@happyvertical/smrt-education 0.57.51 → 0.57.52

package/dist/manifest.json

@@ -2,7 +2,7 @@
   "version": "1.0.0",
   "timestamp": 0,
   "packageName": "@happyvertical/smrt-education",
-  "packageVersion": "0.57.51",
+  "packageVersion": "0.57.52",
   "objects": {
     "@happyvertical/smrt-education:CatalogProviderCollection": {
       "name": "catalogprovidercollection",
@@ -15557,6 +15557,15 @@
           "type": "text",
           "required": false
         },
+        "renderedSha256": {
+          "type": "text",
+          "required": false
+        },
+        "renderedSizeBytes": {
+          "type": "integer",
+          "required": true,
+          "default": 0
+        },
         "recordedAt": {
           "type": "text",
           "required": false
@@ -15636,11 +15645,16 @@
           "field": "sizeBytes",
           "rule": "required",
           "fieldType": "integer"
+        },
+        {
+          "field": "renderedSizeBytes",
+          "rule": "required",
+          "fieldType": "integer"
         }
       ],
       "schema": {
         "tableName": "course_bundle_members",
-        "ddl": "CREATE TABLE IF NOT EXISTS \"course_bundle_members\" (\n  \"id\" UUID PRIMARY KEY NOT NULL,\n  \"slug\" TEXT NOT NULL,\n  \"context\" TEXT NOT NULL DEFAULT '',\n  \"created_at\" TIMESTAMP NOT NULL DEFAULT current_timestamp,\n  \"updated_at\" TIMESTAMP NOT NULL DEFAULT current_timestamp,\n  \"tenant_id\" UUID NOT NULL,\n  \"import_id\" UUID NOT NULL,\n  \"ordinal\" INTEGER NOT NULL DEFAULT 0,\n  \"path\" TEXT,\n  \"sha256\" TEXT,\n  \"size_bytes\" INTEGER NOT NULL DEFAULT 0,\n  \"media_type\" TEXT,\n  \"kind\" TEXT,\n  \"status\" TEXT,\n  \"reason\" TEXT,\n  \"duplicate_of_path\" TEXT,\n  \"role\" TEXT,\n  \"section_key\" TEXT,\n  \"lesson_key\" TEXT,\n  \"recorded_at\" TEXT\n);",
+        "ddl": "CREATE TABLE IF NOT EXISTS \"course_bundle_members\" (\n  \"id\" UUID PRIMARY KEY NOT NULL,\n  \"slug\" TEXT NOT NULL,\n  \"context\" TEXT NOT NULL DEFAULT '',\n  \"created_at\" TIMESTAMP NOT NULL DEFAULT current_timestamp,\n  \"updated_at\" TIMESTAMP NOT NULL DEFAULT current_timestamp,\n  \"tenant_id\" UUID NOT NULL,\n  \"import_id\" UUID NOT NULL,\n  \"ordinal\" INTEGER NOT NULL DEFAULT 0,\n  \"path\" TEXT,\n  \"sha256\" TEXT,\n  \"size_bytes\" INTEGER NOT NULL DEFAULT 0,\n  \"media_type\" TEXT,\n  \"kind\" TEXT,\n  \"status\" TEXT,\n  \"reason\" TEXT,\n  \"duplicate_of_path\" TEXT,\n  \"role\" TEXT,\n  \"section_key\" TEXT,\n  \"lesson_key\" TEXT,\n  \"rendered_sha256\" TEXT,\n  \"rendered_size_bytes\" INTEGER NOT NULL DEFAULT 0,\n  \"recorded_at\" TEXT\n);",
         "columns": {
           "id": {
             "type": "UUID",
@@ -15747,6 +15761,17 @@
             "notNull": false,
             "unique": false
           },
+          "rendered_sha256": {
+            "type": "TEXT",
+            "notNull": false,
+            "unique": false
+          },
+          "rendered_size_bytes": {
+            "type": "INTEGER",
+            "notNull": true,
+            "unique": false,
+            "default": 0
+          },
           "recorded_at": {
             "type": "TEXT",
             "notNull": false,
@@ -15801,7 +15826,7 @@
             ]
           }
         ],
-        "version": "18729a67"
+        "version": "c9f518ec"
       },
       "importPath": "@happyvertical/smrt-education"
     },
@@ -15832,7 +15857,7 @@
       ],
       "schema": {
         "tableName": "course_bundle_members",
-        "ddl": "CREATE TABLE IF NOT EXISTS \"course_bundle_members\" (\n  \"id\" UUID PRIMARY KEY NOT NULL,\n  \"slug\" TEXT NOT NULL,\n  \"context\" TEXT NOT NULL DEFAULT '',\n  \"created_at\" TIMESTAMP NOT NULL DEFAULT current_timestamp,\n  \"updated_at\" TIMESTAMP NOT NULL DEFAULT current_timestamp,\n  \"tenant_id\" UUID NOT NULL,\n  \"import_id\" UUID NOT NULL,\n  \"ordinal\" INTEGER NOT NULL DEFAULT 0,\n  \"path\" TEXT,\n  \"sha256\" TEXT,\n  \"size_bytes\" INTEGER NOT NULL DEFAULT 0,\n  \"media_type\" TEXT,\n  \"kind\" TEXT,\n  \"status\" TEXT,\n  \"reason\" TEXT,\n  \"duplicate_of_path\" TEXT,\n  \"role\" TEXT,\n  \"section_key\" TEXT,\n  \"lesson_key\" TEXT,\n  \"recorded_at\" TEXT\n);",
+        "ddl": "CREATE TABLE IF NOT EXISTS \"course_bundle_members\" (\n  \"id\" UUID PRIMARY KEY NOT NULL,\n  \"slug\" TEXT NOT NULL,\n  \"context\" TEXT NOT NULL DEFAULT '',\n  \"created_at\" TIMESTAMP NOT NULL DEFAULT current_timestamp,\n  \"updated_at\" TIMESTAMP NOT NULL DEFAULT current_timestamp,\n  \"tenant_id\" UUID NOT NULL,\n  \"import_id\" UUID NOT NULL,\n  \"ordinal\" INTEGER NOT NULL DEFAULT 0,\n  \"path\" TEXT,\n  \"sha256\" TEXT,\n  \"size_bytes\" INTEGER NOT NULL DEFAULT 0,\n  \"media_type\" TEXT,\n  \"kind\" TEXT,\n  \"status\" TEXT,\n  \"reason\" TEXT,\n  \"duplicate_of_path\" TEXT,\n  \"role\" TEXT,\n  \"section_key\" TEXT,\n  \"lesson_key\" TEXT,\n  \"rendered_sha256\" TEXT,\n  \"rendered_size_bytes\" INTEGER NOT NULL DEFAULT 0,\n  \"recorded_at\" TEXT\n);",
         "columns": {
           "id": {
             "type": "UUID",
@@ -15939,6 +15964,17 @@
             "notNull": false,
             "unique": false
           },
+          "rendered_sha256": {
+            "type": "TEXT",
+            "notNull": false,
+            "unique": false
+          },
+          "rendered_size_bytes": {
+            "type": "INTEGER",
+            "notNull": true,
+            "unique": false,
+            "default": 0
+          },
           "recorded_at": {
             "type": "TEXT",
             "notNull": false,
@@ -15993,7 +16029,7 @@
             ]
           }
         ],
-        "version": "18729a67"
+        "version": "c9f518ec"
       },
       "importPath": "@happyvertical/smrt-education"
     },

package/dist/smrt-knowledge.json

@@ -3,13 +3,13 @@
   "sensitiveFieldsExcluded": true,
   "generatedAt": "1970-01-01T00:00:00.000Z",
   "packageName": "@happyvertical/smrt-education",
-  "packageVersion": "0.57.51",
+  "packageVersion": "0.57.52",
   "sourceManifestPath": "dist/manifest.json",
   "agentDocPath": "AGENTS.md",
   "sourceHashes": {
-    "manifest": "4f05280aff9a1f96069240ea8cc397534ddf633bd3b5d60f97df6d21515149ff",
-    "packageJson": "225ea71c55e19fa103e784a5bb94760de71e6d467db44767b0c7321f4e1d69a2",
-    "agents": "f95be1f04c4a93378d3397c972a204210b45efa72b3c7fb7d924232eedb0a265",
+    "manifest": "17ca2c9e8a00e15b0501d8f13e528e455809f7c3c9d55bb6e1c53ac1478bec21",
+    "packageJson": "25a063c5585e031763a80004925c7a0a6b90d711e1f3f653d0c4f91fdbf090ae",
+    "agents": "a7197fcc3a842cd27908c11d77ef00019be7aedb95ae7c6b0b4cf5958d9e5b0b",
     "moduleDoc:agents/learner-service.md": "d05e2cc8f824505806540aa0b2e2158f9d7ccb218fd36671d21718d58a44908d",
     "moduleDoc:agents/submission-service.md": "b5c1cb3863e06530c0b575ba5598c8a329c0e907ea42226267faa7317680eaa1",
     "moduleDoc:agents/workbook-service.md": "ae3a0f208b42766e417eac226e4a24ec15fe17c1ba9cde26a325ded982614278",
@@ -18,7 +18,7 @@
     "moduleDoc:agents/course-generation-review.md": "7f6d5e67ed56ec9923bd00df05650b363fb0db30c1b1f004c8b203f9c62cd117",
     "moduleDoc:agents/question-records.md": "335fe247f7294c6b891885d43d4eaccb20885bfb4a861cf66ae245e8be1fa781",
     "moduleDoc:agents/question-extraction.md": "75972dda9b2f837471fb42f963ef143a098005c919a1a49db7ce395a5e5e50a7",
-    "moduleDoc:agents/course-bundle-import.md": "dd463bd79882c47446c71c9f3a1f5c97fa9db00b50a28f4598ee709caa8a3bb4",
+    "moduleDoc:agents/course-bundle-import.md": "4ae9ec59d1ad91cbd7019f73cf38dd788cdd28375c8964c3b08b91feddfc5086",
     "moduleDoc:agents/source-import.md": "30ee823589d655442bcf1abcd86a2132172065b4bc5e3c4d89b1e8b7afcc66bc",
     "moduleDoc:agents/family-intake.md": "540b6d37f438496acb0208b6ac7258264e2142bce3d2f986967b7741606fe4f7",
     "moduleDoc:agents/catalog-previews.md": "5f4e0352274009b3ce00408017a16666036d24df695a7078c27317ba4e66f0a7",
@@ -5215,6 +5215,19 @@
           "required": false,
           "columnType": "TEXT"
         },
+        {
+          "name": "renderedSha256",
+          "type": "text",
+          "required": false,
+          "columnType": "TEXT"
+        },
+        {
+          "name": "renderedSizeBytes",
+          "type": "integer",
+          "required": true,
+          "columnType": "INTEGER",
+          "default": 0
+        },
         {
           "name": "recordedAt",
           "type": "text",
@@ -15085,7 +15098,7 @@
     "polymorphicAssociations": 0,
     "uuidColumns": 774
   },
-  "agentDoc": "# smrt-education\n\nCatalog objects are optional-tenant scoped and have no generated REST, CLI, MCP,\nor AI surface. Applications use `CatalogService`, supplying their actor-bound\nauthorization decision and the trusted `TenantCollection.withTenantCatalogReadScope`\nadapter from smrt-users. Catalog authorization receives the exact shared collection\nexecutor; live session/permission checks must use that executor. That read scope may widen only an allowlisted catalog\ncollection's `list` to self, ancestors, and globals; it never authorizes a get,\nwrite, raw query, sibling, or descendant read.\n\n`personal-use-copy` courses and all `answer-key` lesson assets are private\nfamily material. Do not publish them, promote them, or expose them through a\nlearner-facing result. `LessonAsset` is the sole owner join for lesson files;\ndo not add asset-id fields to lesson subclasses.\n\nLearner objects require exact tenant ownership and close generated surfaces.\n`LearnerService` owns actor/resource authorization and transactional planning and\nstart-progress; see [learner contract](agents/learner-service.md). Guardian\nrelationships require a host decision about their type and effective terms.\nNever expose direct collection writes as learner handlers, reset progress on a\nretry, or treat an unscored nullable decimal as zero.\n\nSubmission capture, guardian grading and private PDF packet splitting use\n`SubmissionService`; see [submission contract](agents/submission-service.md).\nKeep asset metadata on the same transaction and compensate only newly planned\nfiles after rollback; never expose source URIs or accept learner grade patches.\n\nValidation: `pnpm --filter @happyvertical/smrt-education test`,\n`test:postgres`, `typecheck`, and `build`.\n\nPhysical workbook capture, immutable assessment/feedback history and descriptive\nevidence use [WorkbookService](agents/workbook-service.md). Keep provisional\nsharing separate from guardian Grade authority. Photo-first receipt/context\ncapture uses [WorkbookPhotoService](agents/photo-intake.md); covers/references\nnever create learning evidence, and blank exercises need explicit work intent.\n\nRevisioned authored courses, release, archive, enrollment pins and licence use\n[CourseAuthoringService](agents/course-authoring-service.md). Keep history\nappend-only, require request keys, and never weaken personal-use-copy rules.