fix(data): publish migration and cursor corrections (successor to #185)

CHANGELOG.md

@@ -165,6 +165,8 @@ The project publishes 0.x prerelease versions; a stable release line is not yet
 
 ### Fixed
 
+- Enforce text-chunk uniqueness, index nullable memory references, cascade memory relations with workspace deletion, batch generation listings, and paginate relation listings with opaque cursors. Refs #178.
+
 - Follow the shared design language for reading, numeric and action alignment; generate the existing Web color variables from a pinned design-system token snapshot, and use a single consistent empty-state pattern. Refs #211.
 
 - Improve web caption and status contrast in both themes, including tinted danger

docs/MIGRATION_SEQUENCE.md

@@ -6,30 +6,31 @@ These draft changes are cumulative, not independently deployable:
 | --- | --- | --- | --- |
 | 1 | #194 / #183 | 0024 file lexical lane | released/main schema 23 (merged) |
 | 2 | #173 HNSW completion (supersedes #197 HOLD) | 0025 HNSW indexes + text continuation | #194, schema 24 |
-| 3 | #195 / #185 | 0026 data-plane hygiene | schema 25 |
+| 3 | #195 / #185 | 0027 data-plane hygiene | schema 26 |
 
-The PR base chain is `main` → `codex/fix-pr-183` → `codex/fix-pr-180`
-→ `codex/fix-pr-185`. Successor branches must include their predecessor schema and source. Local
-repair branches are rebuilt on current main and replay the original authored
-changes; published commit identities remain in the original PR history.
-Keep this order when retargeting after a predecessor merges.
+Migrations 0024–0026 are now on `main`. PR #195 must integrate that exact
+history before adding data-plane hygiene as 0027. Local repair branches retain
+the original authored commits and add a current-main integration commit rather
+than rewriting published history.
 
-On 2026-09-10, main `2986fe38175f54d99f15dd38a498708c6ecd88cd` and published
-tags `v0.1.0` / `v0.1.1` contain only migrations 0001–0023. This does not prove
-that a private deployment never applied a draft. Consequently migration
-numbers and SQL identities are retained, not renumbered on an assumption.
+The hygiene migration was originally reviewed as draft 0026, but main now owns
+0026 for `memory_producer_agent`. The unmerged hygiene migration therefore
+moves to 0027. An operator who privately applied the old draft under version 26
+must stop and obtain a recovery plan; do not rewrite Goose history to make the
+new main sequence appear valid.
 
 Migration 0025 creates the three cosine HNSW indexes. The shipping text route
 no longer uses `DISTINCT ON (f.id) ORDER BY f.id` as its primary plan: it walks
 cosine-ordered candidates and falls back to that exact query only when a bounded
 scan underfills. Visual cosine-order already matched HNSW. Face DDL is not a
 face-query speedup. #195 still requires predecessor schema 25. Recall and live
-latency remain `#175`, not this migration.
+latency remain `#175`, not this migration. #195 now requires predecessor schema
+26.
 
 This document does not waive a review gate or authorize deployment.
 
 Goose startup remains strict: no `WithAllowMissing` or equivalent option is
-enabled. A database that already applied 26 while omitting 24/25 will correctly
+enabled. A database that already applied 27 while omitting 24/25/26 will correctly
 fail startup against the cumulative schema. Stop and obtain an operator-owned
 recovery plan for such a database; do not edit its migration history, renumber
 its SQL, or apply lower versions out of order to manufacture a pass.
@@ -55,9 +56,10 @@ an operator-owned recovery; do not mark an unverified partial schema applied.
 `scripts/verify.sh integration` creates a separate, owned `_test` database and
 runs `TestMigrationUpgradeSequence`. It applies real Goose migrations to 23,
 seeds a file with duplicate text chunks, then advances one version at a time
-to the branch's declared head (24, 25, or 26). Each step checks full applied
+to the branch's declared head (24 through 27). Each step checks full applied
 history and preserved data; subsequent steps check lexical backfill, valid
-HNSW DDL, and deduplication/unique rejection. Finally the ordinary production
+HNSW DDL, producer-agent indexing, and hygiene deduplication/unique rejection.
+Finally the ordinary production
 `DB.Migrate` startup path must accept the resulting history unchanged.
 
 The dedicated test uses `MEM_MIGRATION_SEQUENCE_TEST_DB`, refuses a database

scripts/verify.sh

@@ -4,7 +4,7 @@ set -euo pipefail
 
 REPO_ROOT="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")/.." && pwd)"
 MODE="${1:-unit}"
-EXPECTED_MIGRATION_HEAD=26
+EXPECTED_MIGRATION_HEAD=27
 MIGRATION_ROLLBACK_TARGET=11
 MODEL_TEXT_CANONICAL_BASE=15
 WORKSPACE_AI_PROFILE_BASE=16

server/internal/api/api.go

@@ -78,7 +78,7 @@ type MemoryService interface {
 	Restore(context.Context, memory.LifecycleCommand) (*memory.MutationResult, error)
 	Forget(context.Context, memory.ForgetCommand) (*memory.ForgetResult, error)
 	CreateRelation(context.Context, memory.CreateRelationCommand) (*memory.CreateRelationResult, error)
-	ListRelations(context.Context, memory.ListRelationsQuery) ([]memory.Relation, error)
+	ListRelations(context.Context, memory.ListRelationsQuery) (*memory.ListRelationsResult, error)
 }
 
 // DurableContextService is the scoped durable-context port (mem#70). Handlers

server/internal/api/handlers_memory.go

@@ -763,18 +763,21 @@ func (s *Server) handleListMemoryRelations(w http.ResponseWriter, r *http.Reques
 	}
 
 	tok := r.Context().Value(ctxToken).(*auth.Token)
-	relations, err := s.Memory.ListRelations(r.Context(), memory.ListRelationsQuery{
+	result, err := s.Memory.ListRelations(r.Context(), memory.ListRelationsQuery{
 		WorkspaceID:  currentWorkspace(r).ID,
 		MemoryID:     id,
 		Direction:    direction,
 		RelationType: relationType,
 		AllowedPaths: tok.Paths,
 		Limit:        limit,
+		Cursor:       r.URL.Query().Get("cursor"),
 	})
 	if err != nil {
 		switch {
 		case errors.Is(err, memory.ErrInvalidCommand):
 			writeError(w, http.StatusBadRequest, "invalid_relation_query", err.Error())
+		case errors.Is(err, memory.ErrInvalidCursor):
+			writeError(w, http.StatusBadRequest, "invalid_cursor", err.Error())
 		case errors.Is(err, memory.ErrNotFound):
 			writeError(w, http.StatusNotFound, "not_found", "memory not found")
 		case errors.Is(err, memory.ErrForgotten):
@@ -789,8 +792,12 @@ func (s *Server) handleListMemoryRelations(w http.ResponseWriter, r *http.Reques
 		}
 		return
 	}
-	if relations == nil {
-		relations = []memory.Relation{}
+	resp := map[string]any{"relations": result.Relations}
+	if result.Relations == nil {
+		resp["relations"] = []memory.Relation{}
 	}
-	writeJSON(w, http.StatusOK, map[string]any{"relations": relations})
+	if result.NextCursor != "" {
+		resp["next_cursor"] = result.NextCursor
+	}
+	writeJSON(w, http.StatusOK, resp)
 }

server/internal/api/handlers_memory_test.go

@@ -107,10 +107,10 @@ func (s *memoryServiceStub) CreateRelation(
 func (s *memoryServiceStub) ListRelations(
 	_ context.Context,
 	q memory.ListRelationsQuery,
-) ([]memory.Relation, error) {
+) (*memory.ListRelationsResult, error) {
 	s.calls++
 	s.listRelationsQuery = q
-	return s.relations, s.controlErr
+	return &memory.ListRelationsResult{Relations: s.relations}, s.controlErr
 }
 
 func memoryHandlerContext(req *http.Request, paths []string) (*http.Request, uuid.UUID, uuid.UUID, uuid.UUID) {

server/internal/db/migration_sequence_test.go

@@ -3,6 +3,7 @@ package db
 import (
 	"context"
 	"database/sql"
+	"errors"
 	"os"
 	"strconv"
 	"strings"
@@ -11,6 +12,7 @@ import (
 
