R6: regime editing API + online editor + second-pass historical revision - #31
Merged
Merged
Conversation
Follows the 2026-06 audit. Periodization and anachronism fixes: Han Confucian orthodoxy dated to Emperor Wu with the early Western Han marked as Huang-Lao Daoist; the Qing Grand Council dated to 1729 with the Deliberative Council and Grand Secretariat named for the period before it; Shang ranks replaced with attested Shang titles instead of Zhou feudal grades, and coordination moved off the Zhou-era Qingshi Liao; Sui quantifies the division it ended. Attribution fixes: the Cao Cao hero speech marked as Sanguo Yanyi, a novel, not a historical source; L%27Etat c%27est moi flagged as apocryphal in both the Louis XIV soul file and the metadata system field. Office-power correction for Athens: the Strategia was the powerful classical office while the Archonship had become largely ceremonial and allotted. Seven further doubts are recorded in the report rather than silently rewritten - among them the Qin Nine Ministers as a Han-era reconstruction and the Manchu or Han temperament framing in the Qing soul file. IDENTITY.md and topology.json were deliberately untouched, so agentCount is unaffected; all 57 regimes still pass validate:regimes.
Edits metadata.json / IDENTITY.md / SOUL.md behind a validation chain that must pass in full before a single byte lands: id whitelist + safeResolve paths, the regime must already exist (edits only, no create/delete), metadata must parse, the IDENTITY role table must compile to >= 1 agent (AGENTS.md rule 2 — a prose rewrite silently yields 0 agents and bricks the regime), metadata.agentCount is resynced to the compiled count (rule 3), and topology validation must pass when topology.json exists. Writes go through temp file + rename with rollback, so a multi-file update is all-or-nothing. The regime cache is invalidated on success. Fixed during integration: the resynced agentCount was computed and returned but only persisted when the caller also sent a metadata payload — so editing just the IDENTITY table answered 200 with agentCount 3 while metadata.json kept saying 2, which would fail validate:regimes on the next run. metadata.json is now rewritten whenever the count drifts. The existing test always sent metadata, which is why it passed; added cases that edit IDENTITY alone and assert the on-disk value, plus one asserting no gratuitous metadata rewrite.
Editor tab edits metadata.json / IDENTITY.md / SOUL.md for the selected regime
through PUT /api/regimes/:region/:id. The IDENTITY box parses the role table as
you type and shows how many agents it would compile to; at zero it turns red and
disables save with an explanation, since a prose rewrite is what silently bricks
a regime (AGENTS.md rule 2). That preview is convenience only — the server
validation is still authoritative, and when it rejects, the error string and
every finding are surfaced verbatim rather than summarized.
SkillLibrary gains a Pending-approval section listing staged skills. Approval
itself stays on the CLI this round because no approval endpoint exists yet; the
button is disabled and says so instead of pretending to work.
Six vitest cases cover the zero-agent lockout, findings rendering, agentCount
refresh on success, and inline JSON syntax errors; the fetch mocks mirror the
real { id, metadata } envelope.
main's list gained the R7-R12 modules (skill-outcome, history-db, history-retriever); r6 added server/services/regime-write.mjs. The conflict was resolved by keeping main's list and inserting r6's entry, so neither side loses a syntax check. Note: `git merge-tree` reported zero conflict markers for this pair while the real merge conflicted on package.json — the newer merge-tree output format does not emit the markers that check grepped for. A conflict probe that returns a clean wrong answer is worse than no probe. CI on the merged tree: 519 backend + 25 frontend + 57 regimes + smoke.