\n\nGeneration proposals into new draft revisions and revision-bound educational\nreview use [CourseGenerationReviewService](agents/course-generation-review.md).\nKeep supplied content unless a decision is recorded; reviews carry no score.\n\nImmutable questions, answer keys/rubrics, stable learning objectives, worksheet\nrender ids and exact-revision references use [QuestionAuthoringService](agents/question-records.md).\nKeep keys out of learner bundles; a missing key stays null. `attachGeneratedPrintable`\nstores the rendered PDFs; the guardian copy is always an `answer-key`, never `printable`.\n\nStructured questions from scans and imported course files use\n[QuestionExtractionService](agents/question-extraction.md) and the `extract-questions`\ningestion handler. Anchors come from provider region catalogs only; match by worksheet\nid and layout, never text; ambiguous matches stay unattached; source-keyed replays.\n\nCourse bundles (zip with a closed manifest, or a file set) use\n[CourseBundleImportService](agents/course-bundle-import.md). Parse only with the\nbounded reader; refuse, never trim, hostile input; unsupported members stay\nvisible; proposals only, nothing auto-published.\n\nSource import: [host boundary and transaction contract](agents/source-import.md).\n`SourceImportService` requires host opt-in, matching active actor/tenant and\nexplicit licensing authorization; never add a system/global write bypass.\n\nFamily-owned unassigned originals and audited learner adoption use\n[FamilyIntakeService](agents/family-intake.md). Keep family source scope immutable;\nassignment alone never creates learning history or grants learner inbox access.\n\nPublisher catalog previews use [the preview contract](agents/catalog-previews.md).\nKeep independent image rights and source revisions; serve via `CatalogPreviewService`\nwith live actor/tenant authorization, never direct provider image URLs.\n\nEnrollment Plan projections and guardian-only parent notes use\n[the Plan contract](agents/learner-plan.md). Keep awaiting review separate from\nContinue; note authority/provenance never grants learner access or learning evidence.\n\nHomeschool group sharing uses [GroupSharingService](agents/group-sharing.md),\nits read-only reader and revision comment service. Group tenants never become\nfamily ancestors or guardians. Enumerate only the group-owned opaque roster index,\nthen revalidate each exact family edge and current policy. Media delivery must call\nthe reader on every request; signed storage URLs are not a sharing grant. Saved\nparent comment visibility applies on every child read. Progress with feedback\ndisabled contains counts only, never correctness, scores or concept labels.\n\nLearner-safe tutor context and append-only tutor transcripts use\n[TutorContextService and TutorSessionService](agents/tutor-context.md). Learner\naudience never carries keys, rubrics, notes or storage URIs; no AI call here.\n",
+  "agentDoc": "# smrt-education\n\nCatalog objects are optional-tenant scoped and have no generated REST, CLI, MCP,\nor AI surface. Applications use `CatalogService`, supplying their actor-bound\nauthorization decision and the trusted `TenantCollection.withTenantCatalogReadScope`\nadapter from smrt-users. Catalog authorization receives the exact shared collection\nexecutor; live session/permission checks must use that executor. That read scope may widen only an allowlisted catalog\ncollection's `list` to self, ancestors, and globals; it never authorizes a get,\nwrite, raw query, sibling, or descendant read.\n\n`personal-use-copy` courses and all `answer-key` lesson assets are private\nfamily material. Do not publish them, promote them, or expose them through a\nlearner-facing result. `LessonAsset` is the sole owner join for lesson files;\ndo not add asset-id fields to lesson subclasses.\n\nLearner objects require exact tenant ownership and close generated surfaces.\n`LearnerService` owns actor/resource authorization and transactional planning and\nstart-progress; see [learner contract](agents/learner-service.md). Guardian\nrelationships require a host decision about their type and effective terms.\nNever expose direct collection writes as learner handlers, reset progress on a\nretry, or treat an unscored nullable decimal as zero.\n\nSubmission capture, guardian grading and private PDF packet splitting use\n`SubmissionService`; see [submission contract](agents/submission-service.md).\nKeep asset metadata on the same transaction and compensate only newly planned\nfiles after rollback; never expose source URIs or accept learner grade patches.\n\nValidation: `pnpm --filter @happyvertical/smrt-education test`,\n`test:postgres`, `typecheck`, and `build`.\n\nPhysical workbook capture, immutable assessment/feedback history and descriptive\nevidence use [WorkbookService](agents/workbook-service.md). Keep provisional\nsharing separate from guardian Grade authority. Photo-first receipt/context\ncapture uses [WorkbookPhotoService](agents/photo-intake.md); covers/references\nnever create learning evidence, and blank exercises need explicit work intent.\n\nRevisioned authored courses, release, archive, enrollment pins and licence use\n[CourseAuthoringService](agents/course-authoring-service.md). Keep history\nappend-only, require request keys, and never weaken personal-use-copy rules.