Add main publication to new entries (IsMain)

.claude/agents/i18n-completeness.md

@@ -1,6 +1,6 @@
 ---
 name: i18n-completeness
-description: Verify the i18n extraction workflow completed end-to-end when new user-visible strings are added in frontend/viewer/**. Extraction freshness, .po file consistency, context comments per I18N_CONTEXT_GUIDE.md. Distinct from viewer-watcher's parser-awareness check.
+description: Verify the i18n extraction workflow completed end-to-end when new user-visible strings are added in frontend/viewer/**. Extraction freshness and .po file consistency. Distinct from viewer-watcher's parser-awareness check.
 tools: Bash, Read, Grep, Glob
 model: sonnet
 ---
@@ -19,9 +19,11 @@ detected; nothing to verify" and stop.
 
 ## Baseline
 
-Read `frontend/viewer/I18N_CONTEXT_GUIDE.md` for the context-comment
-expectations. `frontend/viewer/AGENTS.md` §i18n has the extraction
-workflow.
+`frontend/viewer/AGENTS.md` §i18n has the extraction workflow.
+
+Translator-context (`#.`) comments are **out of scope** — they're owned
+by the `crowdin-merge` skill (via the `i18n-context-writer` agent). Do
+not check for, suggest, or flag them, even informationally.
 
 ## Standards
 
@@ -38,26 +40,7 @@ For each new `$t\`...\`` or `msg\`...\`` introduced by the diff:
 Grep `frontend/viewer/src/locales/en.po` for each new string's
 msgid form.
 
-### B. Context comments
-
-`I18N_CONTEXT_GUIDE.md` requires context comments on strings that:
-- Differ between Classic and Lite views.
-- Are ambiguous in isolation ("Open" — a verb? a noun?).
-- Contain user-supplied variables.
-- Could be confused with similar terms elsewhere.
-
-For each new string, ask: does it need a context comment? If yes and
-absent → 💭 nit; suggest one.
-
-Context-comment syntax in source:
-```typescript
-// i18n: <context>
-$t`Open`
-```
-
-Or in `msg\`...\`` calls via the `context` option.
-
-### C. Cross-locale state
+### B. Cross-locale state
 
 When extraction runs, it adds the new msgid to all `.po` files (as
 untranslated). Spot-check:
@@ -69,14 +52,14 @@ untranslated). Spot-check:
 Missing msgid in non-source locales → 💭 nit (Lingui usually syncs them
 on next extract; might be a partial run).
 
-### D. Don't translate placeholder strings
+### C. Don't translate placeholder strings
 
 If a string only appears in dev / test code (Storybook stories,
 `*.test.ts`, dev-only debug routes) → flag as 💭 nit: *"this string is
 extracted but only used in dev — wrap in `// i18n-ignore` or scope to
 non-extracted context."*
 
