there_is_a_problem():
+    assert "→" not in doctor.Check("x", OK, "all good", fix="unused").line()
+    assert "→ do the thing" in doctor.Check("x", WARN, "hmm", fix="do the thing").line()

tests/test_lipsync.py

@@ -0,0 +1,135 @@
+"""Amplitude → mouth openness, from PCM samples to the drawn frame.
+
+Two halves, tested separately because only one of them needs a display:
+`tts.level_of`/`envelope` turn samples into a 0..1 loudness, and
+`PetWindow._mouth_frame` turns that loudness into a frame index. They meet at
+the `mouth` signal, which the pipeline smoke test covers end to end.
+"""
+
+import sys
+import time
+from pathlib import Path
+
+import numpy as np
+import pytest
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from bolt_pet.audio import tts
+
+
+def frame(amplitude: int, samples: int = 480) -> np.ndarray:
+    return np.full(samples, amplitude, dtype=np.int16)
+
+
+# ── loudness ────────────────────────────────────────────────────────────────
+
+def test_silence_closes_the_mouth():
+    assert tts.level_of(np.zeros(480, dtype=np.int16)) == 0.0
+
+
+def test_a_loud_frame_opens_it_fully():
+    assert tts.level_of(frame(30000)) == 1.0
+
+
+def test_the_level_never_leaves_zero_to_one():
+    """It indexes a frame list; out of range is an IndexError on the UI thread."""
+    for amplitude in (0, 1, 500, 6000, 20000, 32767):
+        assert 0.0 <= tts.level_of(frame(amplitude)) <= 1.0
+
+
+def test_it_rises_with_amplitude():
+    quiet, middling, loud = (tts.level_of(frame(a)) for a in (800, 4000, 12000))
+    assert quiet < middling < loud
+
+
+def test_quiet_speech_still_moves_the_mouth_visibly():
+    """The sqrt curve is the whole point. Speech spends most of its time well
+    below peak, so a linear map leaves the mouth barely open for normal talking
+    and the pet looks like it's mumbling."""
+    assert tts.level_of(frame(1500)) > 0.15
+
+
+def test_an_empty_frame_is_silence_not_a_crash():
+    assert tts.level_of(np.zeros(0, dtype=np.int16)) == 0.0
+
+
+# ── the envelope of a whole clip ────────────────────────────────────────────
+
+def test_an_envelope_has_one_value_per_frame_at_the_requested_rate():
+    one_second = np.zeros(16000, dtype=np.int16)
+    assert len(tts.envelope(one_second, 16000, fps=30)) == pytest.approx(30, abs=1)
+
+
+def test_an_envelope_tracks_loud_and_quiet_stretches():
+    pcm = np.concatenate([np.zeros(8000, dtype=np.int16), frame(20000, 8000)])
+    levels = tts.envelope(pcm, 16000, fps=10)
+    assert max(levels[:4]) == 0.0      # the silent half
+    assert min(levels[-4:]) > 0.5      # the loud half
+
+
+def test_an_empty_clip_has_an_empty_envelope():
+    assert tts.envelope(np.zeros(0, dtype=np.int16), 16000) == []
+
+
+# ── the drawn frame ─────────────────────────────────────────────────────────
+
+@pytest.fixture
+def window():
+    """Needs a QApplication; run with QT_QPA_PLATFORM=offscreen."""
+    from PySide6.QtWidgets import QApplication
+
+    from bolt_pet.ui.pet_window import PetWindow
+
+    _app = QApplication.instance() or QApplication(["test"])
+    win = PetWindow()
+    yield win
+    win.close()
+
+
+class FakeAnimation:
+    """Stands in for sprite.Animation — _mouth_frame only wants `.frames`."""
+
+    def __init__(self, frames):
+        self.frames = list(frames)
+
+
+def test_the_talking_frame_follows_the_level(window):
+    """The talking frames are an openness ramp — closed first, widest last —
+    specifically so loudness can index them directly."""
+    animation = FakeAnimation(["closed", "a", "b", "c", "d", "open"])
+
+    assert window._mouth_frame(animation) is None      # no level set yet
+
+    window.set_mouth(0.0)
+    assert window._mouth_frame(animation) == "closed"
+
+    window.set_mouth(1.0)
+    assert window._mouth_frame(animation) == "open"
+
+    window.set_mouth(0.5)
+    assert window._mouth_frame(animation) not in ("closed", "open")
+
+
+def test_a_stale_level_gives_the_animation_back(window):
+    """If the audio thread stops sending levels — TTS died, the clip ended
+    without a final zero — the mouth must not freeze half-open forever. After a
+    moment it falls back to the ordinary looping animation."""
