Recorded demo · synthetic data
Test a reconciliation agent.
Your team defines the records, permitted tools and checks for a correct result. AgentEnv runs agents in that test environment and records the outcomes.
Inspect the recorded results
Two cases. Two correct outcomes.
This page replays existing results. Download the environment to run your own agent.
These are authored development cases, not a held-out benchmark. They do not establish production reliability or improvement over a rules baseline. No GTS systems or data were used.
Run it yourself
Keep the cases. Change the agent.
The dataset includes the runnable environment, thirteen authored cases, recorded evaluations and the application source. Start with the rules baseline below; it needs Python 3.11.4+, Git and Docker running.
python -m pip install 'agentenv-framework==0.9.1324'
python -m pip install 'agentenv-hf @ git+https://github.com/earakely-scale/agentenv-hf-plugin@c116bf1b321fa33d671632b0ee1b77406dbe3b58'
agent-env hf run AlexSchwartzman/TradeOpsSim-AgentEnv --bundle tradeops-baseline --task late_fill_duplicateThe baseline needs no model key. The tradeops-agent bundle supports a configured API model; follow its README for setup.