memorix 1.2.0 → 1.2.2

package/src/store/persistence.ts

@@ -1,291 +1,291 @@
-/**
- * Persistence Layer — Runtime + Migration
- *
- * Runtime responsibilities:
- *   - Data directory resolution (flat global ~/.memorix/data/)
- *   - Orama DB file path helpers
- *   - Legacy per-project subdirectory migration
- *
- * JSON I/O helpers have been moved to persistence-json.ts.
- * They are re-exported here for backward compatibility during the transition.
- *
- * SQLite is the sole canonical runtime store.
- * JSON/JSONL files are only used for migration, export/import, and debug.
- */
-
-import { promises as fs } from 'node:fs';
-import path from 'node:path';
-import os from 'node:os';
-
-// ── Re-export JSON helpers for backward compat ────────────────────
-export {
-  getGraphFilePath,
-  saveGraphJsonl,
-  loadGraphJsonl,
-  saveObservationsJson,
-  loadObservationsJson,
-  saveIdCounter,
-  loadIdCounter,
-  saveMiniSkillsJson,
-  loadMiniSkillsJson,
-  loadMiniSkillsCounter,
-  saveMiniSkillsCounter,
-  saveSessionsJson,
-  loadSessionsJson,
-} from './persistence-json.js';
-
-/** Default base data directory — overridable via MEMORIX_DATA_DIR env var */
-function resolveDefaultDataDir(): string {
-  return process.env.MEMORIX_DATA_DIR || path.join(os.homedir(), '.memorix', 'data');
-}
-
-/**
- * Sanitize a projectId for use as a directory name.
- * Used only during migration to locate legacy per-project subdirectories.
- */
-function sanitizeProjectId(projectId: string): string {
-  return projectId.replace(/\//g, '--').replace(/[<>:"|?*\\]/g, '_');
-}
-
-/**
- * Get the data directory for Memorix storage.
- *
- * Returns the FLAT global directory (~/.memorix/data/) regardless of projectId.
- * projectId is stored as metadata inside observations, not used for directory partitioning.
- * This ensures all IDEs share the same data directory even if they detect different projectIds.
- *
- * @param _projectId - Ignored for directory purposes (kept for API compat)
- */
-export async function getProjectDataDir(_projectId: string, baseDir?: string): Promise<string> {
-  // All projects share one flat data directory — projectId is metadata only
-  const base = baseDir ?? resolveDefaultDataDir();
-  await fs.mkdir(base, { recursive: true });
-  return base;
-}
-
-/**
- * Get the base data directory (parent of all project dirs).
- */
-export function getBaseDataDir(baseDir?: string): string {
-  return baseDir ?? resolveDefaultDataDir();
-}
-
-/**
- * List all project data directories.
- * Used for cross-project (global) search.
- */
-export async function listProjectDirs(baseDir?: string): Promise<string[]> {
-  const base = baseDir ?? resolveDefaultDataDir();
-  try {
-    const entries = await fs.readdir(base, { withFileTypes: true });
-    return entries
-      .filter((e) => e.isDirectory())
-      .map((e) => path.join(base, e.name));
-  } catch {
-    return [];
-  }
-}
-
-/**
- * Migrate legacy per-project subdirectories into the flat base directory.
- *
- * Before v0.9.6, data was stored in per-project subdirectories:
- *   ~/.memorix/data/AVIDS2--memorix/observations.json
- *   ~/.memorix/data/local--myproject/observations.json
- *
- * This caused data fragmentation when different IDEs detected different projectIds.
- * Now all data lives in ~/.memorix/data/ directly.
- *
- * Migration:
- *   1. Scan all subdirectories under base dir
- *   2. Merge observations from all subdirs into base dir (remap IDs to avoid collision)
- *   3. Merge graph.jsonl (deduplicate entities by name)
- *   4. Move subdirectories to .migrated-subdirs/ backup
- */
-export async function migrateSubdirsToFlat(baseDir?: string): Promise<boolean> {
-  const base = baseDir ?? resolveDefaultDataDir();
-  await fs.mkdir(base, { recursive: true });
-
-  // Find all subdirectories that contain observations.json
-  let entries: import('node:fs').Dirent[];
-  try {
-    entries = await fs.readdir(base, { withFileTypes: true });
-  } catch {
-    return false;
-  }
-
-  const dataDirs = entries
-    .filter((e) => e.isDirectory() && !e.name.startsWith('.'))