+    animation = FakeAnimation(["a", "b", "c"])
+    window.set_mouth(0.9)
+    assert window._mouth_frame(animation) is not None
+
+    window._mouth_at = time.monotonic() - 5.0
+    assert window._mouth_frame(animation) is None
+
+
+def test_a_single_frame_animation_falls_back_instead_of_indexing(window):
+    """Art with one talking frame can't lip-sync; it must not try."""
+    window.set_mouth(1.0)
+    assert window._mouth_frame(FakeAnimation(["only"])) is None
+    assert window._mouth_frame(FakeAnimation([])) is None
+
+
+def test_no_animation_at_all_is_handled(window):
+    window.set_mouth(0.5)
+    assert window._mouth_frame(None) is None

tests/test_pipeline_smoke.py

@@ -0,0 +1,289 @@
+"""End-to-end: microphone in, speech out, over real HTTP.
+
+Every other test in this suite injects a fake at the seam it cares about, and
+every one of them passed all week while these got through to production:
+
+  - notifications sitting unspoken for minutes (a clock stamped in the wrong
+    order, two correct units)
+  - the pet saying the same thing twice (server discarded prose, model repeated
+    it — both sides behaving as written)
+  - `[laughing]` read out loud (a tag that means something to one model and
+    nothing to the next one down the pipe)
+  - a device command written as prose (extractor fine, prompt fine, no marker)
+
+They were all *interaction* bugs. So this one runs the actual pipeline against
+a real socket: a threaded HTTP server that speaks the desk protocol, the real
+`server_client` doing real requests (including NDJSON streaming), the real
+controller loop and state machine. The only fakes are where the hardware is —
+the mic stream and the speakers — because those are the two things a test
+genuinely cannot have.
+"""
+
+import json
+import sys
+import threading
+from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
+from pathlib import Path
+
+import numpy as np
+import pytest
+from PySide6.QtWidgets import QApplication
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from bolt_pet import controller as controller_mod
+from bolt_pet.notifications import Notification
+from bolt_pet.state import PetState
+
+_app = QApplication.instance() or QApplication(["test"])
+
+
+# ── a desk server that actually listens on a port ───────────────────────────
+
+class FakeDesk:
+    """Scripted responses, real HTTP. Set `.script` per test."""
+
+    def __init__(self):
+        self.script = {}
+        self.requests = []
+        self._server = ThreadingHTTPServer(("127.0.0.1", 0), self._handler())
+        self._thread = threading.Thread(target=self._server.serve_forever, daemon=True)
+        self._thread.start()
+
+    @property
+    def url(self) -> str:
+        host, port = self._server.server_address[:2]
+        return f"http://{host}:{port}"
+
+    def stop(self) -> None:
+        self._server.shutdown()
+        self._server.server_close()
+
+    def _handler(self):
+        desk = self
+
+        class Handler(BaseHTTPRequestHandler):
+            def log_message(self, *_args):
+                pass  # the test output is not an access log
+
+            def do_GET(self):
+                path = self.path.split("?")[0]
+                desk.requests.append(("GET", path))
+                self._json(desk.script.get(path, {"files": []}))
+
+            def do_POST(self):
+                length = int(self.headers.get("Content-Length") or 0)
+                body = json.loads(self.rfile.read(length) or b"{}")
+                path = self.path.split("?")[0]
+                desk.requests.append(("POST", path, body))
+                response = desk.script.get(path)
+                if callable(response):
+                    response = response(body)
+                if path.endswith("converse_stream"):
+                    self._ndjson(response or [])
+                else:
+                    self._json(response if response is not None else {"type": "reply", "text": "ok"})
+
+            def _json(self, payload):
+                data = json.dumps(payload).encode()
+                self.send_response(200)
+                self.send_header("Content-Type", "application/json")
+                self.send_header("Content-Length", str(len(data)))
+                self.end_headers()
+                self.wfile.write(data)
+
+            def _ndjson(self, events):
+                self.send_response(200)
+                self.send_header("Content-Type", "application/x-ndjson")
+                self.end_headers()
+                for event in events:
+                    self.wfile.write((json.dumps(event) + "\n").encode())
+                    self.wfile.flush()
+
+        return Handler
+
+
+class FakeMic:
+    """Loud frames then quiet ones, so the VAD ends the utterance on its own."""
+
+    def __init__(self, loud=8, quiet=60):
+        self.frames = ([np.full((320, 1), 4000, dtype=np.int16)] * loud
+                       + [np.zeros((320, 1), dtype=np.int16)] * quiet)
+
+    def read(self, _n):
+        return (self.frames.pop(0) if self.frames
+                else np.zeros((320, 1), dtype=np.int16)), None
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, *_a):
+        return False
+
+
+@pytest.fixture
+def pipeline(monkeypatch):
+    """A controller wired to a real local server, with fake ears and mouth."""
