Releases: tonyhyworld/TraceArena
Release list
TraceArena v0.1.12 — Investment Agent Benchmark v1
TraceArena v0.1.12 — Investment Agent Benchmark v1
TraceArena now ships a reproducible evaluation contract for its flagship capital-market world.
What is new
- Investment Agent Benchmark v1 contract baseline with a checked-in JSON report and readable leaderboard.
- Evidence → decision → simulated order → deterministic settlement evaluation across return, benchmark excess return, maximum drawdown, turnover, and transaction costs.
- Integrity verification for fixture, full replay, semantic replay, and generated report.
- Cross-platform reproduction in CI on Ubuntu and macOS.
- English and Chinese benchmark cards linked from both READMEs.
Reproduce it without an API key
Fastest path: open the pinned Colab notebook. It runs v0.1.12 and verifies the generated report against the published baseline.
PYTHONPATH=backend python backend/scripts/investment_benchmark.py \
--output runs/investment_agent_benchmark_v1The run disables network access and brokerage execution. It uses a synthetic fixture and a simulated portfolio ledger.
Important claim boundary
The two bundled entrants are deterministic scripted controls. This release proves the benchmark contract; it is not yet a model leaderboard. A model result will only be promoted when its provider/model identity, fixture and scenario digests, seeds, role rotation, complete replay, and verifier output are available.
The synthetic fixture remains marked internal_synthetic_pending_approval, and its proposed CC0-1.0 dedication is still pending maintainer confirmation. The report exposes this status explicitly.
Evidence bundle
Download tracearena-v0.1.12-investment-benchmark-evidence.zip to inspect the bilingual benchmark cards, report, leaderboard, run manifest, summary, and authoritative replay.
SHA-256: 353026af03de551020ce927a0579ebcebd1f9836772a35127ee5d30ff426496b
No financial advice. No real orders. No brokerage connection.
本版本发布首个投资智能体评测协议基线：可离线复现、可校验、可回放，并同时报告收益、超额收益、最大回撤、换手率与交易成本。当前两名参赛者是脚本控制组，不是模型排行榜；网络和券商均禁用，不构成投资建议。
TraceArena v0.1.11 — Professional bilingual evaluation
Professional bilingual evaluation
TraceArena v0.1.11 turns the Capital Market Public Edition into a professional, bilingual evaluation experience.
Highlights
- complete English and Simplified Chinese presentation across the viewer, operator console, archives, model analysis, training-data export, and user management;
- non-disruptive language switching: the active run keeps its tick, evidence, and portfolio state, while the next reset adopts the selected Agent language;
- localized capital-market roles, actions, settlement explanations, and operator schema;
- automated localization contract coverage plus green frontend and public-runtime CI;
- repository hardening with least-privilege workflow tokens, immutable Action pins, Dependabot updates, malware alerts, and private vulnerability reporting.
Start here
The public scenario uses simulated settlement, does not connect to a brokerage, and is not investment advice.
TraceArena v0.1.10 — Capital Market Public Edition
Capital Market Public Edition
TraceArena v0.1.10 publishes the first public investment-decision scenario pack.
What is included
backend/scenarios/capital_market/as a loadable Public Edition scenario pack;- two competing investment agents, shared constraints, evidence requirements and simulated portfolio settlement;
- no-key deterministic replay for a first run;
- optional BYO model provider and authorized read-only market research tools;
- bilingual scenario orientation and public deployment/first-login guidance;
- Apache-2.0 scenario metadata with asset provenance.
Safety boundary
This release does not connect to a brokerage, submit real orders, ship credentials,
or provide investment advice. Market and financial data integrations are optional,
must be configured by the user, and remain subject to each provider's terms and
redistribution limits.
Start with the Quickstart,
then read the Capital Market Public Edition guide.
TraceArena v0.1.9 — Demo narrative consistency
TraceArena v0.1.9 — Demo narrative consistency
This patch release keeps the public demo aligned with the Physical World OS positioning.
- Replaced the remaining HUD label
DATA FACTORYwithVERIFIED EVIDENCE. - Reframed the pipeline label as a decision-evidence pipeline.
- Updated the finale to emphasize worlds, rules, competing paths, and verifiable evidence.
Live demo: https://tonyworld888-tracearena-demo.static.hf.space/index.html
TraceArena v0.1.8 — Physical World OS direction
TraceArena v0.1.8 — Physical World OS direction
TraceArena now presents a sharper public preview of its core idea: an open-source runtime for loading physical-world contracts—goals, resources, rules, tools, and settlement authority—then letting agents compete over paths and execute toward outcomes with replayable evidence.
What changed
- Reframed the README, Chinese README, demo narrative, and evidence factory around the Physical World OS direction.