LeoLin990405
pushed a commit
that referenced
this pull request
Aug 6, 2026
- wanikua#37: Replace static AUTH_TOKEN with dynamic getAuthToken() to fix 401 after login - #29: Change silijian agent id from 'main' to 'silijian' to avoid conflicts - #19: Remove deprecated fields (runTimeoutSeconds, maxConcurrent), fix api field, add troubleshooting - #20: Add --with-gui / --no-gui flag for optional Dashboard installation - #31: Add comments recommending openclaw CLI over direct config editing - #26: Add docs/windows-wsl.md for WSL2 setup guide - #25/#17: Add hanlinyuan (翰林院) agent as optional configuration
LeoLin990405
added a commit
that referenced
this pull request
Aug 6, 2026
…ion (#31) * content: second-pass historical revision across 10 regimes Follows the 2026-06 audit. Periodization and anachronism fixes: Han Confucian orthodoxy dated to Emperor Wu with the early Western Han marked as Huang-Lao Daoist; the Qing Grand Council dated to 1729 with the Deliberative Council and Grand Secretariat named for the period before it; Shang ranks replaced with attested Shang titles instead of Zhou feudal grades, and coordination moved off the Zhou-era Qingshi Liao; Sui quantifies the division it ended. Attribution fixes: the Cao Cao hero speech marked as Sanguo Yanyi, a novel, not a historical source; L%27Etat c%27est moi flagged as apocryphal in both the Louis XIV soul file and the metadata system field. Office-power correction for Athens: the Strategia was the powerful classical office while the Archonship had become largely ceremonial and allotted. Seven further doubts are recorded in the report rather than silently rewritten - among them the Qin Nine Ministers as a Han-era reconstruction and the Manchu or Han temperament framing in the Qing soul file. IDENTITY.md and topology.json were deliberately untouched, so agentCount is unaffected; all 57 regimes still pass validate:regimes. * feat(server): PUT /api/regimes/:region/:id — validated regime editing Edits metadata.json / IDENTITY.md / SOUL.md behind a validation chain that must pass in full before a single byte lands: id whitelist + safeResolve paths, the regime must already exist (edits only, no create/delete), metadata must parse, the IDENTITY role table must compile to >= 1 agent (AGENTS.md rule 2 — a prose rewrite silently yields 0 agents and bricks the regime), metadata.agentCount is resynced to the compiled count (rule 3), and topology validation must pass when topology.json exists. Writes go through temp file + rename with rollback, so a multi-file update is all-or-nothing. The regime cache is invalidated on success. Fixed during integration: the resynced agentCount was computed and returned but only persisted when the caller also sent a metadata payload — so editing just the IDENTITY table answered 200 with agentCount 3 while metadata.json kept saying 2, which would fail validate:regimes on the next run. metadata.json is now rewritten whenever the count drifts. The existing test always sent metadata, which is why it passed; added cases that edit IDENTITY alone and assert the on-disk value, plus one asserting no gratuitous metadata rewrite. * feat(frontend): regime editor tab and a read-only staged-skill view Editor tab edits metadata.json / IDENTITY.md / SOUL.md for the selected regime through PUT /api/regimes/:region/:id. The IDENTITY box parses the role table as you type and shows how many agents it would compile to; at zero it turns red and disables save with an explanation, since a prose rewrite is what silently bricks a regime (AGENTS.md rule 2). That preview is convenience only — the server validation is still authoritative, and when it rejects, the error string and every finding are surfaced verbatim rather than summarized. SkillLibrary gains a Pending-approval section listing staged skills. Approval itself stays on the CLI this round because no approval endpoint exists yet; the button is disabled and says so instead of pretending to work. Six vitest cases cover the zero-agent lockout, findings rendering, agentCount refresh on success, and inline JSON syntax errors; the fetch mocks mirror the real { id, metadata } envelope. * build: cover regime-write.mjs in lint:syntax
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Round 6, built in parallel by three workers in isolated worktrees. 321 backend + 20 frontend tests green; five-machine home-cluster CI green at
bcaba56(including a Node 26 / macOS arm64 runner).What's here
PUT /api/regimes/:region/:id— editsmetadata.json/IDENTITY.md/SOUL.mdbehind a validation chain that must pass in full before any byte lands: id whitelist +safeResolvepaths, regime must already exist (edits only), metadata must parse, the IDENTITY role table must compile to ≥1 agent,agentCountis resynced to the compiled count, topology validation runs whentopology.jsonexists. Temp-file + rename with rollback, so multi-file updates are all-or-nothing; the regime cache is invalidated on success.Editor tab — live agent-count preview while editing the IDENTITY table; at zero it disables save and explains why, since a prose rewrite is exactly what silently bricks a regime. Server rejections surface the error and every finding verbatim. SkillLibrary gains a read-only staged-skill list; approval stays on the CLI (no endpoint yet) with the button disabled rather than faked.
Historical revision, 10 regimes — periodization and anachronism fixes (Han Confucian orthodoxy dated to Emperor Wu with the early Western Han marked Huang-Lao; the Qing Grand Council dated to 1729; Shang ranks replaced with attested Shang titles rather than Zhou feudal grades), attribution fixes (the Cao Cao hero speech marked as a novel; "L'État, c'est moi" flagged apocryphal), and an office-power correction for Athens. Seven further doubts are recorded in the report rather than silently rewritten.
IDENTITY.mdandtopology.jsonuntouched, soagentCountis unaffected.Bug found during integration
The write API computed the resynced
agentCountand returned it, but only persisted it when the caller also sent a metadata payload — so editing just the IDENTITY table answered200 {agentCount: 3}whilemetadata.jsonkept saying2, which would failvalidate:regimeson the next run. The existing test always sent metadata, which is why it passed. Fixed, with cases that edit IDENTITY alone and assert the on-disk value; reverting the fix turns exactly that case red.Test plan
npm run ci— 321 backend, 57 regimes, 20 frontendbcaba56