environments/
and can be copied with hud init; specialized agents, protocols, and training loops live in
cookbooks/. Entries without a
walkthrough are best read from their source.
Walkthroughs
Coding environment
Run a real SWE-bench task in an isolated repository workspace with hidden-test grading.
Source:
environments/coding.A2A chat
Serve a chat task over the A2A protocol and talk to it from any client.
Source:
cookbooks/a2a-chat.Ops diagnostics
An investigation task where the agent integrates evidence into a diagnosis.
Robot benchmark
Run a VLA policy against a containerized robot sim, graded by task success.
Fireworks Serverless RL
Train a Fireworks LoRA adapter on HUD environments, with grouped rewards,
calibration, checkpoints, and evaluation.
RL on Daytona sandboxes
Evals and GRPO training at 256 parallel sandboxes: 35.9% to 81.2% held-out
on a bug-fixing task.
More runnable examples
These ship in the repo without a separate walkthrough. Read the README in each directory to run them.RL training
On-policy RL: roll out a taskset with the current weights, train on the
resulting trajectories, and serve the updated weights for the next rollout,
all under one trainable model string.
Connect Four self-play
Symmetric self-play GRPO on a 6×7 Connect Four board, training both sides
from a single rollout.