It would be awesome if this repo started to add other LED controls #45
Replies: 14 comments 83 replies
-
|
Hey @JenniferHatches. Thanks for the feedback. It is good to hear that others find the repo useful and it's not just relegated to my own LVAs. Thanks for letting me know about @cknic 's repo as well. It wasn't on my radar yet. When you mention the LED controls from cknic's repo, are you referring to the hardware of the LEDs or the LED effects? If it is hardware, I suspect that should be supportable via the config.json file as you can pass (Dotstar or Neopixel), (GPIO or SPI) as well as identifying the data and clock (for dotstar) pins in there. I am fighting with permission when trying to run Neopixel on GPIO though, but Dotstar work great. If you are wanting some more advanced LED effects, that is something that I'm intending to improve as well. I just started with the most common effects up front just to get the code working. |
Beta Was this translation helpful? Give feedback.
-
|
More than willing to help with the XVF3800 LEDs. |
Beta Was this translation helpful? Give feedback.
-
|
I see that it is using WS2812 LEDs. If they are running on SPI, it may work with the following section added to the config.json file: If not, I am more that happy to get it added to the project. |
Beta Was this translation helpful? Give feedback.
-
|
I'll give this a test tomorrow during the day. I don't think SPI will work tho, i think you have to use the xvf_host control software |
Beta Was this translation helpful? Give feedback.
-
|
I'm not using the ESP32 XVF3800, just the USB version. |
Beta Was this translation helpful? Give feedback.
-
|
See my response below, SPI did not work for me. I had to use the XVF_Host to control the LEDs. My project still has some work to go but it is working really well for my specific needs. |
Beta Was this translation helpful? Give feedback.
-
|
i actually used this GitHub https://github.com/KasunThushara/reSpeakerXVF_rpi and then call the xvf_host commands via my led_controller_xvf3800.py |
Beta Was this translation helpful? Give feedback.
-
|
i was having a heck of a time getting SPI to work, this way it just worked natively even if you weren't using the ESP32 on the XVF3800, i'm actually just using it USB connected to a RPI and i still get the LED Commands |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for that repo. I just ordered a xvf3800 and I am eager to get it working. I will start working to incorporating your work into the led_controller as a configuration driven option. I will update this thread when I have potentially working code for testing. |
Beta Was this translation helpful? Give feedback.
-
|
check this repo too: Official firmware updates and xvf_host, python examples, etc. firmware was updated today lol |
Beta Was this translation helpful? Give feedback.
-
|
btw, thanks to you both, this is awesome |
Beta Was this translation helpful? Give feedback.
-
|
If you have a ReSpeaker XVF3800 4-Mic Array (mine is in the mail) and you’re willing to help test integration with the Linux Voice Assistant (LVA), please follow the steps below and paste the outputs. We’ll do this in two tiers: Tier 1: Quick basic info (fast, low effort – ideal starting point)Tier 2: Deeper audio probe (only if you have time / are comfortable)Prep: Clone the XVF3800 branchPlease clone the XVF3800 branch of the LVA: cd ~
git clone --branch XVF3800 --single-branch https://github.com/imonlinux/linux-voice-assistant.git
cd linux-voice-assistantRun the script/setup to establish the venv environment and dependencies. script/setupTier 1 – Basic system + audio info (recommended for everyone)Please run these commands and paste the full outputs in your reply. It’s fine to redact IPs, usernames, or anything you consider sensitive. 1. System infouname -acat /etc/os-release2. ALSA view of the XVF3800arecord -larecord -Laplay -l3. PulseAudio / PipeWire viewIf you have PulseAudio / pipewire-pulse: pactl list short sourcespactl list short sinksIf pactl isn’t available, just mention that in your reply. 4. LVA’s view of audio devicesFrom inside the linux-voice-assistant repo: cd ~/linux-voice-assistant
