commit dfb5b0b7e5675f2d282eeb3e5281d4a9b093135f
parent 24a0396ac6a36d53d7b227bb6c0bd3e6c0c4774b
Author: ling0x <ling0x@users.noreply.github.com>
Date: Mon, 6 Jul 2026 09:06:56 +0100
refactor: test to speech
Diffstat:
4 files changed, 147 insertions(+), 4 deletions(-)
diff --git a/commands/disk.txt b/commands/disk.txt
@@ -15,3 +15,39 @@ for d in ~/.config/*/ ~/.config/.[!.]*/; do
when=$(date -d "@${ts%.*}" '+%Y-%m-%d' 2>/dev/null)
printf '%-12s %8s %s\n' "$when" "$sz" "$(basename "$d")"
done | sort | head -40
+
+====================================================================
+
+Suggested order of attack
+
+ Run these one by one and recheck with df -h:
+
+ ```bash
+ # 1. Empty trash
+ rm -rf ~/.local/share/Trash/*
+
+ # 2. Pacman cache
+ sudo paccache -rk3
+
+ # 3. yay cache
+ rm -rf ~/.cache/yay/*
+
+ # 4. Journal
+ sudo journalctl --vacuum-size=500M
+
+ # 5. Docker
+ docker system prune -a --volumes
+
+ # 6. Rust debug builds only (safer than full cargo clean)
+ find /home/quintessa/code -type d -path "*/target/debug" -prune -exec rm
+ -rf {} +
+
+ # 7. Huggingface cache
+ rm -rf ~/.cache/huggingface/*
+ ```
+
+ That sequence alone should free well over 1.5 TB, probably closer to 2 TB,
+ without touching any of your actual source code, games, or personal files.
+
+ After that, decide on Ollama models, Steam games, and old videos if you
+ still need more space.
diff --git a/commands/linux_audio.txt b/commands/linux_audio.txt
@@ -117,4 +117,18 @@ Quick overview (start here)
├───────────────────┼─────────────────────────────────────────────────────┤
│ Test recording │ arecord -d 5 -f cd /tmp/test.wav && aplay │
│ │ /tmp/test.wav │
- └───────────────────┴─────────────────────────────────────────────────────┘
-\ No newline at end of file
+ └───────────────────┴─────────────────────────────────────────────────────┘
+
+============================================================================
+
+Turn off power-save / auto-turn-off for bluetooth earphone in linux:
+
+echo 'ACTION=="add", SUBSYSTEM=="usb", ATTR{idVendor}=="8087", ATTR{idProduct}=="0032", TEST=="power/control", ATTR{power/control}="on"' | sudo tee
+ /etc/udev/rules.d/50-bluetooth-no-autosuspend.rules
+sudo udevadm control --reload-rules
+echo on | sudo tee /sys/bus/usb/devices/5-7/power/control
+
+
+The last line applies it immediately (no reboot); the udev rule makes it persist across reboots. Verify with:
+
+cat /sys/bus/usb/devices/5-7/power/control # should say: on
diff --git a/commands/piper-tts.txt b/commands/piper-tts.txt
@@ -12,4 +12,21 @@ How to use it to play audio directly:
cat test.txt | piper-tts \
--model /usr/share/piper-voices/en/en_US/bryce/medium/en_US-bryce-medium.onnx \
- --output-raw | aplay -r 22050 -f S16_LE -t raw -
-\ No newline at end of file
+ --output-raw | aplay -r 22050 -f S16_LE -t raw -
+
+====================================================================
+
+Speech Dispatcher
+
+/etc/speech-dispatcher/modules/piper-tts-generic.conf
+
+Here you set up the default voice for reading PDFs which can
+integrate with Okular or Calibre
+
+Restart:
+
+systemctl --user status speech-dispatcher
+
+Test saying sth:
+
+spd-say "This is a test"
diff --git a/commands/stt.txt b/commands/stt.txt
@@ -1,3 +1,81 @@
+ Quick recovery (when bluetooth audio dies)
+
+ # Option A — lighter touch
+ bluetoothctl disconnect B8:84:11:6B:7D:09
+ sleep 2
+ bluetoothctl connect B8:84:11:6B:7D:09
+ # Option B — full reset (what you did at 19:15)
+ systemctl --user restart wireplumber pipewire pipewire-pulse
+ sudo systemctl restart bluetooth
+
+===============================================================================
Restart after stt crash in coding agent:
systemctl --user restart wireplumber pipewire pipewire-pulse
+
+systemctl --user status pipewire pipewire-pulse wireplumber pulseaudio
+
+===============================================================================
+
+ Your adapter is an Intel AX210 (8087:0032), which has a long history of flaky A2DP/CIS behavior with Linux Bluetooth stacks.
+
+ Why it keeps happening
+
+ Common triggers on AX210 + earbuds:
+
+ 1. Earbuds idle/sleep → stale link → first audio after wake fails
+ 2. A2DP profile switch (especially AAC) → packet timing issues
+ 3. A2DP ↔ HFP switch (mic use, /voice, browser tab) → transport teardown race
+ 4. Kernel/firmware regression — you're on kernel 7.0.12, which is very new
+
+ Quick recovery (when audio dies)
+
+ # Option A — lighter touch
+ bluetoothctl disconnect <id>
+ sleep 2
+ bluetoothctl connect <id>
+ # Option B — full reset (what you did at 19:15)
+ systemctl --user restart wireplumber pipewire pipewire-pulse
+ sudo systemctl restart bluetooth
+
+ Mitigations worth trying (most impact first)
+
+ 1. Force SBC instead of AAC (AAC is often more fragile on AX210)
+
+ Create ~/.config/wireplumber/wireplumber.conf.d/51-bluez-a2dp-sbc.conf:
+
+ bluez_monitor.rules = {
+ {
+ matches = {
+ { { "device.name", "matches", "bluez_card.*" } },
+ },
+ apply_properties = {
+ ["bluez5.auto-connect"] = "[ a2dp_sink ]",
+ ["bluez5.codecs"] = "[ sbc ]",
+ },
+ },
+ }
+
+ Then:
+
+ systemctl --user restart wireplumber
+
+ 2. Keep firmware updated
+
+ sudo pacman -Syu linux-firmware bluez pipewire wireplumber
+
+ 3. If it got worse on kernel 7.0, try the LTS kernel as a test — AX210 issues often track kernel versions.
+
+ 4. Long-term workaround: a dedicated USB Bluetooth dongle (Broadcam/MediaTek-based ones often behave better than AX210 for A2DP).
+
+ 5. For /voice specifically: after the earphones have been idle, play a second of audio or reconnect before recording — the HFP switch is extra stress right after a broken A2DP
+ link.
+
+ ────────────────────────────────────────
+
+ Bottom line: At 19:11 your Intel BT adapter lost sync with the A2DP stream to the OpenFit Pro. WirePlumber logged it and kept running; audio died because the BT transport
+ broke, not because PipeWire segfaulted. The sleep-idle → wake → stream pattern fits this well.
+
+ If you want, I can add that WirePlumber config to your dotfiles repo and a small reconnect script for when this happens.
+
+