+    desk = FakeDesk()
+    spoken: list[str] = []
+    levels: list[float] = []
+
+    monkeypatch.setattr(controller_mod.config, "SERVER_URL", desk.url)
+    monkeypatch.setattr(controller_mod.config, "API_KEY", "test-key")
+    monkeypatch.setattr(controller_mod.config, "SESSION_ID", "pet-smoke")
+    monkeypatch.setattr(controller_mod.config, "SILENCE_END_SEC", 0.2)
+    monkeypatch.setattr(controller_mod.config, "MIN_UTTERANCE_S", 0.0)
+    monkeypatch.setattr(controller_mod.config, "RECEIVE_FILES", False)
+    # Off by default so each test picks its own path; the streaming tests opt in.
+    monkeypatch.setattr(controller_mod.config, "STREAMING_REPLIES", False)
+    monkeypatch.setattr(controller_mod.screen_context, "context_for", lambda text: text)
+    monkeypatch.setattr(controller_mod.screen_context, "is_fullscreen_active", lambda: False)
+    # Deepgram and the speakers are the two things a test can't have.
+    monkeypatch.setattr(controller_mod.stt, "transcribe", lambda pcm: "what's on my disk?")
+    monkeypatch.setattr(controller_mod.stt_stream.StreamingTranscriber, "open",
+                        classmethod(lambda cls, **kw: None))
+
+    def fake_speak(text, on_error=None, should_stop=None, voice_id=None, on_level=None):
+        spoken.append(text)
+        if on_level is not None:
+            on_level(0.8)   # the mouth opens while a word plays...
+            levels.append(0.8)
+        return True
+
+    monkeypatch.setattr(controller_mod.tts, "speak", fake_speak)
+
+    ctrl = controller_mod.PetController()
+    ctrl._stream = FakeMic()
+    try:
+        yield ctrl, desk, spoken, levels
+    finally:
+        desk.stop()
+
+
+# ── the whole path ──────────────────────────────────────────────────────────
+
+def test_a_plain_turn_goes_mic_to_speaker(pipeline):
+    ctrl, desk, spoken, _levels = pipeline
+    desk.script["/desk/converse"] = {"type": "reply", "text": "About sixty percent full."}
+
+    ctrl._handle_conversation_turn()
+
+    assert spoken == ["About sixty percent full."]
+    assert ctrl._state.state == PetState.IDLE
+    posted = [r for r in desk.requests if r[0] == "POST"]
+    assert posted[0][1] == "/desk/converse"
+    assert "what's on my disk?" in posted[0][2]["text"]
+    assert ctrl.history.entries()[-1].text == "About sixty percent full."
+
+
+def test_a_streamed_turn_speaks_each_sentence_as_it_lands(pipeline, monkeypatch):
+    """The NDJSON is parsed by the real client over a real socket — the layer
+    that a mocked `converse` can never exercise."""
+    ctrl, desk, spoken, _levels = pipeline
+    monkeypatch.setattr(controller_mod.config, "STREAMING_REPLIES", True)
+    desk.script["/desk/converse_stream"] = [
+        {"type": "say", "text": "The disk is fine."},
+        {"type": "say", "text": "About sixty percent used."},
+        {"type": "reply", "text": "The disk is fine. About sixty percent used.",
+         "already_spoken": True},
+    ]
+
+    ctrl._handle_conversation_turn()
+
+    assert spoken == ["The disk is fine.", "About sixty percent used."]
+    # The final reply must not be spoken a third time.
+    assert len(spoken) == 2
+
+
+def test_a_tool_turn_says_give_me_a_sec_then_the_answer(pipeline, monkeypatch):
+    """Holding line, relayed command, and the real answer — in that order."""
+    ctrl, desk, spoken, _levels = pipeline
+    ran = []
+    monkeypatch.setattr(controller_mod.server_client, "run_local_command",
+                        lambda cmd: ran.append(cmd) or "[exit 0]\n60% used")
+    desk.script["/desk/converse"] = {
+        "type": "command", "command": "df -h /", "token": "tok",
+        "say": "Let me check that for you.",
+    }
+    desk.script["/desk/tool_result"] = {"type": "reply", "text": "Sixty percent used."}
+
+    ctrl._handle_conversation_turn()
+
+    assert spoken == ["Let me check that for you.", "Sixty percent used."]
+    assert ran == ["df -h /"]
+    relayed = next(r for r in desk.requests if r[0] == "POST" and r[1] == "/desk/tool_result")
+    assert relayed[2]["output"].endswith("60% used")
+
+
+def test_a_device_command_never_reaches_the_shell(pipeline, monkeypatch):
+    ctrl, desk, spoken, _levels = pipeline
+    ran = []
+    monkeypatch.setattr(controller_mod.server_client, "run_local_command", lambda cmd: ran.append(cmd))
+    actions = []
+    ctrl.action.connect(actions.append)
+    desk.script["/desk/converse"] = {
+        "type": "command", "command": "petctl emote wave", "token": "tok"}
+    desk.script["/desk/tool_result"] = {"type": "reply", "text": "There you go."}
+
+    ctrl._handle_conversation_turn()
+
+    assert ran == []
+    assert actions == [{"action": "emote", "emote": "wave"}]
+    assert spoken == ["There you go."]