-### E. Pluralization & ICU
+### D. Pluralization & ICU
 
 Plural forms use Lingui's plural component / API:
 ```typescript
@@ -86,7 +69,7 @@ $t`${count, plural, one {# entry} other {# entries}}`
 A new user-visible count without pluralization handling → ⚠️ important
 *"will read 'You have 1 entries' for count=1"*.
 
-### F. Don't compose translated strings at runtime
+### E. Don't compose translated strings at runtime
 
 Even if a translatable phrase looks like literals at extraction time,
 concatenating them at runtime breaks grammar in other languages:
@@ -109,12 +92,10 @@ const label = $t`${noun} modified by ${author}`;
 - `src/locales/en.po` — check each new msgid is present.
 - `src/locales/*.po` — spot-check other locales were extracted
   too.
-- `// i18n:` comment density vs new-string density.
 
 ## Severity quick map
 
 - New strings without extraction run → ⚠️ important.
-- Missing context comment on ambiguous string → 💭 nit.
 - Missing plural form on count-bearing string → ⚠️ important.
 - Runtime-composed translated string → ⚠️ important.
 - Dev-only string extracted → 💭 nit.

AGENTS.md

@@ -53,6 +53,7 @@ Key documentation for this project:
 ### Testing
 
 - ✅ **DO run unit tests via the CLI**, filtered to the tests relevant to your changes (e.g. `dotnet test backend/FwLite/FwLiteShared.Tests --filter "FullyQualifiedName~MyTestClass"`). Verify tests you wrote or changed actually pass before handing work back. Never run whole suites just to "see if anything broke".
+- ✅ **`MiniLcm.Tests`, `LcmCrdt.Tests`, `FwDataMiniLcmBridge.Tests` need no infrastructure** — these are the unit/conformance tests for the two `IMiniLcmApi` implementations; run them filtered to your change like any unit test (this is where a `*TestsBase` change should be verified, including its FwData side). The FwData/LCM-backed selections load a real project (a handful of tests = seconds, a whole class = a minute or two), so keep the filter tight; don't reach for the whole class or project.
 - ✅ **FwLite integration tests** (e.g. `FwLiteProjectSync.Tests`) need no infrastructure but are slow. Run a **targeted selection** (specific tests, not necessarily whole classes) when you touched critical sync code **and believe the work is finished** — not on every iteration. Waiting on tests burns time; be deliberate about which runs buy real signal.
 - ✅ **`backend/Testing` contains unit tests too** — only tests marked `Category=Integration|FlakyIntegration|RequiresDb` (and the `Testing.Browser` namespace) need infrastructure. Its unit tests are fine to run: `task test:unit -- <filter>` excludes those categories for you.
 - ✅ **FwLite viewer Playwright tests MAY be run** — they're cheap: `task playwright-test-standalone -- <test-name-filter>` (from `frontend/viewer/`) auto-starts the vite dev server with the in-browser demo project; no lexbox stack, chromium only. Always filter to relevant tests; details in `frontend/viewer/AGENTS.md`.

backend/FwLite/AGENTS.md

@@ -253,6 +253,12 @@ if (entity?.DeletedAt is not null) return;
 
 ---
 
+## Validation
+
+Imperative validation in `MiniLcmApiValidationWrapper` (rules that need an async lookup, so they can't be FluentValidation rules) must throw `FluentValidation.ValidationException` — the same type the validators throw — not `InvalidOperationException`. Otherwise the same rule reports different exception types per code path. Reserve `InvalidOperationException` for genuine "can't happen" data-integrity guards, not user-input validation.
+
+---
+
 ## Important Files Quick Reference
 
 | File | Purpose | Risk Level |

backend/FwLite/FwDataMiniLcmBridge/Api/FwDataMiniLcmApi.cs

@@ -364,12 +364,25 @@ private Publication FromLcmPossibility(ICmPossibility lcmPossibility)
         var possibility = new Publication
         {
             Id = lcmPossibility.Guid,
-            Name = FromLcmMultiString(lcmPossibility.Name)
+            Name = FromLcmMultiString(lcmPossibility.Name),
+            IsMain = lcmPossibility.IsProtected
         };
 
         return possibility;
     }
 
+    // The Main Dictionary is the publication FieldWorks protects from deletion (IsProtected). LibLCM treats anything
+    // other than exactly one protected publication as an error; we mirror that for >1 but tolerate 0 so a project
+    // without a Main Dictionary can still create entries (it just won't auto-add one).
+    private ICmPossibility? FindMainPublication()
+    {
+        var protectedPublications = Publications.PossibilitiesOS.Where(pub => pub.IsProtected).ToArray();
+        if (protectedPublications.Length > 1)
+            throw new InvalidOperationException(
+                $"Expected at most one protected (main) publication but found {protectedPublications.Length}.");
+        return protectedPublications.FirstOrDefault();
+    }
+
     public Task<Publication> UpdatePublication(Guid id, UpdateObjectInput<Publication> update)
     {
         var lcmPublication = GetLcmPublication(id);
@@ -992,8 +1005,16 @@ public Task<int> GetEntryIndex(Guid entryId, string? query = null, IndexQueryOpt
 
     public async Task<Entry> CreateEntry(Entry entry, CreateEntryOptions? options = null)
     {
-        options ??= CreateEntryOptions.Everything;
+        options ??= CreateEntryOptions.WithMainPublication;
         entry.Id = entry.Id == default ? Guid.NewGuid() : entry.Id;
+        if (options.AutoAddMainPublication)
+        {
+            var mainPublication = FindMainPublication();
+            if (mainPublication is not null && entry.PublishIn.All(pub => pub.Id != mainPublication.Guid))
+            {
+                entry.PublishIn.Add(FromLcmPossibility(mainPublication));
+            }
+        }
         try
         {
             UndoableUnitOfWorkHelper.DoUsingNewOrCurrentUOW("Create Entry",

backend/FwLite/FwDataMiniLcmBridge/Api/UpdateProxy/UpdatePublicationProxy.cs

@@ -17,6 +17,12 @@ public UpdatePublicationProxy(ICmPossibility lcmPublication, FwDataMiniLcmApi le
         _lexboxLcmApi = lexboxLcmApi;
     }
 
+    public override bool IsMain
+    {
+        get => _lcmPublication.IsProtected;
+        set { } // IsMain is mapped from IsProtected in FW; don't write back
+    }
+
     public override MultiString Name
     {
         get => new UpdateMultiStringProxy(_lcmPublication.Name, _lexboxLcmApi);

backend/FwLite/FwLiteProjectSync.Tests/EntrySyncTests.cs

@@ -311,18 +311,20 @@ public async Task CanSyncRandomEntries(ApiType? roundTripApiType)
         // We expect the final result to be equivalent to this "raw"/untouched, requested state.
         var expected = after.Copy();
 
+        // Don't auto-add the main publication when staging data: the sync path under test (EntrySync) never does.
+        var asIsEntryOptions = CreateEntryOptions.AsIs;
         if (roundTripApi is not null)
         {
             // round-tripping ensures we're dealing with realistic data
             // (e.g. in fwdata ComplexFormComponents do not have an Id)
-            before = await roundTripApi.CreateEntry(before);
+            before = await roundTripApi.CreateEntry(before, asIsEntryOptions);
             await roundTripApi.DeleteEntry(before.Id);
-            after = await roundTripApi.CreateEntry(after);
+            after = await roundTripApi.CreateEntry(after, asIsEntryOptions);
             await roundTripApi.DeleteEntry(after.Id);
         }
 
         // before should not be round-tripped here. That's handled above.
-        await Api.CreateEntry(before);
+        await Api.CreateEntry(before, asIsEntryOptions);
 
         // act
         await EntrySync.SyncFull(before, after, Api);