\n\nGeneration proposals into new draft revisions and revision-bound educational\nreview use [CourseGenerationReviewService](agents/course-generation-review.md).\nKeep supplied content unless a decision is recorded; reviews carry no score.\n\nImmutable questions, answer keys/rubrics, stable learning objectives, worksheet\nrender ids and exact-revision references use [QuestionAuthoringService](agents/question-records.md).\nKeep keys out of learner bundles; a missing key stays null. `attachGeneratedPrintable`\nstores the rendered PDFs; the guardian copy is always an `answer-key`, never `printable`.\n\nStructured questions from scans and imported course files use\n[QuestionExtractionService](agents/question-extraction.md) and the `extract-questions`\ningestion handler. Anchors come from provider region catalogs only; match by worksheet\nid and layout, never text; ambiguous matches stay unattached; source-keyed replays.\n\nCourse bundles (zip with a closed manifest, or a file set; Word/PowerPoint members render through an injected `DocumentConverter`, fail closed) use\n[CourseBundleImportService](agents/course-bundle-import.md). Parse only with the\nbounded reader; refuse, never trim, hostile input; unsupported members stay\nvisible; proposals only, nothing auto-published.\n\nSource import: [host boundary and transaction contract](agents/source-import.md).\n`SourceImportService` requires host opt-in, matching active actor/tenant and\nexplicit licensing authorization; never add a system/global write bypass.\n\nFamily-owned unassigned originals and audited learner adoption use\n[FamilyIntakeService](agents/family-intake.md). Keep family source scope immutable;\nassignment alone never creates learning history or grants learner inbox access.\n\nPublisher catalog previews use [the preview contract](agents/catalog-previews.md).\nKeep independent image rights and source revisions; serve via `CatalogPreviewService`\nwith live actor/tenant authorization, never direct provider image URLs.\n\nEnrollment Plan projections and guardian-only parent notes use\n[the Plan contract](agents/learner-plan.md). Keep awaiting review separate from\nContinue; note authority/provenance never grants learner access or learning evidence.\n\nHomeschool group sharing uses [GroupSharingService](agents/group-sharing.md),\nits read-only reader and revision comment service. Group tenants never become\nfamily ancestors or guardians. Enumerate only the group-owned opaque roster index,\nthen revalidate each exact family edge and current policy. Media delivery must call\nthe reader on every request; signed storage URLs are not a sharing grant. Saved\nparent comment visibility applies on every child read. Progress with feedback\ndisabled contains counts only, never correctness, scores or concept labels.\n\nLearner-safe tutor context and append-only tutor transcripts use\n[TutorContextService and TutorSessionService](agents/tutor-context.md). Learner\naudience never carries keys, rubrics, notes or storage URIs; no AI call here.\n",
   "moduleDocs": [
     {
       "path": "agents/learner-service.md",
@@ -15130,7 +15143,7 @@
     {
       "path": "agents/course-bundle-import.md",
       "module": "course-bundle-import",
-      "content": "# Course bundle import\n\n`CourseBundleImportService` (SMRT #4578, extends `CourseAuthoringService`,\nconsumed by Hearth #94) imports a course bundle, a zip with an optional manifest\nor a plain file set, as bounded evidence, hands its supported members to the\nhost for ingestion extraction and proposes sections and lessons for review.\nRisk is high: untrusted archive parsing and new schema. It reuses the course\nauthorizer: `importBundle` needs `education.course.author`, `readImport`\n`education.course.read-draft`, `proposeStructure` `education.course.generate`\n(via `CourseGenerationReviewService`). Operations: `import-course-bundle`,\n`read-course-bundle`. Server only (Node `zlib`/`crypto`); no new dependency.\n\n## Flow\n\n1. `importBundle({ ...actor, bundle, limits?, onProgress?, receiveMember? })`\n   authorizes **before parsing**, unpacks (stage `unpack`), classifies and lays\n   out the structure (`classify`), writes the evidence in one transaction and\n   then calls `receiveMember` for each accepted member (`extract`), outside any\n   transaction. `bundle` is `{ archive: Uint8Array }` or `{ files: [{ path,\n   bytes }] }`. Hostile or damaged input throws `CourseBundleRefusedError`\n   (`reason`, optional `entry`) and writes nothing.\n2. `receiveMember(member, { replayed })` is the host's ingestion seam: call\n   `IngestionService.receive` with `sourceKind: 'course_import'` and the member\n   bytes, then queue extraction and `preserveQuestionExtractionAnalysis` as in\n   [question extraction](question-extraction.md). Members carry `role`\n   (manifest page role), `lessonKey`, `sectionKey`. It runs on a replay too\n   (ingestion dedups by content hash), so an interrupted first run resumes. Its\n   return values are in `result.receipts`.