 	"github.com/google/uuid"
 	"github.com/jackc/pgx/v5"
+	"github.com/jackc/pgx/v5/pgconn"
 	"github.com/pressly/goose/v3"
 )
 
@@ -107,8 +109,12 @@ func TestMigrationUpgradeSequence(t *testing.T) {
 			}
 		}
 		var chunks int
-		if err := sqldb.QueryRowContext(ctx, "SELECT count(*) FROM embeddings_text WHERE file_id=$1", fileID).Scan(&chunks); err != nil || chunks != 2 {
-			t.Fatalf("preserved chunks=%d, want=2, err=%v", chunks, err)
+		wantChunks := 2
+		if version >= 27 {
+			wantChunks = 1
+		}
+		if err := sqldb.QueryRowContext(ctx, "SELECT count(*) FROM embeddings_text WHERE file_id=$1", fileID).Scan(&chunks); err != nil || chunks != wantChunks {
+			t.Fatalf("preserved chunks=%d, want=%d, err=%v", chunks, wantChunks, err)
 		}
 		if version >= 26 {
 			var producerIdx int
@@ -118,6 +124,13 @@ func TestMigrationUpgradeSequence(t *testing.T) {
 		}
 		t.Logf("PASS: strict Goose upgrade to %d; complete history and populated data preserved", version)
 	}
+	if head >= 27 {
+		_, err := sqldb.ExecContext(ctx, "INSERT INTO embeddings_text(file_id,chunk_index,chunk_text) VALUES($1,0,'duplicate')", fileID)
+		var pgErr *pgconn.PgError
+		if !errors.As(err, &pgErr) || pgErr.Code != "23505" {
+			t.Fatalf("expected duplicate rejection 23505, got %v", err)
+		}
+	}
 	// The real startup path must accept the upgraded history unchanged.
 	if err := (&DB{url: dsn}).Migrate(ctx); err != nil {
 		t.Fatalf("production startup after sequential upgrade: %v", err)