-    .map((e) => path.join(base, e.name));
-
-  if (dataDirs.length === 0) return false;
-
-  // Check which subdirs actually have observation data
-  const subdirData: Array<{ dir: string; obs: any[]; graph: { entities: any[]; relations: any[] } }> = [];
-  for (const dir of dataDirs) {
-    const obsPath = path.join(dir, 'observations.json');
-    try {
-      const data = await fs.readFile(obsPath, 'utf-8');
-      const obs = JSON.parse(data);
-      if (Array.isArray(obs) && obs.length > 0) {
-        // Also try to load graph
-        let graph = { entities: [] as any[], relations: [] as any[] };
-        try {
-          const graphData = await fs.readFile(path.join(dir, 'graph.jsonl'), 'utf-8');
-          const lines = graphData.split('\n').filter((l: string) => l.trim());
-          for (const line of lines) {
-            const item = JSON.parse(line);
-            if (item.type === 'entity') graph.entities.push(item);
-            if (item.type === 'relation') graph.relations.push(item);
-          }
-        } catch { /* no graph */ }
-        subdirData.push({ dir, obs, graph });
-      }
-    } catch { /* no observations */ }
-  }
-
-  if (subdirData.length === 0) return false;
-
-  // Load existing base-level data (if any)
-  let baseObs: any[] = [];
-  try {
-    const data = await fs.readFile(path.join(base, 'observations.json'), 'utf-8');
-    baseObs = JSON.parse(data);
-    if (!Array.isArray(baseObs)) baseObs = [];
-  } catch { /* no existing base data */ }
-
-  let baseGraph = { entities: [] as any[], relations: [] as any[] };
-  try {
-    const graphData = await fs.readFile(path.join(base, 'graph.jsonl'), 'utf-8');
-    const lines = graphData.split('\n').filter((l: string) => l.trim());
-    for (const line of lines) {
-      const item = JSON.parse(line);
-      if (item.type === 'entity') baseGraph.entities.push(item);
-      if (item.type === 'relation') baseGraph.relations.push(item);
-    }
-  } catch { /* no graph */ }
-
-  // Merge all observations: collect, sort by createdAt, remap IDs
-  const allObs: any[] = [...baseObs];
-  for (const { obs } of subdirData) {
-    for (const o of obs) {
-      // Deduplicate by title+createdAt+projectId (same observation from migration overlap)
-      const isDuplicate = allObs.some(
-        (existing) => existing.title === o.title && existing.createdAt === o.createdAt,
-      );
-      if (!isDuplicate) {
-        allObs.push(o);
-      }
-    }
-  }
-
-  // Sort by createdAt then remap IDs sequentially
-  allObs.sort((a, b) => (a.createdAt || '').localeCompare(b.createdAt || ''));
-  for (let i = 0; i < allObs.length; i++) {
-    allObs[i].id = i + 1;
-  }
-
-  // Merge graphs (deduplicate entities by name)
-  const entityMap = new Map<string, any>();
-  for (const e of baseGraph.entities) entityMap.set(e.name, e);
-  for (const { graph } of subdirData) {
-    for (const e of graph.entities) {
-      if (!entityMap.has(e.name)) {
-        entityMap.set(e.name, e);
-      } else {
-        // Merge observations lists
-        const existing = entityMap.get(e.name);
-        const obsSet = new Set([...(existing.observations || []), ...(e.observations || [])]);
-        existing.observations = [...obsSet];
-      }
-    }
-  }
-
-  const relationSet = new Set<string>();
-  const mergedRelations: any[] = [];
-  for (const rel of [...baseGraph.relations, ...subdirData.flatMap((d) => d.graph.relations)]) {
-    const key = `${rel.from}|${rel.to}|${rel.relationType}`;
-    if (!relationSet.has(key)) {
-      relationSet.add(key);
-      mergedRelations.push(rel);
-    }
-  }
-
-  // Write merged data to base directory
-  await fs.writeFile(path.join(base, 'observations.json'), JSON.stringify(allObs, null, 2), 'utf-8');
-  await fs.writeFile(
-    path.join(base, 'counter.json'),
-    JSON.stringify({ nextId: allObs.length + 1 }),
-    'utf-8',
-  );
-
-  // Write merged graph
-  const graphLines = [
-    ...[...entityMap.values()].map((e) => JSON.stringify({ type: 'entity', name: e.name, entityType: e.entityType, observations: e.observations })),
-    ...mergedRelations.map((r) => JSON.stringify({ type: 'relation', from: r.from, to: r.to, relationType: r.relationType })),
-  ];
-  if (graphLines.length > 0) {
-    await fs.writeFile(path.join(base, 'graph.jsonl'), graphLines.join('\n'), 'utf-8');
-  }
-
-  // Also merge sessions if present