script/run --list-input-devices
script/run --list-output-devices👉 In your reply, please highlight the lines that correspond to the ReSpeaker XVF3800 (copy the exact text). Tier 2 – Deeper XVF3800 audio probe (optional but very helpful)If you’re comfortable going a bit further, this will tell us exactly how many channels LVA sees and whether we need explicit 2→1 downmixing. These commands assume you’re still in ~/linux-voice-assistant and your venv (if any) is active. source .venv/bin/activate5. Short arecord test with VU outputFirst, find the card number N for the XVF3800 from arecord -l (e.g. card 4: Array [reSpeaker XVF3800 4-Mic Array] → N=4). Then run: arecord -D plughw:N,0 -c 2 -r 16000 -f S16_LE -d 5 -vv /tmp/xvf3800-test.wavReplace N with your actual card number! You can optionally check playback: aplay /tmp/xvf3800-test.wav6. USB descriptor (if available)If you’re comfortable sharing USB info, run: lsusb
lsusb -v -d 2886:001aIf 2886:001a is not present, just paste the lsusb output and we’ll adjust. 7. Python probe script (from XVF3800 branch)In the XVF3800 branch there is a small probe script at tests/xvf3800_probe.py. Run: cd ~/linux-voice-assistant
python tests/xvf3800_probe.py --device "<XVF3800 device name from --list-input-devices>"Examples:If --list-input-devices shows something like python tests/xvf3800_probe.py --device "reSpeaker XVF3800 4-Mic Array: USB Audio (hw:4,0)"or python tests/xvf3800_probe.py --device 3Please paste the entire output of this probe script, including: The device list it prints The Opened stream section The Recorded array shape: ... line What we’ll do with thisWith these outputs we can: Confirm how the XVF3800 appears at each layer (ALSA → PipeWire → PortAudio / sounddevice). See whether LVA receives 1 or 2 channels from the device. Decide if we need any XVF3800-specific handling, or just a generic “if channels > 1, downmix to mono” in the audio engine. Thank you for helping test the XVF3800 integration! |
Beta Was this translation helpful? Give feedback.
-
|
I know this isn't the LED controller yet, but I want to be systematic in the implementation. I suspect that audio will work out of the box, but I'm a dotted i and crossed t kind of guy. LEDs and the mute button are up next. |
Beta Was this translation helpful? Give feedback.
-
|
I actually have a bunch of this done already. As soon as I get back home I will post an update to this. I know at a high level the USB devices were exposed for Mic and then the headphone jack. The wired speaker port I wasn’t using. I did find I stuck with ALSA. Pulse wave did work but was acting funny. Pipewire was not playing nicely for me. I believe I ended up with PCM devices at the end but I’ll look. |
Beta Was this translation helpful? Give feedback.
-
|
Whereas I'm running mine on debian trixie and using pipewire, and yes, the audio works great out of the box. Here's tier 1, I'll do two tomorrow hopefully. note: I have disabled the built in hdmi and headphone jacks on the rpi4 using wireplumber so the only devices are the xvf3800 ones. jennifer@BedroomPi:~ $ uname -a jennifer@BedroomPi:~ $ cat /etc/os-release jennifer@BedroomPi:~ $ arecord -l jennifer@BedroomPi:~ $ arecord -L jennifer@BedroomPi:~ $ aplay -l jennifer@BedroomPi:~ $ pactl list short sources jennifer@BedroomPi:~ $ pactl list short sinks jennifer@BedroomPi:~/testing/linux-voice-assistant $ script/run --list-input-devices
|
Beta Was this translation helpful? Give feedback.
-
|
(.venv) jennifer@BedroomPi:~/testing/linux-voice-assistant $ sudo arecord -D plughw:3,0 -c 2 -r 16000 -f S16_LE -d 5 -vv /tmp/xvf3800-test.wav |
Beta Was this translation helpful? Give feedback.
-
|
(.venv) jennifer@BedroomPi:~/testing/linux-voice-assistant $ lsusb -v -d 2886:001a Bus 001 Device 005: ID 2886:001a Seeed Technology Co., Ltd. reSpeaker XVF3800 4-Mic Array |
Beta Was this translation helpful? Give feedback.
