feat(kebab-store-sqlite): p9-fb-23 task 3 — V006 migration + put/get_document round-trip version stamps

Add V006__incremental_ingest.sql to persist last_chunker_version and
last_embedding_version on the documents table. Wire both columns into
upsert_document (INSERT + ON CONFLICT UPDATE) and get_document (SELECT +
row mapper), replacing the previous hardcoded None. Add two round-trip
tests in tests/incremental_ingest.rs covering the set and None cases.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

crates/kebab-store-sqlite/src/documents.rs

@@ -165,7 +165,8 @@ impl kebab_core::DocumentStore for SqliteStore {
                     doc_id, asset_id, workspace_path, title, lang,
                     source_type, trust_level, parser_version,
                     doc_version, schema_version, metadata_json,
-                    provenance_json, created_at, updated_at
+                    provenance_json, created_at, updated_at,
+                    last_chunker_version, last_embedding_version
                 FROM documents WHERE doc_id = ?",
                 params![id.0],
                 document_row_from_sql,
@@ -221,8 +222,8 @@ impl kebab_core::DocumentStore for SqliteStore {
             // under that invariant.
             schema_version: row.schema_version as u32,
             doc_version: row.doc_version as u32,
-            last_chunker_version: None,
-            last_embedding_version: None,
+            last_chunker_version: row.last_chunker_version.map(kebab_core::ChunkerVersion),
+            last_embedding_version: row.last_embedding_version.map(kebab_core::EmbeddingVersion),
         }))
     }
 
@@ -367,6 +368,8 @@ struct DocumentRow {
     provenance_json: String,
     // source_type / trust_level are loaded back via metadata_json round-trip,
     // so we do not need separate fields here for `get_document`.
+    last_chunker_version: Option<String>,
+    last_embedding_version: Option<String>,
 }
 
 fn document_row_from_sql(row: &rusqlite::Row<'_>) -> rusqlite::Result<DocumentRow> {
@@ -385,6 +388,10 @@ fn document_row_from_sql(row: &rusqlite::Row<'_>) -> rusqlite::Result<DocumentRo
         schema_version: row.get(9)?,
         metadata_json: row.get(10)?,
         provenance_json: row.get(11)?,
+        // 12: created_at, 13: updated_at — not stored in DocumentRow
+        // (only needed for list_documents). Columns 14-15 follow.
+        last_chunker_version: row.get(14)?,
+        last_embedding_version: row.get(15)?,
     })
 }
 
@@ -505,24 +512,27 @@ fn upsert_document(
             doc_id, asset_id, workspace_path, title, lang,
             source_type, trust_level, parser_version,
             doc_version, schema_version, metadata_json,
-            provenance_json, created_at, updated_at
-        ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+            provenance_json, created_at, updated_at,
+            last_chunker_version, last_embedding_version
+        ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
         ON CONFLICT(doc_id) DO UPDATE SET
-            asset_id        = excluded.asset_id,
-            workspace_path  = excluded.workspace_path,
-            title           = excluded.title,
-            lang            = excluded.lang,
-            source_type     = excluded.source_type,
-            trust_level     = excluded.trust_level,
-            parser_version  = excluded.parser_version,
+            asset_id              = excluded.asset_id,
+            workspace_path        = excluded.workspace_path,
+            title                 = excluded.title,
+            lang                  = excluded.lang,
+            source_type           = excluded.source_type,
+            trust_level           = excluded.trust_level,
+            parser_version        = excluded.parser_version,
             -- doc_version: bump on update. excluded.doc_version is the
             -- caller's submitted value; we ignore it and add 1 to the
             -- existing column so each re-ingest cleanly increments.
-            doc_version     = documents.doc_version + 1,
-            schema_version  = excluded.schema_version,
-            metadata_json   = excluded.metadata_json,
-            provenance_json = excluded.provenance_json,
-            updated_at      = excluded.updated_at",
+            doc_version           = documents.doc_version + 1,
+            schema_version        = excluded.schema_version,
+            metadata_json         = excluded.metadata_json,
+            provenance_json       = excluded.provenance_json,
+            updated_at            = excluded.updated_at,
+            last_chunker_version  = excluded.last_chunker_version,
+            last_embedding_version = excluded.last_embedding_version",
         params![
             doc.doc_id.0,
             doc.source_asset_id.0,
@@ -538,6 +548,8 @@ fn upsert_document(
             provenance_json,
             created_at,
             now,
+            doc.last_chunker_version.as_ref().map(|v| v.0.as_str()),
+            doc.last_embedding_version.as_ref().map(|v| v.0.as_str()),
         ],
     )
     .map_err(StoreError::from)?;