server/internal/db/migrations/0027_data_plane_hygiene.sql

@@ -0,0 +1,85 @@
+-- +goose Up
+-- Data-plane hygiene from the index audit (#178), sequenced after main's 0026.
+-- Three independent fixes bundled into one migration because each is a single
+-- DDL statement and none warrants its own schema version.
+
+-- Item 1: embeddings_text uniqueness on (file_id, chunk_index).
+-- The write path (indexer.go) DELETEs all chunks for a file before re-inserting,
+-- so duplicates should never exist in practice. Deduplicate defensively before
+-- adding the constraint: if any duplicates survived, keep the row with the
+-- smallest UUID for deterministic selection (UUID order is not insert order).
+-- +goose StatementBegin
+DELETE FROM embeddings_text
+ WHERE id NOT IN (
+    SELECT DISTINCT ON (file_id, chunk_index) id
+      FROM embeddings_text
+     ORDER BY file_id, chunk_index, id
+ );
+-- +goose StatementEnd
+
+-- +goose StatementBegin
+ALTER TABLE embeddings_text
+    ADD CONSTRAINT uq_embeddings_text_file_chunk UNIQUE (file_id, chunk_index);
+-- +goose StatementEnd
+
+-- Item 2: partial indexes for ON DELETE SET NULL lookups on memories.
+-- Without these, every file or user delete takes a RowExclusiveLock on memories
+-- and performs a sequential scan to find the rows to null out.
+-- +goose StatementBegin
+CREATE INDEX IF NOT EXISTS idx_memories_source_file_id
+    ON memories (source_file_id) WHERE source_file_id IS NOT NULL;
+-- +goose StatementEnd
+
+-- +goose StatementBegin
+CREATE INDEX IF NOT EXISTS idx_memories_created_by_user_id
+    ON memories (created_by_user_id) WHERE created_by_user_id IS NOT NULL;
+-- +goose StatementEnd
+
+-- Item 3: memory_relations FKs must cascade with memories.
+-- memories.workspace_id is ON DELETE CASCADE, so a workspace delete removes
+-- memories rows. Without matching cascade on memory_relations, the delete then
+-- fails on any edge touching those memories. Align the referential actions.
+-- +goose StatementBegin
+ALTER TABLE memory_relations
+    DROP CONSTRAINT IF EXISTS memory_relations_workspace_id_fkey,
+    DROP CONSTRAINT IF EXISTS memory_relations_source_id_fkey,
+    DROP CONSTRAINT IF EXISTS memory_relations_target_id_fkey;
+-- +goose StatementEnd
+
+-- +goose StatementBegin
+ALTER TABLE memory_relations
+    ADD CONSTRAINT memory_relations_workspace_id_fkey
+        FOREIGN KEY (workspace_id) REFERENCES workspaces(id) ON DELETE CASCADE,
+    ADD CONSTRAINT memory_relations_source_id_fkey
+        FOREIGN KEY (source_id) REFERENCES memories(id) ON DELETE CASCADE,
+    ADD CONSTRAINT memory_relations_target_id_fkey
+        FOREIGN KEY (target_id) REFERENCES memories(id) ON DELETE CASCADE;
+-- +goose StatementEnd
+
+-- +goose Down
+-- +goose StatementBegin
+ALTER TABLE memory_relations
+    DROP CONSTRAINT IF EXISTS memory_relations_workspace_id_fkey,
+    DROP CONSTRAINT IF EXISTS memory_relations_source_id_fkey,
+    DROP CONSTRAINT IF EXISTS memory_relations_target_id_fkey;
+-- +goose StatementEnd
+
+-- +goose StatementBegin
+ALTER TABLE memory_relations
+    ADD CONSTRAINT memory_relations_workspace_id_fkey
+        FOREIGN KEY (workspace_id) REFERENCES workspaces(id),
+    ADD CONSTRAINT memory_relations_source_id_fkey
+        FOREIGN KEY (source_id) REFERENCES memories(id),
+    ADD CONSTRAINT memory_relations_target_id_fkey
+        FOREIGN KEY (target_id) REFERENCES memories(id);
+-- +goose StatementEnd
+
+-- +goose StatementBegin
+DROP INDEX IF EXISTS idx_memories_created_by_user_id;
+DROP INDEX IF EXISTS idx_memories_source_file_id;
+-- +goose StatementEnd
+
+-- +goose StatementBegin
+ALTER TABLE embeddings_text
+    DROP CONSTRAINT IF EXISTS uq_embeddings_text_file_chunk;
+-- +goose StatementEnd

