Releases¶
Reverse-chronological timeline of tagged BDPD releases. The
foundational release (v0.6) has a full release notebook with
manifests, regression invariants, and figures; later releases ship as
summary entries here plus their git tag, with the deep narrative
folded into the relevant paper.
| Tag | Date | Theme | What landed |
|---|---|---|---|
| v1.1.4 | 2026-06-29 | Static replay player | Server-less web/replay/ page: step through a recorded trace.json (commons / wealth / governance / cheap-talk / pacts), drag-drop offline or browse a served catalog with deep-links. kind:"world" multi-arena replay with cross-arena flow + per-round event markers; git-tracked curated showcase (scripts/curate_fixtures.mjs); --trace on the runner and the S2/S3 pilots. No engine / agent behaviour change. |
| v1.1.3 | 2026-06-23 | Verification method + docs hardening | New docs/method/ section (LLM-assisted research workflow: golden rules, audit-to-convergence, prompt library, panel / mimo-mem tooling); round-1/2/3 review fixes applied across BDPD⁰–³, the mini-course and the pitch decks; doc↔code reconciliation; repo hygiene (dropped JOSS/arXiv artefacts, tightened .gitignore). No code-behaviour change. |
| v1.1.2 | 2026-06-19 | Documentation & editorial release | No code change from v1.1.1 (engine / platform / agents / scripts / tests / tools bit-identical). Node >=20 baseline, transparency-statement harmonization, bibliography cleanup, citation-in-context fixes across the four papers, the mini-course and the blog book. |
| v1.1.1 | 2026-06-14 | Software Zenodo deposit | Tracked-only source snapshot + security scrub for the 9th Zenodo deposit (Software record, DOI 10.5281/zenodo.20687387). No behavioural code change from v1.1. |
| v1.1 | 2026-05-31 | Seneca polycentric governance | Bardi R/C/P substrate, RCP agent, pollution links, S0–S3 + OFAT robustness (paper BDPD³). E1 setPollution, E2 setCapital + capital_shock perturbations. |
| v1.0 | 2026-05-28 | Nested arenas / polycentric | World, Link, Treaty, exclude perturbation, world-level meta-agent; D1–D5 polycentric vignettes (paper BDPD²). Retagged at e4d658a for docs theme + BDPD¹ alignment. |
| v0.9.1 | 2026-05-28 | Paper BDPD¹ finalised | D1 / D2 / D3 promoted to \(N = 5\), monotone preservation ladder, D3 ladder-geometry sweep. |
| v0.9 | 2026-05-27 | Governance MVP | PactRegistry, MessageBus, cheap-talk surface (broadcast / announce_intended_harvest / propose_pact / accept_pact / pledge), 5 governance metrics, sanction and sanction_graduated perturbations, B1–C1–D1 pilots. See pacts, cheap-talk, governance metrics. |
| v0.8 | 2026-05-20 | Agent SDK | bdpd_sdk Python SDK (BDPDAgent, PlayerMemory, @tool, TraceWriter), DeepSeek v4 client, prompt-cache discipline, tool-loop with on_tool_call hook. See Agent SDK. |
| v0.7 | 2026-05-19 | Replication library | Paper-grade experiment runner, archive replication artefacts, B-series statistical inference report. |
| v0.6.5 | 2026-05-19 | Provenance & manifests | Per-run manifest hashes (sha256), PNG tEXt chunk provenance, regression invariants vs baseline. |
| v0.6 | 2026-05-19 | Open Platform | Registry refactors, plugin system, scheduler API, observability transforms, onboarding docs. Full release notebook: v0.6. |
Earlier development tags (v0.4.x, v0.5.x)¶
The pre-v0.6 history (v0.4.1, v0.5.0, v0.5.1) precedes the
documentation that ships with the platform today. The canonical
foundational paper (BDPD⁰) is calibrated against v0.6 and forward;
pre-v0.6 numerical results were superseded by the v0.6 regression
baseline (see tools/check_regression.sh).
Reading guide¶
- For the scientific narrative of each thematic release block, see the matching paper (BDPD⁰ = v0.6/v0.7, BDPD¹ = v0.9.1, BDPD² = v1.0, BDPD³ = v1.1). The papers are the durable artefact; the release notes are the technical changelog.
- For the full commit-level changelog, see
docs/dev/changelog.md. - For how to reproduce the experiments associated with each
release, see
docs/experiments/reproduce.md. - The regression baseline (5 canonical sweeps that all releases
≥ v0.6 must match byte-for-byte) is invoked by
npm testor./tools/check_regression.shand is the load-bearing guarantee that none of these releases silently broke a numerical result.