-
🧪 XVF3800 Mute Button / HID Probe – Help WantedFirst round of XVF3800 testing confirmed that audio works out of the box with LVA using PipeWire (and the audio stack handles the 2-channel → mono conversion for us) I have added an initial tutorial to the XVF3800 branch. Next goal: fully integrate the XVF3800 mute button + red mute LED with the LVA’s existing mute path (MQTT, tray icon, etc.). To do that, we need to see how the XVF3800 reports its mute button over USB HID. I’ve added a small probe script on the If you have an XVF3800 connected, you can help by running this probe and posting the output. ❤️ 1. Clone the XVF3800 branch (if you haven’t already)If you’re starting fresh: cd ~
git clone --branch XVF3800 --single-branch https://github.com/imonlinux/linux-voice-assistant.git
cd linux-voice-assistantIf you already have the repo: cd ~/linux-voice-assistant
git fetch origin
git checkout XVF3800Then (re)run the setup to ensure dependencies + venv are ready: cd ~/linux-voice-assistant
script/setupThat should create/refresh 2. Install the HID library in the LVA virtualenvThe probe script uses From the cd ~/linux-voice-assistant
# Activate the LVA venv if your shell doesn’t auto-activate it
source .venv/bin/activate
# Install hidapi inside this venv
pip install hidapiYou can confirm it installed with: python -c "import hid; print('hid version OK')"If that prints without errors, you’re good. 3. Run the XVF3800 HID mute probeMake sure your XVF3800 is:
Then make sure that the .venv is still source, if not run: cd ~/linux-voice-assistant
source .venv/bin/activateThen run: python tests/xvf3800_hid_mute_probe.pyYou should see something like: Now:
4. What to post back in the DiscussionPlease paste:
For example: You don’t need to redact the serial unless you want to; it’s only used for debugging. Once we have a couple of these samples from different testers, we’ll:
Thanks in advance to anyone who runs the probe—this gets us much closer to full, “just like a smart speaker” behavior on the XVF3800. 🙌 |
Beta Was this translation helpful? Give feedback.
-
|
=== XVF3800 HID Mute Probe === NOTE: Using index 0 by default. If this is not the HID interface, Using device: Opening HID device by path: b'1-1.1:1.5' ^C So, the probe runs, but doesn't detect any button presses. If i press the reset button instead, it crashes the process with a timeout read error, so it seems to be looking at the correct device. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for this @JenniferHatches. I should have plugged the repo that you flagged much sooner in this process (https://github.com/respeaker/reSpeaker_XVF3800_USB_4MIC_ARRAY/). It is the key to getting all of this working. I don't think that we are going to need to do any more test scripts. I am looking into implementing everything via pyusb and it should be a big bang update to the XVF3800 branch. I will drop another comment when that code is ready for testing. ChatGPT has been pretty good lately, but I'm sure that there will be bugs to test through. |
Beta Was this translation helpful? Give feedback.
-
|
I find Claude and Grok has been excellent for coding, although i hear the new Gemini is fanstatic. Anyway... i'm having the same issue as Jennifer when testing your script. |
Beta Was this translation helpful? Give feedback.
-
|
I agree with you that Gemini is good too. I periodically have either ChatGPT or Gemini completely forget the LVA project and then reintroduce it to them as a fresh code review (usually of the other LLMs code). This has found quite a few issues and I think made for a better, cleaner code base. It also prevents the LLM from getting too caught up in previous decisions that may not be the best long term approach. |
Beta Was this translation helpful? Give feedback.
-
looks like they just released a new firmware that also allows individual LED control... neat. |
Beta Was this translation helpful? Give feedback.
-
|
FYI if you are looking to upgrade the firmware do NOT wget the files, clone the repo. wget is corrupting them somehow |
Beta Was this translation helpful? Give feedback.