server/internal/db/migrations_test.go

@@ -0,0 +1,109 @@
+package db
+
+import (
+	"context"
+	"errors"
+	"os"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/jackc/pgx/v5/pgconn"
+	"github.com/jackc/pgx/v5/pgxpool"
+)
+
+func TestEmbeddingsTextUniqueConstraint(t *testing.T) {
+	dsn := os.Getenv("MEM_TEST_DB")
+	if dsn == "" {
+		t.Skip("MEM_TEST_DB not set; skipping DB integration test")
+	}
+	config, err := pgxpool.ParseConfig(dsn)
+	if err != nil {
+		t.Fatalf("parse MEM_TEST_DB: %v", err)
+	}
+	if !strings.HasSuffix(config.ConnConfig.Database, "_test") {
+		t.Fatalf("refusing to modify non-test database %q", config.ConnConfig.Database)
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
+	defer cancel()
+
+	pool, err := pgxpool.NewWithConfig(ctx, config)
+	if err != nil {
+		t.Fatalf("connect: %v", err)
+	}
+	defer pool.Close()
+
+	db := &DB{Pool: pool, url: dsn}
+	if err := db.Migrate(ctx); err != nil {
+		t.Fatalf("migrate: %v", err)
+	}
+	tx, err := pool.Begin(ctx)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer tx.Rollback(context.Background())
+
+	var userID, workspaceID, fileID uuid.UUID
+	if err := tx.QueryRow(ctx, `
+		INSERT INTO users (email, password_hash) VALUES ($1, 'x')
+		RETURNING id
+	`, uuid.NewString()+"@example.test").Scan(&userID); err != nil {
+		t.Fatalf("seed user: %v", err)
+	}
+	if err := tx.QueryRow(ctx, `
+		INSERT INTO workspaces (name, resource_owner_user_id)
+		VALUES ('unique-ctest-ws', $1)
+		ON CONFLICT (resource_owner_user_id) DO UPDATE SET name = EXCLUDED.name
+		RETURNING id
+	`, userID).Scan(&workspaceID); err != nil {
+		t.Fatalf("seed workspace: %v", err)
+	}
+	if err := tx.QueryRow(ctx, `
+		INSERT INTO files (user_id, name, path, size, sha256, mime, storage_key)
+		VALUES ($1, 'unique-ctest.txt', '/unique-ctest.txt', 0, '', 'text/plain', 'test://unique')
+		RETURNING id
+	`, userID).Scan(&fileID); err != nil {
+		t.Fatalf("seed file: %v", err)
+	}
+	if _, err := tx.Exec(ctx, `
+		INSERT INTO embeddings_text (file_id, chunk_index, chunk_text, provider)
+		VALUES ($1, 0, 'chunk zero', 'test')
+	`, fileID); err != nil {
+		t.Fatalf("first insert: %v", err)
+	}
+	// Replay the actual migration over populated, pre-constraint data inside
+	// this rollback-only transaction. A fresh-schema migrate alone misses this.
+	if _, err := tx.Exec(ctx, `ALTER TABLE embeddings_text DROP CONSTRAINT uq_embeddings_text_file_chunk`); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := tx.Exec(ctx, `INSERT INTO embeddings_text (file_id, chunk_index, chunk_text)
+		VALUES ($1, 0, 'legacy duplicate')`, fileID); err != nil {
+		t.Fatal(err)
+	}
+	migration, err := migrationsFS.ReadFile("migrations/0027_data_plane_hygiene.sql")
+	if err != nil {
+		t.Fatal(err)
+	}
+	up := strings.SplitN(string(migration), "-- +goose Down", 2)[0]
+	if _, err := tx.Exec(ctx, up); err != nil {
+		t.Fatalf("migrate populated table: %v", err)
+	}
+	var survivors int
+	if err := tx.QueryRow(ctx, `SELECT count(*) FROM embeddings_text WHERE file_id=$1`, fileID).Scan(&survivors); err != nil || survivors != 1 {
+		t.Fatalf("deduplicated rows = %d, err=%v", survivors, err)
+	}
+
+	_, err = tx.Exec(ctx, `
+		INSERT INTO embeddings_text (file_id, chunk_index, chunk_text, provider)
+		VALUES ($1, 0, 'duplicate chunk', 'test')
+	`, fileID)
+	if err == nil {
+		t.Fatal("expected duplicate (file_id, chunk_index) to be rejected")
+	}
+	var pgErr *pgconn.PgError
+	if !errors.As(err, &pgErr) || pgErr.Code != "23505" {
+		t.Fatalf("expected unique violation 23505, got %v", err)
+	}
+}