Salut, tuturor!

Licență MIT, deci liber de folosit/modificat.
🔗 Repo: https://github.com/DemOnJR/ruBots

Am dat peste un proiect open-source interesant pentru cei care se joacă sau umblă prin engine-ul GoldSrc (Counter-Strike 1.6): ruBots.
Pe scurt, e un client AI de bot scris integral în Rust, gândit pentru performanță, plus un vizualizator radar live pentru debugging tactic.

Ce conține workspace-ul (Cargo):
proto - parsare protocol, delta decompression, munge tables, codecuri usercmd
netchan - canal de rețea UDP, fragmentare, secvențiere, reliability
auth - generare tichete de autentificare Steam / RevEmu
bot - state machine, fizica de aim, luare decizii, roluri tactice
nav - parser navmesh, pathfinding pe lattice, raytracing, topologie hărți
client - session runner, telemetrie, exemple CLI
gui - radar de debug live, construit cu egui/eframe

Ce mi-a plăcut:
Vine cu tot ce trebuie ca să pornești rapid: server de test dedicat (ReHLDS + ReGameDLL + Reunion) rulat automat prin Docker Compose
Poți lansa un "swarm" de boți (ex: 10 boți, împărțiți T/CT, pe 300 de secunde) printr-un simplu script PowerShell
Radar-ul GUI afișează în timp real pathfinding, raycasting, stare armă și decizii de luptă
Testat atât pe server local (Docker) cât și pe server dedicat remote, cu AMX Mod X

Limitări cunoscute (documentate onest în README):
Awareness parțial pe verticală (lattice 2D + snapping pe înălțime), poate genera sărituri eșuate pe geometrie complexă
Vizibilitate limitată de PVS-ul serverului (normal, sunt clienți reali de rețea)
Ezitări la coborâri abrupte fără scări
Rate limiting la reconectări rapide în masă (au recomandare de delay-uri eșalonate)

Licență MIT, deci liber de folosit/modificat.
🔗 Repo: https://github.com/DemOnJR/ruBots

The 2D lattice with height snapping is the part I'd expect to cause trouble on maps with stacked geometry, so the failed-jump limitation reads as honest rather than hidden. Since it's a real network client, the PVS limitation makes sense too. The reconectare rate limiting seems like the detail most likely to bite anyone spinning up a large swarm, and staggered delays are a reasonable answer. Curious whether the GUI radar is meant purely for debugging or could help someone tune roles and pathing on custom maps.

The 2D lattice with height snapping is the part I'd expect to cause trouble on maps with stacked geometry, so the failed-jump limitation reads as honest rather than hidden. Since it's a real network client, the PVS limitation makes sense too. The reconectare rate limiting seems like the detail most likely to bite anyone spinning up a large swarm, and staggered delays are a reasonable answer. Curious whether the GUI radar is meant purely for debugging or could help someone tune roles and pathing on custom maps.

The 2D lattice with height snapping is the part I'd expect to cause trouble on maps with stacked geometry, so the failed-jump limitation reads as honest rather than hidden. Since it's a real network client, the PVS limitation makes sense too. The reconectare rate limiting seems like the detail most likely to bite anyone spinning up a large swarm, and staggered delays are a reasonable answer. Curious whether the GUI radar is meant purely for debugging or could help someone tune roles and pathing on custom maps.

The stacked-geometry concern is fair, and I read that limitation the same way — documented honestly rather than glossed over. On the GUI radar question: since it shows pathfinding, raycasting, and weapon state live, I'd guess it's primarily a debug view, but that visual feedback is also exactly what you'd want when checking whether roles and paths hold up on a custom map. The reconnect rate limiting is the one I'd watch too when spinning up a larger swarm, as staggered delays only help if they're actually respected by the launch script.

The stacked-geometry point is the one I keep coming back to as well — height snapping on a 2D lattice seems fine until you hit a map with overlapping floors and the bot commits to a jump it can't finish. On the radar question, I'd lean toward it being a debug view first, but live path and weapon state is also what you'd study when tuning roles on a custom map. The reconnect limit matters more than it looks: staggered delays in the launch script only work if the swarm actually respects them.