+
+
+def test_the_mouth_moves_with_the_audio(pipeline):
+    """Lip-sync is only real if the level actually reaches the window."""
+    ctrl, desk, _spoken, levels = pipeline
+    mouth = []
+    ctrl.mouth.connect(mouth.append)
+    desk.script["/desk/converse"] = {"type": "reply", "text": "Talking now."}
+
+    ctrl._handle_conversation_turn()
+
+    assert levels, "TTS was never given a level callback"
+    assert 0.8 in mouth              # opened while speaking...
+    assert mouth[-1] == 0.0          # ...and closed at the end
+
+
+def test_a_notification_is_forwarded_and_spoken(pipeline):
+    """The path that was silently sitting for five to ten minutes."""
+    ctrl, desk, spoken, _levels = pipeline
+    ctrl._notification_gate = controller_mod.notifications.NotificationGate("", 0)
+    desk.script["/desk/converse"] = {"type": "reply", "text": "Harry says he's around."}
+
+    ctrl._queue_notification(Notification(app="Signal", summary="Harry", body="you there?"))
+    ctrl._maybe_heartbeat()
+
+    assert spoken == ["Harry says he's around."]
+    forwarded = next(r for r in desk.requests if r[0] == "POST" and r[1] == "/desk/converse")
+    assert "Harry" in forwarded[2]["text"]
+
+
+def test_streaming_falls_back_when_the_server_is_older(pipeline, monkeypatch):
+    """A server without /desk/converse_stream (or one that answers with nothing)
+    must not cost a turn — the client drops to the plain endpoint. This is how
+    the pet keeps working against a container that hasn't been updated yet."""
+    ctrl, desk, spoken, _levels = pipeline
+    monkeypatch.setattr(controller_mod.config, "STREAMING_REPLIES", True)
+    desk.script["/desk/converse_stream"] = []          # nothing streamed back
+    desk.script["/desk/converse"] = {"type": "reply", "text": "Still here."}
+
+    ctrl._handle_conversation_turn()
+
+    assert spoken == ["Still here."]
+    paths = [r[1] for r in desk.requests if r[0] == "POST"]
+    assert paths == ["/desk/converse_stream", "/desk/converse"]
+
+
+def test_a_dead_server_leaves_the_pet_usable(pipeline):
+    """It should flash an error and go back to listening, not wedge."""
+    ctrl, desk, spoken, _levels = pipeline
+    desk.stop()   # the server disappears mid-session
+    states = []
+    ctrl.state_changed.connect(states.append)
+
+    ctrl._handle_conversation_turn()
+
+    assert states[-2:] == ["error", "idle"]
+    assert spoken == []

tests/test_speech.py

@@ -0,0 +1,260 @@
+"""The rules about talking — the four kinds of utterance and the mic policy.
+
+These were four near-copies in the controller before, and the copies had
+drifted: one didn't arm barge-in, one skipped the follow-up rule. The value of
+having one `Speaker` is only real if the differences between the kinds stay
+*visible*, so this asserts on the differences rather than on the machinery.
+"""
+
+import sys
+from pathlib import Path
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from bolt_pet import config, speech
+from bolt_pet.speech import Speaker, Utterance
+from bolt_pet.state import PetState, PetStateMachine
+
+
+class FakeTts:
+    def __init__(self, completed=True):
+        self.completed = completed
+        self.calls = []
+
+    def speak(self, text, on_error=None, should_stop=None, voice_id=None, on_level=None):
+        self.calls.append({"text": text, "voice_id": voice_id,
+                           "interruptible": should_stop is not None})
+        if on_level is not None:
+            on_level(0.7)
+        return self.completed
+
+    def play_pcm(self, pcm, sample_rate, should_stop=None, on_level=None):
+        self.calls.append({"pcm": pcm, "rate": sample_rate,
+                           "interruptible": should_stop is not None})
+        return self.completed
+
+
+class FakeBargeIn:
+    def __init__(self):
+        self.resets = 0
+
+    def reset(self):
+        self.resets += 1
+
+    def check(self, _frame=None):
+        return False
+
+
+def build(**kwargs):
+    """A speaker plus the things worth asserting on."""
+    state = PetStateMachine()
+    tts = kwargs.pop("tts", None) or FakeTts()
+    said, logged, recorded, levels = [], [], [], []
+    speaker = Speaker(
+        state=state, tts=tts,
+        history=recorded.append,
+        on_said=said.append,
+        on_log=logged.append,
+        on_level=levels.append,
+        **kwargs,
+    )
+    return speaker, state, tts, said, logged, recorded, levels
+
+
+# ── what distinguishes the four kinds ───────────────────────────────────────
+
+def test_a_reply_is_recorded_but_a_holding_line_is_not():
+    """Filler must not push the actual answer out of the transcript."""