-
|
i didn't even realize there was a python script now.. interested to see what you can do with that, but with the normal xvf_host, i can mute the device, with "sudo ./xvf_host GPO_WRITE_VALUE 30 1" that sets X0D30 to high, which is mute, and 0 which is unmute. I was also disabling the LEDs, as i didn't want LOA all the time by setting 33 to 0. |
Beta Was this translation helpful? Give feedback.
-
|
Yep, I am going down the python path. I just pushed some changes to the XVF3800 branch that should support the existing LED effects defined in the LVA and controllable by MQTT using the legacy XVF3800 effects. I will dive into the new LED functionality provided by the new firmware once I have my hands on one of the devices. I am drafting a new note to walk through the changes and what to include in the config.json file to test it. |
Beta Was this translation helpful? Give feedback.
-
Call for Testing: XVF3800 USB LED and Mute Button Support in
|
Beta Was this translation helpful? Give feedback.
-
|
@cknic @JenniferHatches Feel free to toss over any debug logs that you have for these issues. The XVF3800 is hard to get lately. I have ordered 3 of them from different sources. Looks like the they are all still a week or more out. |
Beta Was this translation helpful? Give feedback.
-
|
I'll try to get debug logs. |
Beta Was this translation helpful? Give feedback.
-
|
So, what logs would be useful? |
Beta Was this translation helpful? Give feedback.
-
|
Hey @JenniferHatches, 2 of my XVF3800s were just delivered this weekend. I will be doing a lot of testing. One thing that I have seen so far is that on initial boot of the RPi with the XVF3800 attached, the microphone doesn't seem to be working. I have to unplug it and reconnect followed by a restart of the LVA. It seems to be OK after that. I suspect that it is the order in which everything is starting up and that I need to includes some Depends and Wants in the LVA service file. Are you seeing that too? |
Beta Was this translation helpful? Give feedback.
-
|
not with my rpi5 but yes with my rpi4. I have to do this on the rpi4 any time i change/modify/set the config at all using xvf_host - on the next boot, the service will be running, but not actually listening on the mic. If i play an announcement through the rpi4, the microphone works for like 5 seconds or something, then stops again. Have to unplug or reflash the xvf_3800 to get it back. |
Beta Was this translation helpful? Give feedback.
-
|
I vaguely remember coming across that during my adventures and had to change the order of things. I’ll poke through my git for what I did because I don’t specifically remember |
Beta Was this translation helpful? Give feedback.
-
|
Same with it going to sleep after a while and not responding to the wake word but I believe that was a process hanging and spiking the CPU |
Beta Was this translation helpful? Give feedback.
-
|
ok i'm having a strange issue on a re-install. It responds to Alexa or whatever the wake word is but is getting stuck. So it detects the wakeword, plays the wake word sound, changes to the pulsing blue but never actually follows up and gives me "2025-12-22 12:03:22 DEBUG linux_voice_assistant.event_bus: Publishing event to 1 listeners on topic 'set_listening_color' |
Beta Was this translation helpful? Give feedback.
-
|
Hey @cknic, I have reverted a refractory bug fix of audio_engine.py in the XVF3800 branch that was merged last week. Let me know if this resolves your issue. |
Beta Was this translation helpful? Give feedback.
-
|
ok i'm an idiot.. it was not on you at all. Somehow HA had the Speech to Text unset.. no idea how that happened. |
Beta Was this translation helpful? Give feedback.
-
|
I have seen that before. I will put the newer version of the audio_engine.py file back in the branch. Thanks for letting me know. |
Beta Was this translation helpful? Give feedback.
-
|
of course. What else can i test with. I know one thing i wanted to commit up was the updated config.json |
Beta Was this translation helpful? Give feedback.
-
|
my big question is, are we actually using the correct channels on the microphone? by default, the two channels (stereo mic) are two different feeds - I can't imagine that pipewire just mashing the two together is the optimal configuration. it works, yes, but optimal? Surely detection would be improved with a change? |
Beta Was this translation helpful? Give feedback.
