SkyEye
SkyEye is an AI-powered Air Intercept Controller (AIC) bot built by dharmab for DCS World. It replaces the stock in-game E-2/E-3/A-50 AWACS AI, which uses simplified, often inaccurate brevity and has an annoying cadence and voice.
SkyEye brings natural language processing (NLP) to DCS with speech recognition and natural-sounding text-to-speech locally, reading live aircraft data via Tacview’s ACMI telemetry feed, and talks to players over DCS-SRS radio frequencies. We run it on both Panther Den servers because it gives every pilot a real AIC on frequency at all times — accurate brevity, and automatic threat monitoring — without needing a human to sit in the AWACS seat for hours at a time.
Set your in-game name to the callsign format Callsign 1-1 | YourName. Here are all the commands you can make with SkyEye:
- RADIO CHECK to confirm the bot hears you and sees you on scope.
- ALPHA CHECK gives your current position in bullseye (BE) format
- BOGEY DOPE gives BRAA (bearing/range/altitude/aspect) to the nearest air threat
- DECLARE has you give a bullseye or BRAA, and the bot tells you hostile/friendly/furball/clean
- PICTURE gives you a broadcast call using BE format for all players
- SNAPLOCK is a fast BRAA-based IFF check for BVR timelines
- SPIKED/STROBE has you give a bearing to a radar spike, and the bot reports the likely source.
SkyEye also pushes automatic broadcast calls with no request needed:
- SUNRISE means controller online
- THREAT means hostile closing on a friendly
- MERGED means fixed-wing threat within 3nm
- FADED means a tracked hostile drops off scope inside weapons range — not a kill confirmation
SkyEye’s speech-to-text runs on an embedded Whisper model and we’re continually tuning our AI model to cut down transmission-to-response latency. The upstream project deliberately avoids using an LLM for the actual tactical logic — threat prioritization and BRAA math are deterministic and run in milliseconds — so our tuning work is focused entirely on making the voice recognition layer faster and more accurate, not on the decision-making, which is already near-instant.
Right now Panther Den runs two voices: Alan (masculine, British English) and Jenny (feminine, Irish English), both running on local neural TTS. As we bring more voice models online, we’ll rotate them in on “shifts” to keep things fresh and dynamic.
Example Comms
Section titled “Example Comms”DARKSTAR, CALLSIGN 1-1, RADIO CHECKDARKSTAR, CALLSIGN 1-1, ALPHA CHECKDARKSTAR, CALLSIGN 1-1, BOGEY DOPEDARKSTAR, CALLSIGN 1-1, DECLARE TWO THREE ZERO, TWELVE, TWELVE THOUSANDDARKSTAR, CALLSIGN 1-1, DECLARE BRAA ZERO NINE ZERO, TWENTY FIVEDARKSTAR, CALLSIGN 1-1, PICTUREDARKSTAR, CALLSIGN 1-1, SNAPLOCK ONE TWO FIVE, TEN, EIGHT THOUSANDDARKSTAR, CALLSIGN 1-1, SPIKED ZERO EIGHT ZERO