-  let allSessions: any[] = [];
-  try {
-    const data = await fs.readFile(path.join(base, 'sessions.json'), 'utf-8');
-    allSessions = JSON.parse(data);
-    if (!Array.isArray(allSessions)) allSessions = [];
-  } catch { /* no sessions */ }
-  for (const { dir } of subdirData) {
-    try {
-      const data = await fs.readFile(path.join(dir, 'sessions.json'), 'utf-8');
-      const sessions = JSON.parse(data);
-      if (Array.isArray(sessions)) allSessions.push(...sessions);
-    } catch { /* no sessions */ }
-  }
-  if (allSessions.length > 0) {
-    await fs.writeFile(path.join(base, 'sessions.json'), JSON.stringify(allSessions, null, 2), 'utf-8');
-  }
-
-  // Move subdirectories to backup
-  const backupDir = path.join(base, '.migrated-subdirs');
-  await fs.mkdir(backupDir, { recursive: true });
-  for (const { dir } of subdirData) {
-    const dirName = path.basename(dir);
-    try {
-      await fs.rename(dir, path.join(backupDir, dirName));
-    } catch {
-      // If rename fails (cross-device), try to just leave it
-      // The important thing is the merged data is written
-    }
-  }
-
-  // Also move remaining empty subdirectories
-  for (const dir of dataDirs) {
-    const dirName = path.basename(dir);
-    try {
-      await fs.access(dir);
-      await fs.rename(dir, path.join(backupDir, dirName));
-    } catch { /* already moved or doesn't exist */ }
-  }
-
-  return true;
-}
-
-/**
- * Get the file path for the Orama database file.
- */
-export function getDbFilePath(projectDir: string): string {
-  return path.join(projectDir, 'memorix.msp');
-}
-
-/**
- * Check if a database file exists for the given project.
- */
-export async function hasExistingData(projectDir: string): Promise<boolean> {
-  try {
-    await fs.access(getDbFilePath(projectDir));
-    return true;
-  } catch {
-    return false;
-  }
-}
+/**
+ * Persistence Layer — Runtime + Migration
+ *
+ * Runtime responsibilities:
+ *   - Data directory resolution (flat global ~/.memorix/data/)
+ *   - Orama DB file path helpers
+ *   - Legacy per-project subdirectory migration
+ *
+ * JSON I/O helpers have been moved to persistence-json.ts.
+ * They are re-exported here for backward compatibility during the transition.
+ *
+ * SQLite is the sole canonical runtime store.
+ * JSON/JSONL files are only used for migration, export/import, and debug.
+ */
+
+import { promises as fs } from 'node:fs';
+import path from 'node:path';
+import os from 'node:os';
+
+// ── Re-export JSON helpers for backward compat ────────────────────
+export {
+  getGraphFilePath,
+  saveGraphJsonl,
+  loadGraphJsonl,
+  saveObservationsJson,
+  loadObservationsJson,
+  saveIdCounter,
+  loadIdCounter,
+  saveMiniSkillsJson,
+  loadMiniSkillsJson,
+  loadMiniSkillsCounter,
+  saveMiniSkillsCounter,
+  saveSessionsJson,
+  loadSessionsJson,
+} from './persistence-json.js';
+
+/** Default base data directory — overridable via MEMORIX_DATA_DIR env var */
+function resolveDefaultDataDir(): string {
+  return process.env.MEMORIX_DATA_DIR || path.join(os.homedir(), '.memorix', 'data');
+}
+
+/**
+ * Sanitize a projectId for use as a directory name.
+ * Used only during migration to locate legacy per-project subdirectories.
+ */
+function sanitizeProjectId(projectId: string): string {
+  return projectId.replace(/\//g, '--').replace(/[<>:"|?*\\]/g, '_');
+}
+
+/**
+ * Get the data directory for Memorix storage.
+ *
+ * Returns the FLAT global directory (~/.memorix/data/) regardless of projectId.
+ * projectId is stored as metadata inside observations, not used for directory partitioning.
+ * This ensures all IDEs share the same data directory even if they detect different projectIds.
+ *
+ * @param _projectId - Ignored for directory purposes (kept for API compat)
+ */
+export async function getProjectDataDir(_projectId: string, baseDir?: string): Promise<string> {
+  // All projects share one flat data directory — projectId is metadata only
+  const base = baseDir ?? resolveDefaultDataDir();
+  await fs.mkdir(base, { recursive: true });
+  return base;
+}
+
+/**
+ * Get the base data directory (parent of all project dirs).
+ */
+export function getBaseDataDir(baseDir?: string): string {
+  return baseDir ?? resolveDefaultDataDir();
+}
+
+/**
+ * List all project data directories.
+ * Used for cross-project (global) search.
