Skip to content

feat(L): all 57 regimes rewritten to canonical IDENTITY template#6

Merged
LeoLin990405 merged 2 commits intomainfrom
feat/l-full-regime-rewrite
Apr 14, 2026
Merged

feat(L): all 57 regimes rewritten to canonical IDENTITY template#6
LeoLin990405 merged 2 commits intomainfrom
feat/l-full-regime-rewrite

Conversation

@LeoLin990405
Copy link
Copy Markdown
Owner

What

Full rewrite of every regime's IDENTITY.md to a canonical structure so each civilization's Agent Team authentically mirrors its historical governance.

How

Parallel delegation across 9 AI workers (7 CN backends + Codex + Gemini). Each worker was given:

  • The canonical template (docs/IDENTITY-TEMPLATE.md)
  • Tang as gold exemplar
  • Regime metadata.json + existing IDENTITY.md for reference

Wall time: ~15 min for 56 regimes + 2 retry rounds.

Canonical structure

Every IDENTITY.md now has:

  1. 制度简介 (System Overview) — era, duration, significance
  2. 组织架构图 (Organization Chart) — ASCII hierarchy, 3–10 organs
  3. 角色映射表 (Role Mapping Table) — ≥5 rows with canonical AI roles
  4. 决策流程 / 协作流程 (Decision Flow) — step-by-step governance flow
  5. 制度特点 (Characteristics) — mechanism-level bullets
  6. Pattern 映射 — orchestration pattern from 6 canonical options
  7. 历史参考 (Historical Sources) — 3–5 references

Attribution by worker

Worker Regimes Notes
cc-mimo byzantine, persian, ottoman, mongol, russian, soviet, ming, qing, north-south, western-xia, napoleon, us-federal 1M ctx, archive-heavy civs + retries
cc-glm song, yuan, han, sui, five-dynasties, three-kingdoms Chinese reasoning
cc-qwen xia, shang, zhou, qin, jin Early Chinese dynasties
cc-doubao taiping, liao, jin-jurchen, roc, viking General coding
cc-stepfun sparta, prussia, zulu, meiji, maurya, inca Discipline/logic civs
cc-minimax carthage, venice, swiss, aztec, mali, sumeria Fast inference, smaller polities
codex roman-republic, roman-empire, british European structures
gemini athens, caliphate, egypt, shogunate, habsburg, khmer, safavid, hre, joseon, mughal, polish Cross-continental + retries

Rejection incidents

  • cc-kimi returned {"error":{"type":"invalid_request_error","message":"The request was rejected because it was considered high risk"}} for all 6 assigned regimes (ming/qing/joseon/polish/hre/mughal). Content filter triggered by political governance topics. Reassigned to cc-mimo + gemini.
  • codex timed out mid-write on napoleon/us-federal. Reassigned to cc-mimo.

Test plan

  • npm run validate:regimes — 0 errors, 40 alias warnings (expected)
  • All 57 regimes have required canonical H2 sections
  • Role mapping tables have ≥5 rows each
  • npm test — 9/9 unit tests pass
  • Deferred: historical accuracy review (sample 5-10 regimes with a human historian)
  • Deferred: end-to-end tournament run to verify agent counts match new tables

Risks

  • AI-authored historical content may have inaccuracies. The provenance is recorded in commit message; future corrections can target specific regimes.
  • All 7 CN backends' outputs vary in style/depth. Visual consistency is not guaranteed.
  • No single human reviewed all 57 files; quality varies by worker and by regime.

Parallel rewrite delegated across 9 AI workers (+ 3 retry rounds):
- cc-mimo  (1M ctx):  persian, ottoman, mongol, russian, soviet, byzantine,
                      + retries: ming, qing, north-south, western-xia,
                      + final: napoleon, us-federal
- cc-glm:             song, yuan, han, sui, five-dynasties, three-kingdoms
- cc-qwen:            xia, shang, zhou, qin, jin
- cc-doubao:          taiping, liao, jin-jurchen, roc, viking
- cc-stepfun:         sparta, prussia, zulu, meiji, maurya, inca
- cc-minimax:         carthage, venice, swiss, aztec, mali, sumeria
- codex (GPT-5.4):    roman-republic, roman-empire, british
- gemini (2.5 Pro):   caliphate, egypt, shogunate, habsburg, khmer, safavid,
                      athens, hre, joseon, mughal, polish

cc-kimi rejected all 6 (content filter on political topics) — reassigned.

Each IDENTITY.md now has required sections:
- 制度简介 (System Overview)
- 组织架构图 (Organization Chart) — ASCII hierarchy, 3-10 organs
- 角色映射表 (Role Mapping Table) — ≥5 rows with canonical AI roles
- 决策流程 (Decision Flow) or 协作流程 (Workflow — accepted alias)
- 制度特点 (Characteristics)
- Pattern 映射
- 历史参考 (Historical Sources)

Validator (test/regime-validator.mjs) enforces structure.
Canonical template: docs/IDENTITY-TEMPLATE.md

All 57 regimes pass npm run validate:regimes (0 errors, only alias warnings).
- README.md: comprehensive v5 rewrite (345 lines) — 30s quickstart,
  learning-loop diagram, 10-model matrix, 57 regime index, 6 modes table,
  architecture diagram, full CLI reference, v5 workflow walkthrough,
  tournament mode, CI/testing, extension guide, design limits, credits
- regimes/china/tang/IDENTITY.md: remove 司礼监/批红 (Ming-Qing anachronisms
  misplaced in Tang); replace with 中书舍人 (Zhongshu Sheren — actual Tang
  drafter office). Clarify '1300 years' claim — only 三省六部 framework
  persisted, not Tang's specific configuration.
- regimes/REVIEW-FINDINGS-v5.md: Codex audit findings, documenting what
  was fixed and what deferred to v5.1 (byzantine/roman-republic/prussia/
  ottoman have accuracy gaps identified but not yet patched).
@LeoLin990405 LeoLin990405 merged commit 4a8a6b2 into main Apr 14, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant