BotVerse

AI Agent Training Platform · PPO · DQN · SAC · A2C

Active runs
4
Agents trained
218
Inference RPS
42.3 /s
PPO · stable-baselines3
Live
pathing · pathing-tutorial env
AlgoPPO
Steps14,300
Episode42
Status● training
Last reward
+8.42
Best ep
+12.74
+10
Mean reward (100ep)
8.21 +0.34
Episode length
62.4
Policy loss
0.0143
Value loss
0.0091
Entropy
0.671
Active Training Runs 4 running · GPU: 87%
Python SDK · Inference API 60 req/min/token
from botverse import BotVerseClient
client = BotVerseClient(api_token="botv_live_...")
policy = client.load_policy("path-ppo-7c2a") # schema validated locally
for obs in env_loop():
action = policy(obs) # inference: ~12ms p50
ELO Leaderboard k=32 · pathing
Active Tournament round 2/3
🏆
Pathing Pro Cup
8 agents · single-elim · prize 250 ⬡
live
path-ppo-7c2a9.21
path-dqn-4b186.34
path-sac-0e91 winner11.04
path-a2c-22d87.88
Reward · last 100 episodes
+8.21 μ 7.85 · σ 1.42
Environments
pathing
grid_world
physics_sandbox
strategy_arena
obstacle_course