crates/kebab-store-sqlite/tests/incremental_ingest.rs

@@ -0,0 +1,119 @@
+//! Round-trip tests for `last_chunker_version` / `last_embedding_version`
+//! columns added by the V006 migration (p9-fb-23 task 3).
+
+use std::path::PathBuf;
+
+use kebab_core::{
+    AssetId, AssetStorage, Block, CanonicalDocument, Checksum, ChunkerVersion, CommonBlock,
+    DocumentId, DocumentStore, EmbeddingVersion, HeadingBlock, Lang, MediaType, Metadata,
+    ParserVersion, Provenance, RawAsset, SourceSpan, SourceType, SourceUri, TrustLevel,
+    WorkspacePath,
+};
+use kebab_store_sqlite::SqliteStore;
+use time::OffsetDateTime;
+
+mod common;
+
+fn make_asset() -> RawAsset {
+    let bytes = b"incremental-ingest-test";
+    RawAsset {
+        asset_id: AssetId("f".repeat(32)),
+        source_uri: SourceUri::File(PathBuf::from("/tmp/inc.md")),
+        workspace_path: WorkspacePath::new("notes/inc.md".into()).unwrap(),
+        media_type: MediaType::Markdown,
+        byte_len: bytes.len() as u64,
+        checksum: Checksum(blake3::hash(bytes).to_hex().to_string()),
+        discovered_at: OffsetDateTime::from_unix_timestamp(1_700_000_000).unwrap(),
+        stored: AssetStorage::Reference {
+            path: PathBuf::from("/tmp/inc.md"),
+            sha: Checksum(blake3::hash(bytes).to_hex().to_string()),
+        },
+    }
+}
+
+fn make_doc() -> CanonicalDocument {
+    let doc_id = DocumentId("d".repeat(32));
+    let block = Block::Heading(HeadingBlock {
+        common: CommonBlock {
+            block_id: kebab_core::BlockId("b".repeat(32)),
+            heading_path: vec![],
+            source_span: SourceSpan::Line { start: 1, end: 1 },
+        },
+        level: 1,
+        text: "Incremental Title".into(),
+    });
+    let metadata = Metadata {
+        aliases: vec![],
+        tags: vec![],
+        created_at: OffsetDateTime::from_unix_timestamp(1_700_000_000).unwrap(),
+        updated_at: OffsetDateTime::from_unix_timestamp(1_700_000_000).unwrap(),
+        source_type: SourceType::Markdown,
+        trust_level: TrustLevel::Primary,
+        user_id_alias: None,
+        user: Default::default(),
+    };
+    CanonicalDocument {
+        doc_id,
+        source_asset_id: AssetId("f".repeat(32)),
+        workspace_path: WorkspacePath::new("notes/inc.md".into()).unwrap(),
+        title: "Incremental Title".into(),
+        lang: Lang("en".into()),
+        blocks: vec![block],
+        metadata,
+        provenance: Provenance { events: vec![] },
+        parser_version: ParserVersion("test-parser".into()),
+        schema_version: 1,
+        doc_version: 1,
+        last_chunker_version: None,
+        last_embedding_version: None,
+    }
+}
+
+#[test]
+fn put_then_get_document_roundtrips_version_stamps() {
+    let env = common::TestEnv::new();
+    let store = SqliteStore::open(&env.config()).unwrap();
+    store.run_migrations().unwrap();
+
+    let asset = make_asset();
+    store.put_asset(&asset).unwrap();
+
+    let mut doc = make_doc();
+    doc.last_chunker_version = Some(ChunkerVersion("md-heading-v1".into()));
+    doc.last_embedding_version = Some(EmbeddingVersion("multilingual-e5-small@v1".into()));
+
+    store.put_document(&doc).unwrap();
+    let loaded = store
+        .get_document(&doc.doc_id)
+        .unwrap()
+        .expect("doc round-trips");
+
+    assert_eq!(loaded.last_chunker_version, doc.last_chunker_version);
+    assert_eq!(loaded.last_embedding_version, doc.last_embedding_version);
+}
+
+#[test]
+fn put_then_get_document_roundtrips_none_stamps() {
+    let env = common::TestEnv::new();
+    let store = SqliteStore::open(&env.config()).unwrap();
+    store.run_migrations().unwrap();
+
+    let asset = make_asset();
+    store.put_asset(&asset).unwrap();
+
+    let doc = make_doc(); // both version stamps are None by default
+    store.put_document(&doc).unwrap();
+    let loaded = store
+        .get_document(&doc.doc_id)
+        .unwrap()
+        .expect("doc round-trips");
+
+    assert!(
+        loaded.last_chunker_version.is_none(),
+        "last_chunker_version must be None when not set"
+    );
+    assert!(
+        loaded.last_embedding_version.is_none(),
+        "last_embedding_version must be None when not set"
+    );
+}

migrations/V006__incremental_ingest.sql

@@ -0,0 +1,6 @@
+-- p9-fb-23: incremental ingest needs to know which chunker / embedding
+-- versions were used to populate this document so a re-ingest can
+-- decide whether to skip (versions match) or re-process (any mismatch).
+-- parser_version is already on documents from V001.
+ALTER TABLE documents ADD COLUMN last_chunker_version TEXT;
+ALTER TABLE documents ADD COLUMN last_embedding_version TEXT;