+    speaker, state, _tts, _said, _logged, recorded, _levels = build()
+    state.transition(PetState.LISTENING)
+    state.transition(PetState.THINKING)
+
+    speaker.say(Utterance.holding("Give me a sec."))
+    speaker.say(Utterance.reply("Sixty percent used."))
+
+    assert recorded == ["Sixty percent used."]
+
+
+def test_a_holding_line_resumes_the_turn_it_interrupted():
+    """The turn isn't over — a tool is still running — so it must go back to
+    THINKING, not drop to IDLE and end the turn."""
+    speaker, state, _tts, *_ = build()
+    state.transition(PetState.LISTENING)
+    state.transition(PetState.THINKING)
+
+    speaker.say(Utterance.holding("Let me check."))
+
+    assert state.state == PetState.THINKING
+
+
+def test_a_holding_line_cannot_be_talked_over():
+    """Cutting off "give me a sec" strands the tool that's already running."""
+    detector = FakeBargeIn()
+    speaker, state, tts, *_ = build(barge_in=lambda: detector)
+    state.transition(PetState.LISTENING)
+    state.transition(PetState.THINKING)
+
+    speaker.say(Utterance.holding("One sec."))
+    assert tts.calls[-1]["interruptible"] is False
+
+    speaker.say(Utterance.reply("Done."))
+    assert tts.calls[-1]["interruptible"] is True
+
+
+def test_a_streamed_sentence_stays_talking_between_sentences():
+    """Otherwise the sprite flickers idle-talking-idle down a long answer."""
+    speaker, state, *_ = build()
+    state.transition(PetState.LISTENING)
+    state.transition(PetState.THINKING)
+
+    speaker.say(Utterance.stream_chunk("The disk is fine."))
+
+    assert state.state == PetState.TALKING
+
+
+def test_a_scene_is_recorded_because_the_user_heard_it():
+    speaker, state, _tts, _said, _logged, recorded, _levels = build()
+    state.transition(PetState.LISTENING)
+    state.transition(PetState.THINKING)
+
+    speaker.say_pcm(Utterance.scene("Once upon a time."), pcm=b"\x00\x00", sample_rate=44100)
+
+    assert recorded == ["Once upon a time."]
+    assert state.state == PetState.THINKING
+
+
+# ── barge-in ordering, which is what the copies got wrong ───────────────────
+
+def test_the_detector_is_reset_after_playback_not_only_before():
+    """Playback fed the pet's own voice into the wake model's window. If it
+    isn't cleared afterwards, the idle listener re-hears the last sentence and
+    the pet answers itself."""
+    detector = FakeBargeIn()
+    speaker, state, *_ = build(barge_in=lambda: detector)
+    state.transition(PetState.LISTENING)
+    state.transition(PetState.THINKING)
+
+    speaker.say(Utterance.reply("Hello there."))
+
+    assert detector.resets == 2  # armed before, cleared after
+
+
+def test_the_barge_in_detail_is_captured_before_the_reset():
+    """Read it after the reset and every interruption reports zeroed counters —
+    which reads like hard evidence and is nothing of the sort."""
+    detector = FakeBargeIn()
+    details = ["score 0.81 at frame 12", "score 0.000 at frame 0"]
+    speaker, state, *_ = build(barge_in=lambda: detector,
+                               detail_of=lambda: details.pop(0))
+    state.transition(PetState.LISTENING)
+    state.transition(PetState.THINKING)
+
+    speaker.say(Utterance.reply("Hello there."))
+
+    assert speaker.last_detail == "score 0.81 at frame 12"
+
+
+def test_a_detector_that_appears_late_is_still_used():
+    """The detector is built after the speaker — it needs the mic stream — so
+    it's read through a callable. Holding a copy is how the two drift apart."""
+    detector = None
+    speaker, state, tts, *_ = build(barge_in=lambda: detector)
+    state.transition(PetState.LISTENING)
+    state.transition(PetState.THINKING)
+
+    speaker.say(Utterance.reply("Before."))
+    assert tts.calls[-1]["interruptible"] is False
+
+    detector = FakeBargeIn()
+    speaker.say(Utterance.reply("After."))
+    assert tts.calls[-1]["interruptible"] is True
+
+
+# ── the mouth ───────────────────────────────────────────────────────────────
+
+def test_the_mouth_is_closed_when_the_line_ends():
+    """A pet left mid-vowel after the audio stops looks broken."""
+    speaker, state, _tts, _said, _logged, _recorded, levels = build()
+    state.transition(PetState.LISTENING)
+    state.transition(PetState.THINKING)
+
+    speaker.say(Utterance.reply("Talking."))
