intruder11

The macOS version of SkyEye uses features of macOS (Metal, Speech Synthesis Manager) and the Apple Silicon CPUs (M1/M2/M3/M4) that are not available on the other operating systems. The parts of the code that make SkyEye great on macOS will therefore never be available on Windows and Linux. In theory, I could achieve similar levels of speech recognition performance for people who own certain Nvidia graphics cards (and on Linux, use the proprietary driver from Nvidia instead of the open Nouveau driver) by using CUDA. However, I don't own any Nvidia hardware to test on. That might change late this year/early next year if I build another rig, but for now all my Windows/Linux machines use AMD GPUs. The Speech Synthesis Manager has no direct equivalent on any other OS - technology of that quality simply does not exist outside of macOS. I've tried literally every alternative available. Even the hosted speech models from companies like OpenAI don't compare. (These Apple CPUs are so good. I hadn't bought a laptop for myself in over ten years - after work provisioned me an M4 Macbook, I went out and bought two for myself, plus a Mac Mini that's currently running SkyEye 24/7 for Flashpoint Levant. They're amazing hardware, the best non-gaming computers you can buy for the money IMO.)

I've released 1.4.4. This contains improvements to speech recognition accuracy. https://github.com/dharmab/skyeye/releases/tag/v1.4.4

Added a section on how to exclude the HUD and MFDs from DLSS as of the current patch: https://www.8492sqdn.net/guides/dcs/performance/#dlss My collaborator and I are working on a major update to the VR section for later in the year. We will be buying both Bigscreen and Pimax's new headsets with our own money to provide our honest information and updated tweak guides.

Nope that was Pene from DCS ANZUS!!

I'm sorry, who are you talking to?

I've released 1.4.3, fixing an issue where sometimes broadcast calls in situations where some, but not all, of a flight's members were the target had the callsigns out of order ("2 3, EAGLE 2 1" instead of "EAGLE 2 1, 2 3") https://github.com/dharmab/skyeye/releases/tag/v1.4.3 (1.4.2 was a small dependency update for a very niche use case far off the beaten path, so I didn't bother telling anyone except the one person who reported the issue)

I've released 1.4.1, fixing a regression in cloud speech recognition: https://github.com/dharmab/skyeye/releases/tag/v1.4.1

You can now easily install SkyEye on macOS using Homebrew: https://github.com/dharmab/skyeye/blob/main/docs/ADMIN.md#installation-with-homebrew-recommended EDIT: If your server is on LAN turns out macOS has a "feature" that prevents SkyEye from connecting... working on a fix... https://github.com/dharmab/skyeye/issues/566

I've released v1.4.0 SkyEye's macOS build has been fully overhauled in this release and is no longer marked as experimental; indeed, the macOS version of SkyEye is now the best version of SkyEye. I recommend you hop on the Flashpoint Levant server, which is already using the macOS version, and check out the near-instant speech recognition, and the new AI voice that is nearly indistinguishable from a human voice. Unfortunately the macOS build is not easy to install or deploy yet; you can currently manually run it through Terminal, but my eventual goal is to provide an easy installation through Homebrew. That's currently held up because the build process currently itself has a dependency on Homebrew, which creates a circular dependency... I may need to think of a creative solution, haha! Changelist: * The macOS version of SkyEye now supports GPU acceleration using Apple Silicon and the Neural Engine. In my benchmarks, the macOS version recognizes speech 3-10x faster than OpenAI's API, and about 50-100x faster than my Windows/Linux PCs. * The macOS version of SkyEye can now use Apple's Speech Synthesis Manager to speak with realistic AI voices. This requires some additional setup; if this additional setup is not completed, SkyEye falls back to the classic Siri voice from the iPhone 4S/iPhone 5/iPhone 6. * Fix the `auto-picture` flag not being respected. * Fix a bug where unarmed aircraft triggered threat callouts. * Improvements to speech recognition accuracy. * Other minor internal code improvements. https://github.com/dharmab/skyeye/releases/tag/v1.4.0

Thanks for pointing this out. This actually regressed all the way back in v1.2.1: https://github.com/dharmab/skyeye/pull/523/files#diff-243ebed2765f75e6a54f57167212fefb08c3b2a85967ad2acbc0eb78919019c1L129, I'll fix it when I have a free moment. In the meantime you can work around it by setting the auto picture interval to a long time like "24h". What OS are you running on? On Linux, I'd use the existing skyeye.service systemd unit as a starting point and change it into a "service template" (https://www.freedesktop.org/software/systemd/man/latest/systemd.service.html#Service Templates). That's a systemd-native way of running multiple instances of a background service. You could then have each instance of SkyEye point at a different config file in `/etc/skyeye/`. On Windows, the easiest way is probably to make a second copy of the skyeye-service.exe and skyeye-service.yml files (e.g. skyeye-instance-2-service.exe and skyeye-instance-2-service.yml), edit the YAML file to change the id, name and arguments fields, and repeat the install steps from the admin guide using this second copy. See also the reference docs for WinSW v2: https://github.com/winsw/winsw/blob/v2.12.0/doc/yamlConfigFile.md

Here is the newsletter for April: https://www.8492sqdn.net/posts/newsletter-2025-04/

I've released v1.3.0, with two new features: - SkyEye now supports directing threat and merged calls to flight rather than individually, if players conform to USAF-style callsign formats. Where before your flight might have heard "Chevy 11, Chevy 12, Chevy 13, Chevy 14, group threat..." they can now hear "Chevy 1 flight, group threat...". For more details, see the documentation on Strict Callsigns. - Added support for the GPT4o and GPT4o-mini STT models available through OpenAI's platform. However, I don't recommend using them yet. SkyEye uses some features in the API that are not yet available for these new models, so in practice the current Whisper model works better at this time. - Other documentation updates.