- Clarified that TraceArena is a world runtime and decision-evidence system, not a generic training-data catalog.
- Added a square launch thumbnail for community and directory listings.
- Added a concrete statement-of-work template and a bounded evaluation-pilot path for teams.
- Added a direct scenario-pack contribution path with English/Chinese guides, template links, and acceptance criteria.
- Recorded green CI, frontend build, smoke-test, and OpenSSF Scorecard verification for the public preview.
Start here
- Live demo: https://tonyworld888-tracearena-demo.static.hf.space/index.html
- Quickstart: https://github.com/tonyhyworld/TraceArena/blob/main/docs/quickstart.md
- Propose a scenario pack: #2
- Request an evaluation pilot: #7
This release remains a public preview. The deterministic, no-key replay path is the recommended first run; external integrations and scenario-pack APIs will continue to evolve with community feedback.
TraceArena v0.1.7 — Incident Response Benchmark
What is new
- Added the deterministic, synthetic Incident Response World.
- Added a runnable fixture with an explicitly rejected premature resolution.
- Added the Incident Response Benchmark Card with evaluation questions, metrics, settlement authority, and limitations.
- Added a focused replay test and no-key local commands.
Reproduce
python3 examples/incident_response_world/scripts/replay.py \\
examples/incident_response_world/tests/fixture.json
python3 -m unittest discover -s examples/incident_response_world/tests -p 'test_*.py'The fixture is synthetic, does not contact production systems, and is not a claim about production incident-response performance.
TraceArena v0.1.6 — Deterministic replay and demo sharing
This release improves the first-run trust and sharing path.
Highlights:
- Adds deterministic_replay_sha256 for cross-run semantic comparison while retaining canonical_replay_sha256 as the full-file integrity digest.
- Adds regression coverage for repeated no-key Market Replay runs.
- Updates Quickstart and pilot acceptance language.
- Adds social preview metadata and RUN LOCALLY / BUILD A WORLD links to the static AI World demo source.
- Includes a downloadable synthetic replay evidence bundle: tracearena-v0.1.6-replay.zip.
The public replay remains synthetic, deterministic, no-key, offline, and not financial advice.
TraceArena v0.1.5 — Adoption and onboarding polish
This release improves the path from first discovery to a reproducible first run and a bounded pilot conversation.
Highlights:
- fixed and audited public Markdown links (clean audit: 0 missing internal links);
- published the public candidate scope and qualified-lead log template;
- added permission-aware community distribution targets and targeted outreach drafts;
- improved macOS/Linux and Windows installer diagnostics, including explicit Python interpreter selection;
- linked the v0.1.4 technical discussion from the English and Chinese READMEs;
- kept the HF Demo footer and README connected to feedback and pilot paths.
Start here:
- Demo: https://tonyworld888-tracearena-demo.static.hf.space/index.html
- Quickstart: https://github.com/tonyhyworld/TraceArena/blob/main/docs/quickstart.md
- Technical discussion: #8
- Pilot: #7
This remains a public preview. It does not claim external adoption, revenue, model performance, investment returns, or business outcomes.
TraceArena v0.1.4 — Design partner pilot path
This release makes the first external adoption and commercial pilot path explicit.
Highlights:
- added the Agent Evaluation Pilot one-pager with scope, acceptance criteria, and starting price;
- added bilingual founder outreach messages and an auditable outreach queue;
- added the public design-partner call: #7;
- linked the pilot path from README, founder profile, commercial support, and launch content.
Start here:
- Demo: https://tonyworld888-tracearena-demo.static.hf.space/index.html
- Quickstart: https://github.com/tonyhyworld/TraceArena/blob/main/docs/quickstart.md
- Pilot one-pager: https://github.com/tonyhyworld/TraceArena/blob/main/docs/open_source_release/PILOT_ONE_PAGER.md
This is a public preview. It does not claim customer adoption, revenue, model performance, investment returns, or business outcomes.
TraceArena v0.1.3
First release from the rebuilt public repository.
Highlights:
- New bilingual product documentation focused on auditable multi-agent worlds, verifiable outcomes, and reusable scenario packs.
- A visual runtime lifecycle and scenario-pack authoring guides in English and Simplified Chinese.
- Local self-hosted developer console, deterministic no-key market replay, and offline artifact Viewer.
- Copyright notice updated to 张诺亚.
Start here:
- Run the one-command Quickstart, or open the live demo.
- Join the first 10 runs discussion and submit the first-run feedback form.
- Propose a domain scenario pack through Issue #2.
- For a bounded enterprise evaluation, use the pilot request form.
Safety boundary:
The bundled capital-market example uses synthetic fixture data and a simulated ledger. It does not place real orders and is not investment advice. The no-login console is localhost-only; public deployment requires authentication, authorization, secret storage, and audit integrations.