\n3. `proposeStructure({ ...actor, importId, bundle, courseId,\n   expectedHeadRevisionId })` (stage `propose`) re-supplies the bundle (bytes\n   are never retained), checks it is the imported one, and runs a deterministic\n   `CourseMaterialGenerator` (`smrt-bundle-import`) through\n   `requestGeneration`/`runGeneration`. The result is an ordinary proposal: a\n   new draft revision of additions and fills; conflicts wait for\n   `decideProposal`. Nothing is released, assigned or published. The request key\n   is `bundle:<courseId>:<bundleIdentity>`, so a retry returns the same job; the\n   same key with another head is `CourseRequestKeyConflictError`.\n4. `readImport({ ...actor, importId })` returns the recorded evidence.\n\nProgress is an observer callback (`onProgress({ stage, completed, total })`,\nstages `unpack`, `classify`, `extract`, `propose`; each starts at `completed: 0`).\nIngestion has no public job-progress seam, so none is invented. A throwing\nlistener never affects the import.\n\n## Hard bounds (`COURSE_BUNDLE_LIMITS`; a host may only lower them)\n\n| limit | value |\n|---|---|\n| `maxArchiveBytes` (input) | 100 MiB |\n| `maxEntries` | 200 |\n| `maxEntryBytes` (inflated) | 50 MiB |\n| `maxTotalBytes` (inflated) | 200 MiB |\n| `maxCompressionRatio` (members over `ratioFloorBytes` = 64 KiB) | 100 |\n| `maxPathBytes` / `maxSegmentBytes` / `maxPathDepth` | 512 / 255 / 8 |\n| `maxManifestBytes` | 256 KiB |\n| `maxLessons` (the proposal schema's cap) | 50 |\n| `maxLessonBodyChars` | 100 000 |\n| archive nesting | 1: a zip inside a zip is never expanded |\n\nA lesson whose combined text members exceed `maxLessonBodyChars` refuses the\nbundle (`lesson_too_large`, `entry` = lesson key) at import; text is never\ntrimmed or dropped. A plain file set enforces `maxTotalBytes` on its member\nbytes exactly as an archive does on inflated bytes.\n\nThe reader (`course-bundle-zip.ts`) parses the end record and central directory,\nvalidates every offset and size against the real buffer, inflates with\n`maxOutputLength` equal to the declared size (declared sizes are never trusted:\na mismatch either way is `size_mismatch`), checks CRC32 and re-checks the total\non actual output. Stored and deflate only. Refusals (`CourseBundleRefusalReason`)\ncover: not a zip, truncated, ambiguous end record (two end records that both fit\nthe file, for example one forged in a comment), zip64, multi-disk, central\ndirectory not ending at the end record, entries out of range or overlapping,\nlocal header name/method mismatch, encrypted entries (flags 0/6/13, AES method),\nsymlink and special files (unix mode; DOS reparse flag), unsupported method,\nbad CRC, ratio, entry count, totals, duplicate names (case/NFC-insensitive,\nbackslashes normalised first) and bad paths (`..`, `.`, empty segments, absolute,\ndrive letter or `:` anywhere, NUL/control/bidi characters, lone surrogates,\ntoo long or deep). An **honestly** oversize member (declared over\n`maxEntryBytes`) is not inflated and is recorded `refused / entry_too_large`.\n\n## Members and statuses\n\nBy content (never by extension alone): `pdf` (`%PDF-` in the first 1 KiB),\n`image` (PNG, JPEG, WebP), `text` (`.txt`/`.md`, strict UTF-8, no control\ncharacters, body-sized), `manifest` (root `manifest.json`), else `unsupported`.\nStatus: `accepted` (handed to extraction), `duplicate` (same sha256 as an earlier\nmember in canonical order; recorded once as evidence, `duplicateOfPath`, never\nextracted again), `unsupported` (`unsupported_type`, `nested_archive`,\n`unreadable_text`), `refused` (`encrypted_pdf`, `entry_too_large`), `metadata`\n(the manifest). Nothing is dropped. An encrypted PDF is any PDF with an\n`/Encrypt` token anywhere in its bytes (trailer, xref stream): a false positive\non literal text inside an uncompressed stream is refused, never read.\n\n## Manifest (`manifest.json`, version 1, closed)\n\n```json\n{ \"version\": 1, \"course\": { \"subject\": \"Math\" },\n  \"sections\": [ { \"key\": \"unit-1\", \"title\": \"Unit One\",\n    \"lessons\": [ { \"key\": \"sums\", \"title\": \"Sums\",\n      \"members\": [ { \"path\": \"sheets/sums.pdf\", \"role\": \"exercise\" } ] } ] } ] }\n```\nEvery object is closed: an unknown key anywhere is `manifest_unknown_key` (with\nits path in `entry`); another version is `manifest_version`. Keys match\n`[A-Za-z0-9._:-]{1,120}` and are unique; roles are the extraction page roles\n(`exercise`, `example`, `instructions`, `cover`, `answer_key`, `mixed`,\n`unknown`); a listed path must exist and be listed once. The manifest is a source\nfact: roles are `source`, anything the extractor infers stays `inferred`, and a\nmanifest/extraction disagreement belongs in extraction review, not here. Without\na manifest each accepted member is one lesson, sections come from the top-level\ndirectory and order is natural path order (digit runs compare numerically, no\nlocale). Manifest order: the manifest, declared members, then the rest by path.\nLesson bodies are the readable text members of a lesson, joined; PDFs and images\nare never inlined. Members whose declared role is `answer_key` or `mixed` never\ncontribute text to a lesson body (they remain recorded evidence members); the\nproposal rationale only counts how many were withheld.\n\n## Readers\n\n`BundleReader { formatId, detect, prepare }` in a `BundleReaderRegistry`;\nbuilt-ins are `manifest-v1` and `plain-files`. SCORM and Common Cartridge are\n**not implemented** (their manifests describe launch resources and activity\ntrees that do not map to sections/lessons without design); a host can register\na reader ahead of the built-ins (`register(reader, { first: true })`).