I've updated the guide, especially the section on VR with guidance on navigating the various headsets launching in 2025.

I've released v1.2.3. * Prior to this release, Discord tracing was enabled by a `tracing` config option, but documentation and examples referred to a non-existent `enable-tracing` option. The code now accepts either `tracing` or `enable-tracing`. * Controller now omits HIGH/LOW fill-in if a STACK has only two contacts, since 1 HIGH 1 LOW is implicit in that case * Added missing data for the Tu-160 to SkyEye's internal aircraft database. Controller no longer considers the Tu-160 a potential threat to friendly aircraft, and will use the reporting name "Blackjack" * The fact that SRS must have External AWACS Mode configured is now documented and SkyEye will emit error messages if it detects EAM has been disabled by the server. * Minor internal code improvements. https://github.com/dharmab/skyeye/releases/tag/v1.2.3

Hi, as mentioned in the Getting Help section of the admin guide, please open issues on GitHub rather than using social media. This is important for 2 reasons: 1. The GitHub issue tracker has a form which collects data I need to troubleshoot, such as the full log file and information about your deployment. 2. I don't monitor social media/the forums, I only come here when I have something to say.

Released 1.2.2 with a bugfix for a crash that could occur when responding to certain DECLARE requests. https://github.com/dharmab/skyeye/releases/tag/v1.2.2

Released 1.2.1 with a bugfix for a rare race condition. It's not a new bug, been in there for a while, but was really unlikely, so I didn't get any reports until today. https://github.com/dharmab/skyeye/releases/tag/v1.2.1

Here is the March newsletter: https://www.8492sqdn.net/posts/newsletter-2025-03/

I've released 1.2.0. New feature: SkyEye supports optional file-based locking to support a use case for some squadrons where multiple SkyEye instances are running on the same CPU. This allows each instance to use the CPU one-at-a-time during the processes' heaviest bottlenecks which is a better user experience than multiple instances flooding the CPU at the same exact moment. Learn more: https://github.com/dharmab/skyeye/blob/main/docs/ADMIN.md#multiple-instances-experimental Bugfixes: - Further improvement to the FADED call chaining issue mentioned in my last message. - Fix a bug where the controller would call "merged" when players flew near certain stationary hostile aircraft on the ground.

I've released v1.1.5. This is a small bugfix update: - Fix a bug that caused the GCI's voice to play 8% faster/higher pitched than intended (oops!) - When a hostile aircraft disappears, the bot will wait a short while to see if other hostile aircraft also disappear so it can provide a combined FADED call for multiple aircraft in a group disappearing at around the same time. In some cases where players got "killstreaks" this would chain for a long time and a long wave of FADED calls would then be sent all at once. Adjusted the "killstreak timer" to reduce the occurrence of this. - Updated to Go 1.24 (https://go.dev/blog/go1.24); a couple of the new features are being used for development, but this is otherwise an invisible upgrade for users. https://github.com/dharmab/skyeye/releases/tag/v1.1.5

You can also add extra weight to the AI via lua scripting (setUnitInternalCargo trigger action) to reduce their effective performance.

I've released v1.1.4. This contains a notable change - clan/unit tags within callsigns are now ignored. For example, "[ISAF] Mobius 1" was previously parsed as "ISAF Mobius 1", and is now parsed as "Mobius 1". Other changes: - DECLARE responses now include fill-ins. Most notable, FURBALL responses will now include the number of friendlies and hostiles in the furball. - Reduced how often the warning message is logged about SkyEye's IGRF model (magnetic declination correction) in missions set in 2024 and later. (This isn't a problem at this time because DCS World only uses data through 2022, but could be a problem if ED ever ships new data.) - Minor improvements to parser.

Here is the newsletter for February: https://www.8492sqdn.net/posts/newsletter-2025-02/

I've released v1.1.3. Improve parsing of longer multi-word request keywords like BOGEY DOPE and ALPHA CHECK Fix an issue with parsing SPIKED requests, parsing of these should be more lenient SPIKED responses now have the correct fill-ins If a player requests a DECLARE but didn't explicitly specify BULLSEYE or BRAA, the controller will now read back the BULLSEYE coordinate to hint the player that BULLSEYE is assumed. Other minor improvements to parser https://github.com/dharmab/skyeye/releases/tag/v1.1.3

I've released v1.1.2. This is a minor update that improves the voice recognition success rate slightly. https://github.com/dharmab/skyeye/releases/tag/v1.1.2