+
+    assert levels[0] > 0
+    assert levels[-1] == 0.0
+
+
+# ── text handling ───────────────────────────────────────────────────────────
+
+def test_an_empty_line_is_not_spoken_at_all():
+    """A reply that is nothing but an audio tag reduces to '' — and a silent
+    bubble with no audio is better than the pet announcing "laughing"."""
+    speaker, state, tts, said, *_ = build()
+
+    assert speaker.say(Utterance.reply("[laughing]")) is True
+    assert tts.calls == []
+    assert said == []
+    assert state.state == PetState.IDLE
+
+
+def test_the_bubble_gets_display_text_and_tts_gets_the_original():
+    """The bubble keeps emoji and drops markdown; TTS does its own stripping
+    (inside speak(), so every path is covered) and needs the real text."""
+    speaker, state, tts, said, *_ = build()
+    state.transition(PetState.LISTENING)
+    state.transition(PetState.THINKING)
+
+    speaker.say(Utterance.reply("**Sixty** percent 🎉"))
+
+    assert said == ["Sixty percent 🎉"]
+    assert tts.calls[-1]["text"] == "**Sixty** percent 🎉"
+
+
+def test_a_voice_override_reaches_tts():
+    speaker, state, tts, *_ = build(voice_id=lambda: "voice-abc")
+    state.transition(PetState.LISTENING)
+    state.transition(PetState.THINKING)
+
+    speaker.say(Utterance.reply("In character."))
+
+    assert tts.calls[-1]["voice_id"] == "voice-abc"
+
+
+# ── the follow-up rule ──────────────────────────────────────────────────────
+
+@pytest.fixture
+def follow_ups_on(monkeypatch):
+    monkeypatch.setattr(config, "FOLLOW_UP_LISTEN", True)
+    monkeypatch.setattr(config, "FOLLOW_UP_MAX_TURNS", 3)
+
+
+def test_an_interruption_always_reopens_the_mic(follow_ups_on):
+    """You talked over it — you are mid-sentence, so it has to listen."""
+    keep, why = speech.follow_up_decision("Anything else?", completed=False, follow_ups=99)
+    assert keep is True
+    assert why == "interrupted"
+
+
+def test_a_question_keeps_the_mic_open_without_the_wake_word(follow_ups_on):
+    keep, why = speech.follow_up_decision("Want me to check?", completed=True, follow_ups=0)
+    assert (keep, why) == (True, "question")
+
+
+def test_a_statement_ends_the_turn(follow_ups_on):
+    keep, _why = speech.follow_up_decision("Sixty percent used.", completed=True, follow_ups=0)
+    assert keep is False
+
+
+def test_the_cap_stops_a_server_that_ends_every_reply_with_a_question(follow_ups_on):
+    """Otherwise mic noise loops it forever."""
+    assert speech.follow_up_decision("Ok?", completed=True, follow_ups=2)[0] is True
+    keep, why = speech.follow_up_decision("Ok?", completed=True, follow_ups=3)
+    assert keep is False
+    assert "cap" in why
+
+
+def test_the_rule_can_be_switched_off(monkeypatch):
+    monkeypatch.setattr(config, "FOLLOW_UP_LISTEN", False)
+    assert speech.follow_up_decision("Ok?", completed=True, follow_ups=0)[0] is False

tests/test_speech_text.py

@@ -4,6 +4,8 @@
 import sys
 from pathlib import Path
 
+import pytest
+
 sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
 
 from bolt_pet.speech_text import for_display, for_speech, is_question
@@ -77,3 +79,33 @@ def test_is_question_ignores_question_marks_that_are_not_spoken():
     assert not is_question("Docs are at https://example.com/x?y=1")
     assert not is_question("")
     assert not is_question(None)
+
+
+# ── ElevenLabs v3 delivery tags (observed live 2026-07-31) ──────────────────
+# "[laughing] That one came through clean" was spoken as "laughing That one
+# came through clean": the tags mean something to the dialogue model inside a
+# dialoguectl scene, and nothing at all to the ordinary reply voice.
+
+def test_delivery_tags_are_never_spoken_aloud():
+    spoken = for_speech("[laughing] That one came through clean.")
+    assert "laughing" not in spoken
+    assert spoken.startswith("That one came through clean")
+
+
+def test_the_bubble_does_not_caption_a_laugh_nobody_heard():
+    assert "[laughing]" not in for_display("[laughing] All good.")
+
+
+@pytest.mark.parametrize("tag", [
+    "[whispering]", "[cheerfully]", "[sighs]", "[laughs]", "[clears throat]",
+    "[nervously]", "[excited]", "[pause]", "[shouting]",
+])
+def test_the_common_tags_are_all_covered(tag):
+    assert "" == for_speech(tag).strip()
+
+
+def test_ordinary_bracketed_text_survives():
+    """The tags are matched narrowly on purpose — real bracketed content is
+    part of what the user asked to hear."""