\n\n## Identity and replay\n\n`bundleIdentity` (`cbi1-...`) digests the sorted (path, sha256) pairs, so\nre-zipping, reordering and timestamps replay. One `CourseBundleImport` per\n(tenant, identity); a concurrent first import replays once on the unique index.\n\n## Models (closed, required tenant, no REST/CLI/MCP/AI, append-only)\n\n`CourseBundleImport` (`course_bundle_imports`: identity, format, reader,\nmanifest version, counts, actor) and `CourseBundleMember` (`course_bundle_members`:\nunique tenant/import/path; ordinal, sha256, size, media type, kind, status,\nreason, duplicate-of, role, section and lesson key). Hashes and placements only:\nbytes stay with the host's ingestion receipts. Rewrites and deletes throw.\nNormal manifest migrations create both tables.\n\n## Not in this release\n\n- Ingestion's receipt contract is unchanged: there is no `application/zip` part\n  type or `parentEvidenceId` in ingestion. The education side unpacks and the\n  host receives members one by one, so archive-to-member lineage lives in\n  `course_bundle_members` (and the host's receipts), not in ingestion evidence.\n  Changing ingestion's published receipt is a separate design decision.\n- A manifest/extraction disagreement is not compared automatically.\n- SCORM / Common Cartridge readers; zip64; nested archives; encrypted zips.\n"
+      "content": "# Course bundle import\n\n`CourseBundleImportService` (SMRT #4578, extends `CourseAuthoringService`,\nconsumed by Hearth #94) imports a course bundle, a zip with an optional manifest\nor a plain file set, as bounded evidence, hands its supported members to the\nhost for ingestion extraction and proposes sections and lessons for review.\nRisk is high: untrusted archive parsing and new schema. It reuses the course\nauthorizer: `importBundle` needs `education.course.author`, `readImport`\n`education.course.read-draft`, `proposeStructure` `education.course.generate`\n(via `CourseGenerationReviewService`). Operations: `import-course-bundle`,\n`read-course-bundle`. Server only (Node `zlib`/`crypto`); no new dependency.\n\n## Flow\n\n1. `importBundle({ ...actor, bundle, limits?, onProgress?, receiveMember? })`\n   authorizes **before parsing**, unpacks (stage `unpack`), classifies and lays\n   out the structure (`classify`), writes the evidence in one transaction and\n   then calls `receiveMember` for each accepted member (`extract`), outside any\n   transaction. `bundle` is `{ archive: Uint8Array }` or `{ files: [{ path,\n   bytes }] }`. Hostile or damaged input throws `CourseBundleRefusedError`\n   (`reason`, optional `entry`) and writes nothing.\n2. `receiveMember(member, { replayed })` is the host's ingestion seam: call\n   `IngestionService.receive` with `sourceKind: 'course_import'` and the member\n   bytes, then queue extraction and `preserveQuestionExtractionAnalysis` as in\n   [question extraction](question-extraction.md). Members carry `role`\n   (manifest page role), `lessonKey`, `sectionKey`. It runs on a replay too\n   (ingestion dedups by content hash), so an interrupted first run resumes. Its\n   return values are in `result.receipts`.\n3. `proposeStructure({ ...actor, importId, bundle, courseId,\n   expectedHeadRevisionId })` (stage `propose`) re-supplies the bundle (bytes\n   are never retained), checks it is the imported one, and runs a deterministic\n   `CourseMaterialGenerator` (`smrt-bundle-import`) through\n   `requestGeneration`/`runGeneration`. The result is an ordinary proposal: a\n   new draft revision of additions and fills; conflicts wait for\n   `decideProposal`. Nothing is released, assigned or published. The request key\n   is `bundle:<courseId>:<bundleIdentity>`, so a retry returns the same job; the\n   same key with another head is `CourseRequestKeyConflictError`.\n4. `readImport({ ...actor, importId })` returns the recorded evidence.\n\nProgress is an observer callback (`onProgress({ stage, completed, total })`,\nstages `unpack`, `classify`, `render`, `extract`, `propose`; each starts at `completed: 0`;\n`render` is emitted only when the bundle has Word/PowerPoint members).\nIngestion has no public job-progress seam, so none is invented. A throwing\nlistener never affects the import.\n\n## Hard bounds (`COURSE_BUNDLE_LIMITS`; a host may only lower them)\n\n| limit | value |\n|---|---|\n| `maxArchiveBytes` (input) | 100 MiB |\n| `maxEntries` | 200 |\n| `maxEntryBytes` (inflated) | 50 MiB |\n| `maxTotalBytes` (inflated) | 200 MiB |\n| `maxCompressionRatio` (members over `ratioFloorBytes` = 64 KiB) | 100 |\n| `maxPathBytes` / `maxSegmentBytes` / `maxPathDepth` | 512 / 255 / 8 |\n| `maxManifestBytes` | 256 KiB |\n| `maxLessons` (the proposal schema's cap) | 50 |\n| `maxLessonBodyChars` | 100 000 |\n| `maxOfficeMembers` (Word/PowerPoint members rendered per bundle) | 50 |\n| `maxRenderedBytes` (PDF accepted back from the converter) | 50 MiB |\n| `conversionTimeoutMs` (one Office member) | 60 000 |\n| archive nesting | 1: a zip inside a zip is never expanded |\n\nA lesson whose combined text members exceed `maxLessonBodyChars` refuses the\nbundle (`lesson_too_large`, `entry` = lesson key) at import; text is never\ntrimmed or dropped. A plain file set enforces `maxTotalBytes` on its member\nbytes exactly as an archive does on inflated bytes.