TraceArena v0.1.4: run one auditable multi-agent world

[tonyhyworld]

TraceArena v0.1.4 is now public. The goal is simple: let multiple agents act inside one constrained world, then preserve evidence → action → event → settlement → outcome as replayable data.

If you work on agent evaluation, tool use, orchestration, or scenario design, please run the no-key replay once:

• Demo: https://tonyworld888-tracearena-demo.static.hf.space/index.html
• Quickstart: https://github.com/tonyhyworld/TraceArena/blob/main/docs/quickstart.md
• First-run feedback: https://github.com/tonyhyworld/TraceArena/issues/new?template=first-run-feedback.yml
• Design-partner pilot: Design partner call: auditable agent evaluation pilots / 设计合作伙伴试点招募 #7

Please reply with only three things: time to first result, first confusing step, and whether the trace is useful for your work. No API key or private data is needed. This is a simulation/evaluation demo, not investment advice.

中文：欢迎运行一次无密钥回放，反馈首次结果耗时、最困惑的步骤，以及证据链对 Agent 评测/训练是否有价值。

[tonyhyworld]

Update: the public preview now includes deterministic replay evidence and a downloadable v0.1.6 bundle. The semantic digest stays stable across identical fixture runs while the full archive digest remains an integrity check.\n\nTry the browser-local demo: https://tonyworld888-tracearena-demo.static.hf.space/index.html\nRun locally: https://github.com/tonyhyworld/TraceArena/blob/main/docs/quickstart.md\nDownload the replay bundle: https://github.com/tonyhyworld/TraceArena/releases/tag/v0.1.6\n\nWe are looking for one concrete technical response: which world fact, action constraint, or settlement rule would you want to inspect first in an agent evaluation? Please reply here or propose a scenario pack at #2.