-
|
i was stalking your commits.. works now. |
Beta Was this translation helpful? Give feedback.
-
|
I have just uploaded a couple of changes. I have made the volume_sync function off by default and included a config for setting the max_volume_percent (for quite devices that need that extra boost) in the config.json file. To enable and set just add these two values to the end of the audio section in the config.json file. I have updated the config.json.example file to include these values. This changes the previous behavior where volume_sync was enabled by default. |
Beta Was this translation helpful? Give feedback.
-
|
What’s scenario would you think you would want it to be false? I’m just thinking through it wouldn’t you always want it to be true? |
Beta Was this translation helpful? Give feedback.
-
|
I am making it an intentional decision to turn it on as it does break the per/application volume that is a key component of both PipeWire and PulseAudio. This wouldn't impact a single use device like setting up a LVA satellite, but if you use the code on a workstation with the tray client or want to setup Snapcast or Squeezebox clients on the device then it could be an issue. |
Beta Was this translation helpful? Give feedback.
-
|
For example, I'm running it on an rpi5 that is also used for movies, videos, retro gaming, etc. Per application volume is nice lol thank you for making it optional. |
Beta Was this translation helpful? Give feedback.
-
|
@JenniferHatches i'm starting to mess with Music Assistant on this device.. it's telling me its working, but i'm getting this error in the logs. 2026-01-05 08:54:07 ERROR linux_voice_assistant.mpv_player: mpv[ffmpeg/audio]: mp3float: Header missing |
Beta Was this translation helpful? Give feedback.
-
|
@cknic , other than the extended pause issue, is the LVA not playing the audio from Music Assistant? Or are you also seeing the errors in the logs when the player is running? For the player not responding after a long pause, have you implemented the udev rule that prevents the USB from going into sleep mode? There is now a udev rule in the XVF3800 folder (99-respeaker-xvf3800.rules). You just need to copy it into /etc/udev/rules.d/ and reboot. I will keep looking into it otherwise. |
Beta Was this translation helpful? Give feedback.
-
|
Oooh I'll try the udev rule to see if it fixes it |
Beta Was this translation helpful? Give feedback.
-
|
@JenniferHatches @cknic, The latest HA update 2026.1 breaks LLM conversation agents. This is the bug report: I'm going to see if restoring from last nights backup restores the functionality. |
Beta Was this translation helpful? Give feedback.
-
|
Ugh that explains the issue I had today. I thought I broke something. I’ll be doing the same thing as you, restoring. |
Beta Was this translation helpful? Give feedback.
-
|
Restoring the HA core to 2025.12.5 has resolved the issue for me. |
Beta Was this translation helpful? Give feedback.
-
|
Luckily, I use local LLMs running on a multi-gpu machine I also play games on. I use ollama and the gpt-oss 20b model. Works fine still. |
Beta Was this translation helpful? Give feedback.
-
|
I have the same but I didn’t enable tool servers for web searches on my local LLMs and so sometimes I want it to go out to the internet for when my kids asks something silly like the weight of an elephant or whatever |
Beta Was this translation helpful? Give feedback.
-
|
@JenniferHatches @cknic Planning to create the PR to merge the XVF3800 branch into main tomorrow or shortly after. I have updated the linux-voice-assistant-xvf3800.md (install and config document) and pushed some code changes that addressed some identified issues and general cleanup of the code. I really want to thank you both for all your help getting this branch started and stable. Couldn't/wouldn't have done it without you. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you, it's a nice step up in usability from where we started! I've switched over to the main branch, it's working well. It really seems to me like there's not much improvement to be done until we start training wake words and such on the users voice, as far as the voice assistant functionality goes. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
I think this repo currently has the best LVA client; I've been trying them all. I use respeaker vxf3800 microphones on my two lva machines, and I'd love it if you could integrate the LED controls from this fork, which uses lots of your work already:
https://github.com/cknic/linux-voice-assistant
Beta Was this translation helpful? Give feedback.
All reactions