+    assert "1" in for_speech("See reference [1] for details.")
+    assert "docs" in for_speech("It's in [the docs] somewhere.")

tests/test_stt_stream.py

@@ -0,0 +1,173 @@
+"""Streaming speech-to-text: the protocol, and the fallback that makes it safe
+to switch on at all.
+
+A fake websocket throughout — no network, no Deepgram account, no audio.
+"""
+
+import json
+import sys
+import time
+from pathlib import Path
+
+import numpy as np
+import pytest
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from bolt_pet.audio import mic
+from bolt_pet.audio.stt_stream import StreamingTranscriber
+
+
+class FakeSocket:
+    """Records what was sent; replays scripted Deepgram frames."""
+
+    def __init__(self, messages=(), fail_on_send=False):
+        self.sent = []
+        self.closed = False
+        self.fail_on_send = fail_on_send
+        self._messages = list(messages)
+
+    def send_binary(self, data):
+        if self.fail_on_send:
+            raise ConnectionError("socket died")
+        self.sent.append(data)
+
+    def send(self, text):
+        self.sent.append(text)
+
+    def recv(self):
+        if self._messages:
+            return self._messages.pop(0)
+        time.sleep(0.01)
+        raise ConnectionError("closed")
+
+    def close(self):
+        self.closed = True
+
+
+def _results(transcript, is_final=True):
+    return json.dumps({
+        "type": "Results", "is_final": is_final,
+        "channel": {"alternatives": [{"transcript": transcript}]},
+    })
+
+
+def _frame(value=1000):
+    return np.full(320, value, dtype=np.int16)
+
+
+# ── the protocol ────────────────────────────────────────────────────────────
+
+def test_frames_go_up_as_they_are_captured():
+    socket = FakeSocket([_results("what's the weather")])
+    session = StreamingTranscriber(socket)
+
+    for _ in range(3):
+        session.feed(_frame())
+    text = session.finish()
+
+    assert len(socket.sent) == 4          # three frames plus the close message
+    assert text == "what's the weather"
+    assert socket.closed
+
+
+def test_only_final_results_are_kept():
+    """Interim hypotheses change under you; concatenating them would produce
+    "what what's what's the what's the weather"."""
+    socket = FakeSocket([
+        _results("what's", is_final=False),
+        _results("what's the", is_final=False),
+        _results("what's the weather", is_final=True),
+    ])
+    session = StreamingTranscriber(socket)
+    assert session.finish() == "what's the weather"
+
+
+def test_several_final_segments_are_joined():
+    socket = FakeSocket([_results("turn the lights on"), _results("in the kitchen")])
+    session = StreamingTranscriber(socket)
+    assert session.finish() == "turn the lights on in the kitchen"
+
+
+def test_junk_frames_are_ignored_rather_than_killing_the_reader():
+    """An exception on the reader thread would silently end transcription for
+    the rest of the utterance."""
+    socket = FakeSocket(["not json at all", '{"type":"Metadata"}',
+                         _results("still works")])
+    session = StreamingTranscriber(socket)
+    assert session.finish() == "still works"
+
+
+def test_an_empty_frame_means_the_socket_closed():
+    """websocket-client returns "" from recv() on a closed connection, so it
+    ends the read loop rather than being treated as a blank transcript."""
+    socket = FakeSocket([_results("heard this much"), "", _results("never arrives")])
+    session = StreamingTranscriber(socket)
+    assert session.finish() == "heard this much"
+
+
+def test_a_socket_that_dies_mid_utterance_gives_up_quietly():
+    socket = FakeSocket([], fail_on_send=True)
+    session = StreamingTranscriber(socket)
+
+    session.feed(_frame())      # must not raise — recording carries on
+    assert session.finish() == ""
+
+
+# ── opening: failure is an ordinary outcome ────────────────────────────────
+
+def test_open_returns_none_when_it_cannot_connect():
+    """None means "the one-shot path will do it", not an error."""
+    def refuse():
+        raise OSError("no network")
+
+    assert StreamingTranscriber.open(connect=refuse) is None
+
+
+def test_open_returns_a_session_when_it_can():
+    session = StreamingTranscriber.open(connect=lambda: FakeSocket([_results("hi")]))
+    assert session is not None
+    assert session.finish() == "hi"
+
+
+def test_streaming_is_off_without_the_switch_or_a_key(monkeypatch):
+    from bolt_pet.audio import stt_stream
+
+    monkeypatch.setattr(stt_stream.config, "STT_STREAMING", False)
+    assert stt_stream.available() is False
+    monkeypatch.setattr(stt_stream.config, "STT_STREAMING", True)
+    monkeypatch.setattr(stt_stream.config, "DEEPGRAM_API_KEY", "")
+    assert stt_stream.available() is False
+
+
+# ── the capture hook ────────────────────────────────────────────────────────
+
+class _Stream:
+    """Loud frames, then quiet ones, so the VAD ends the utterance."""