\n\nThe reader (`course-bundle-zip.ts`) parses the end record and central directory,\nvalidates every offset and size against the real buffer, inflates with\n`maxOutputLength` equal to the declared size (declared sizes are never trusted:\na mismatch either way is `size_mismatch`), checks CRC32 and re-checks the total\non actual output. Stored and deflate only. Refusals (`CourseBundleRefusalReason`)\ncover: not a zip, truncated, ambiguous end record (two end records that both fit\nthe file, for example one forged in a comment), zip64, multi-disk, central\ndirectory not ending at the end record, entries out of range or overlapping,\nlocal header name/method mismatch, encrypted entries (flags 0/6/13, AES method),\nsymlink and special files (unix mode; DOS reparse flag), unsupported method,\nbad CRC, ratio, entry count, totals, duplicate names (case/NFC-insensitive,\nbackslashes normalised first) and bad paths (`..`, `.`, empty segments, absolute,\ndrive letter or `:` anywhere, NUL/control/bidi characters, lone surrogates,\ntoo long or deep). An **honestly** oversize member (declared over\n`maxEntryBytes`) is not inflated and is recorded `refused / entry_too_large`.\n\n## Members and statuses\n\nBy content (never by extension alone): `docx`/`pptx` (see below), `pdf` (`%PDF-` in the first 1 KiB),\n`image` (PNG, JPEG, WebP), `text` (`.txt`/`.md`, strict UTF-8, no control\ncharacters, body-sized), `manifest` (root `manifest.json`), else `unsupported`.\nStatus: `accepted` (handed to extraction), `duplicate` (same sha256 as an earlier\nmember in canonical order; recorded once as evidence, `duplicateOfPath`, never\nextracted again), `unsupported` (`unsupported_type`, `nested_archive`,\n`unreadable_text`), `refused` (`encrypted_pdf`, `entry_too_large`), `metadata`\n(the manifest). Nothing is dropped. An encrypted PDF is any PDF with an\n`/Encrypt` token anywhere in its bytes (trailer, xref stream): a false positive\non literal text inside an uncompressed stream is refused, never read.\n\n## Word and PowerPoint members (SMRT #4615)\n\nA `.docx`/`.pptx` member is an OOXML package only when the name and the content\nagree: a zip whose entries include `[Content_Types].xml` and `word/document.xml`\n(docx) or `ppt/presentation.xml` (pptx). Before the bytes go anywhere, the\npackage itself passes the same bounded zip reader (ratio, overlap, paths, CRC,\n100 MiB inflated); a failure is `refused / unsafe_office`, never a refusal of the\nwhole bundle. Other zips stay `unsupported / nested_archive`.\n\nTwo separate things happen to an accepted Office member, both after\nauthorization and before the evidence is written:\n\n- **Pages**: a `DocumentConverter` (`document-converter.ts`) renders the original\n  to PDF. The importer enforces the byte and time limits itself (input\n  `maxEntryBytes`, output `maxRenderedBytes`, `conversionTimeoutMs` with an abort\n  signal and a hard race, so an adapter that ignores the signal still times out),\n  and rejects output that is not a PDF or is encrypted. The converter is told a\n  neutral file name (`document.docx`), never the bundle path. Identical documents\n  are rendered once.\n- **Text**: an injectable `OfficeTextReader` (best effort; a failure or a `null`\n  costs the member nothing) returns whole-document text and, for decks, text and\n  speaker notes per slide. Office text is handed to the host on the accepted\n  member (`officeText`) and is **never** inlined into a proposed lesson body:\n  teacher guides and notes carry answers, and audience belongs to manifest v2.\n\n`receiveMember` gets the original `bytes` (store them as the private original\nasset: \"Open in PowerPoint\" for decks) and `rendered` (`{ bytes, sha256, mediaType:\n'application/pdf', converterId }`, store as a private asset beside the original:\nthe pages for \"Present\"/reading). The member row records `renderedSha256` and\n`renderedSizeBytes`; bytes are still not retained here.\n\n**Fail closed.** Any converter problem records that member `unsupported` with a\nreason and writes the rest of the bundle normally: `converter_unavailable` (none\nconfigured), `converter_failed` (unreachable or 5xx), `converter_timeout`,\n`converter_too_large`, `converter_rejected` (HTTP 4xx, not a PDF, encrypted),\n`office_limit` (beyond `maxOfficeMembers`, in natural path order). A member that\nfailed is not a lesson in the proposal, and identical copies of a failed document\nfail the same way rather than becoming duplicates of nothing. Outcomes are part\nof the append-only evidence, so a replay does not retry a recorded failure (to\nretry, change the bundle). A replay does re-render recorded-accepted members so the\nhost receives the bytes again; if it can no longer render one it throws\n`DocumentConversionError` rather than silently dropping an accepted member.\n\n### Host configuration\n\n```ts\nconst service = new CourseBundleImportService(db, authorizer, {\n  converter: new GotenbergDocumentConverter({\n    baseUrl: process.env.GOTENBERG_URL, // for example http://gotenberg:3000\n    timeoutMs: 60_000,\n    maxBytes: 50 * 1024 * 1024,\n  }),\n  officeTextReader, // optional, see below\n});\n```\n\n`GotenbergDocumentConverter` posts to `<baseUrl>/forms/libreoffice/convert`, talks\nto nothing else, refuses redirects and URLs with credentials, and caps the response\nas it streams. Run Gotenberg as an internal-only service with **no outbound\nnetwork**, its own size/time limits and no shared volume with the app; this\npackage assumes untrusted files never run next to the application. Tests use\n`FakeDocumentConverter` (records calls; `failWith`, `failWhen`, `delayMs`, `output`).