+
+    def __init__(self, loud=6, quiet=40):
+        self.frames = ([np.full((320, 1), 3000, dtype=np.int16)] * loud
+                       + [np.zeros((320, 1), dtype=np.int16)] * quiet)
+
+    def read(self, n):
+        return (self.frames.pop(0) if self.frames
+                else np.zeros((320, 1), dtype=np.int16)), None
+
+
+def test_recording_hands_every_speech_frame_to_the_listener():
+    seen = []
+    pcm = mic.record_utterance(_Stream(), on_frame=seen.append,
+                               silence_end_sec=0.2, min_utterance_s=0.0)
+    assert pcm is not None
+    assert len(seen) >= 6            # every frame of speech was streamed
+
+
+def test_a_listener_that_throws_cannot_break_the_recording():
+    """The fallback is about to need this audio — a dead stream must not cost
+    the recording too."""
+    def explode(frame):
+        raise RuntimeError("stream died")
+
+    pcm = mic.record_utterance(_Stream(), on_frame=explode,
+                               silence_end_sec=0.2, min_utterance_s=0.0)
+    assert pcm is not None and len(pcm) > 0

tests/test_window_state.py

@@ -0,0 +1,84 @@
+"""Remembering where the pet was left — and refusing to when that's a trap."""
+
+import json
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from bolt_pet import window_state
+
+
+def test_a_saved_position_comes_back(tmp_path):
+    path = tmp_path / "window.json"
+    window_state.save(1200, 640, path)
+    assert window_state.load(path) == (1200, 640)
+
+
+def test_no_file_yet_is_not_an_error(tmp_path):
+    assert window_state.load(tmp_path / "nope.json") is None
+
+
+def test_a_corrupt_file_falls_back_to_the_default_corner(tmp_path):
+    """Whatever is in there, the pet still has to start."""
+    path = tmp_path / "window.json"
+    for junk in ("", "{", "null", "[]", '{"x": "left"}', '{"y": 3}'):
+        path.write_text(junk)
+        assert window_state.load(path) is None
+
+
+def test_saving_creates_the_cache_directory(tmp_path):
+    path = tmp_path / "deep" / "nested" / "window.json"
+    window_state.save(10, 20, path)
+    assert window_state.load(path) == (10, 20)
+
+
+def test_an_unwritable_location_is_swallowed(tmp_path):
+    """A read-only cache dir is a reason to forget the position, not to crash
+    on every drag."""
+    window_state.save(1, 2, tmp_path / "no" / "\0bad" / "window.json")
+
+
+def test_the_write_is_atomic_and_leaves_no_litter(tmp_path):
+    path = tmp_path / "window.json"
+    window_state.save(5, 5, path)
+    window_state.save(7, 7, path)
+    assert json.loads(path.read_text()) == {"x": 7, "y": 7}
+    assert [p.name for p in tmp_path.iterdir()] == ["window.json"]
+
+
+# ── validating against the screens that exist *now* ─────────────────────────
+
+LAPTOP = (0, 0, 1920, 1080)
+EXTERNAL = (1920, 0, 4480, 1440)
+
+
+def test_a_position_on_a_connected_screen_is_kept():
+    assert window_state.is_visible_on(1700, 900, 128, [LAPTOP])
+
+
+def test_a_position_on_an_unplugged_monitor_is_refused():
+    """The dangerous case: restoring it faithfully puts the pet somewhere you
+    can't see or reach."""
+    assert not window_state.is_visible_on(3000, 700, 128, [LAPTOP])
+    assert window_state.is_visible_on(3000, 700, 128, [LAPTOP, EXTERNAL])
+
+
+def test_mostly_off_screen_counts_as_gone():
+    """A few pixels of ear poking onto the desktop is not "reachable"."""
+    assert not window_state.is_visible_on(1910, 500, 128, [LAPTOP])
+    assert window_state.is_visible_on(1830, 500, 128, [LAPTOP])
+
+
+def test_touching_an_edge_is_not_overlapping():
+    assert not window_state.is_visible_on(1920, 0, 128, [LAPTOP])
+
+
+def test_negative_coordinates_are_fine_when_a_screen_is_there():
+    """Monitors left of or above the primary have negative origins."""
+    left_of_primary = (-1920, 0, 0, 1080)
+    assert window_state.is_visible_on(-900, 400, 128, [left_of_primary, LAPTOP])
+
+
+def test_no_screens_at_all_is_not_visible():
+    assert not window_state.is_visible_on(100, 100, 128, [])