\n\n`OfficeTextReader` is deliberately not wired to the SDK here, so this package\ndoes not depend on an unreleased `@happyvertical/documents`. Once the Office\nprocessors ship (sdk#1485, `extractOffice(bytes, 'docx' | 'pptx')`), implement\n`read` with it; the snippet is in the `OfficeTextReader` doc comment.\n\n## Manifest (`manifest.json`, version 1, closed)\n\n```json\n{ \"version\": 1, \"course\": { \"subject\": \"Math\" },\n  \"sections\": [ { \"key\": \"unit-1\", \"title\": \"Unit One\",\n    \"lessons\": [ { \"key\": \"sums\", \"title\": \"Sums\",\n      \"members\": [ { \"path\": \"sheets/sums.pdf\", \"role\": \"exercise\" } ] } ] } ] }\n```\nEvery object is closed: an unknown key anywhere is `manifest_unknown_key` (with\nits path in `entry`); another version is `manifest_version`. Keys match\n`[A-Za-z0-9._:-]{1,120}` and are unique; roles are the extraction page roles\n(`exercise`, `example`, `instructions`, `cover`, `answer_key`, `mixed`,\n`unknown`); a listed path must exist and be listed once. The manifest is a source\nfact: roles are `source`, anything the extractor infers stays `inferred`, and a\nmanifest/extraction disagreement belongs in extraction review, not here. Without\na manifest each accepted member is one lesson, sections come from the top-level\ndirectory and order is natural path order (digit runs compare numerically, no\nlocale). Manifest order: the manifest, declared members, then the rest by path.\nLesson bodies are the readable text members of a lesson, joined; PDFs and images\nare never inlined. Members whose declared role is `answer_key` or `mixed` never\ncontribute text to a lesson body (they remain recorded evidence members); the\nproposal rationale only counts how many were withheld.\n\n## Readers\n\n`BundleReader { formatId, detect, prepare }` in a `BundleReaderRegistry`;\nbuilt-ins are `manifest-v1` and `plain-files`. SCORM and Common Cartridge are\n**not implemented** (their manifests describe launch resources and activity\ntrees that do not map to sections/lessons without design); a host can register\na reader ahead of the built-ins (`register(reader, { first: true })`).\n\n## Identity and replay\n\n`bundleIdentity` (`cbi1-...`) digests the sorted (path, sha256) pairs, so\nre-zipping, reordering and timestamps replay. One `CourseBundleImport` per\n(tenant, identity); a concurrent first import replays once on the unique index.\n\n## Models (closed, required tenant, no REST/CLI/MCP/AI, append-only)\n\n`CourseBundleImport` (`course_bundle_imports`: identity, format, reader,\nmanifest version, counts, actor) and `CourseBundleMember` (`course_bundle_members`:\nunique tenant/import/path; ordinal, sha256, size, media type, kind, status,\nreason, duplicate-of, role, section and lesson key). Hashes and placements only:\nbytes stay with the host's ingestion receipts. Rewrites and deletes throw.\nNormal manifest migrations create both tables.\n\n## Not in this release\n\n- Manifest version 2 (material roles with audience, lesson pages, course-level\n  members, external video references) waits for the phase 1 role vocabulary\n  (smrt#4614). The page-role vocabulary and the closed v1 schema are unchanged;\n  see the hand-off note on smrt#4615.\n- Re-rendering a member whose first import recorded a converter failure (outcomes\n  are append-only per bundle identity).\n- Ingestion's receipt contract is unchanged: there is no `application/zip` part\n  type or `parentEvidenceId` in ingestion. The education side unpacks and the\n  host receives members one by one, so archive-to-member lineage lives in\n  `course_bundle_members` (and the host's receipts), not in ingestion evidence.\n  Changing ingestion's published receipt is a separate design decision.\n- A manifest/extraction disagreement is not compared automatically.\n- SCORM / Common Cartridge readers; zip64; nested archives; encrypted zips.\n"
     },
     {
       "path": "agents/source-import.md",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@happyvertical/smrt-education",
-  "version": "0.57.51",
+  "version": "0.57.52",
   "smrtJsdoc": "strict",
   "description": "Tenant-aware learning catalog models and authorized catalog access for SMRT",
   "type": "module",
@@ -25,18 +25,18 @@
     }
   },
   "dependencies": {
-    "@happyvertical/smrt-assets": "0.57.51",
-    "@happyvertical/smrt-core": "0.57.51",
-    "@happyvertical/smrt-profiles": "0.57.51",
-    "@happyvertical/smrt-tenancy": "0.57.51",
-    "@happyvertical/smrt-users": "0.57.51",
+    "@happyvertical/smrt-assets": "0.57.52",
+    "@happyvertical/smrt-core": "0.57.52",
+    "@happyvertical/smrt-profiles": "0.57.52",
+    "@happyvertical/smrt-tenancy": "0.57.52",
+    "@happyvertical/smrt-users": "0.57.52",
     "@happyvertical/files": "^0.111.3",
     "@happyvertical/sql": "^0.111.3",
     "pdf-lib": "1.17.1",
-    "@happyvertical/smrt-ingestion": "0.57.51"
+    "@happyvertical/smrt-ingestion": "0.57.52"
   },
   "devDependencies": {
-    "@happyvertical/smrt-vitest": "0.57.51",
+    "@happyvertical/smrt-vitest": "0.57.52",
     "@types/node": "24.13.6",
     "typescript": "6.0.3",
     "vite": "8.1.6",