+ */
+export async function listProjectDirs(baseDir?: string): Promise<string[]> {
+  const base = baseDir ?? resolveDefaultDataDir();
+  try {
+    const entries = await fs.readdir(base, { withFileTypes: true });
+    return entries
+      .filter((e) => e.isDirectory())
+      .map((e) => path.join(base, e.name));
+  } catch {
+    return [];
+  }
+}
+
+/**
+ * Migrate legacy per-project subdirectories into the flat base directory.
+ *
+ * Before v0.9.6, data was stored in per-project subdirectories:
+ *   ~/.memorix/data/AVIDS2--memorix/observations.json
+ *   ~/.memorix/data/local--myproject/observations.json
+ *
+ * This caused data fragmentation when different IDEs detected different projectIds.
+ * Now all data lives in ~/.memorix/data/ directly.
+ *
+ * Migration:
+ *   1. Scan all subdirectories under base dir
+ *   2. Merge observations from all subdirs into base dir (remap IDs to avoid collision)
+ *   3. Merge graph.jsonl (deduplicate entities by name)
+ *   4. Move subdirectories to .migrated-subdirs/ backup
+ */
+export async function migrateSubdirsToFlat(baseDir?: string): Promise<boolean> {
+  const base = baseDir ?? resolveDefaultDataDir();
+  await fs.mkdir(base, { recursive: true });
+
+  // Find all subdirectories that contain observations.json
+  let entries: import('node:fs').Dirent[];
+  try {
+    entries = await fs.readdir(base, { withFileTypes: true });
+  } catch {
+    return false;
+  }
+
+  const dataDirs = entries
+    .filter((e) => e.isDirectory() && !e.name.startsWith('.'))
+    .map((e) => path.join(base, e.name));
+
+  if (dataDirs.length === 0) return false;
+
+  // Check which subdirs actually have observation data
+  const subdirData: Array<{ dir: string; obs: any[]; graph: { entities: any[]; relations: any[] } }> = [];
+  for (const dir of dataDirs) {
+    const obsPath = path.join(dir, 'observations.json');
+    try {
+      const data = await fs.readFile(obsPath, 'utf-8');
+      const obs = JSON.parse(data);
+      if (Array.isArray(obs) && obs.length > 0) {
+        // Also try to load graph
+        let graph = { entities: [] as any[], relations: [] as any[] };
+        try {
+          const graphData = await fs.readFile(path.join(dir, 'graph.jsonl'), 'utf-8');
+          const lines = graphData.split('\n').filter((l: string) => l.trim());
+          for (const line of lines) {
+            const item = JSON.parse(line);
+            if (item.type === 'entity') graph.entities.push(item);
+            if (item.type === 'relation') graph.relations.push(item);
+          }
+        } catch { /* no graph */ }
+        subdirData.push({ dir, obs, graph });
+      }
+    } catch { /* no observations */ }
+  }
+
+  if (subdirData.length === 0) return false;
+
+  // Load existing base-level data (if any)
+  let baseObs: any[] = [];
+  try {
+    const data = await fs.readFile(path.join(base, 'observations.json'), 'utf-8');
+    baseObs = JSON.parse(data);
+    if (!Array.isArray(baseObs)) baseObs = [];
+  } catch { /* no existing base data */ }
+
+  let baseGraph = { entities: [] as any[], relations: [] as any[] };
+  try {
+    const graphData = await fs.readFile(path.join(base, 'graph.jsonl'), 'utf-8');
+    const lines = graphData.split('\n').filter((l: string) => l.trim());
+    for (const line of lines) {
+      const item = JSON.parse(line);
+      if (item.type === 'entity') baseGraph.entities.push(item);
+      if (item.type === 'relation') baseGraph.relations.push(item);
+    }
+  } catch { /* no graph */ }
+
+  // Merge all observations: collect, sort by createdAt, remap IDs
+  const allObs: any[] = [...baseObs];
+  for (const { obs } of subdirData) {
+    for (const o of obs) {
+      // Deduplicate by title+createdAt+projectId (same observation from migration overlap)
+      const isDuplicate = allObs.some(
+        (existing) => existing.title === o.title && existing.createdAt === o.createdAt,
+      );
+      if (!isDuplicate) {
+        allObs.push(o);
+      }
+    }
+  }
+
+  // Sort by createdAt then remap IDs sequentially
+  allObs.sort((a, b) => (a.createdAt || '').localeCompare(b.createdAt || ''));
+  for (let i = 0; i < allObs.length; i++) {
+    allObs[i].id = i + 1;
+  }
+
+  // Merge graphs (deduplicate entities by name)
+  const entityMap = new Map<string, any>();
+  for (const e of baseGraph.entities) entityMap.set(e.name, e);
+  for (const { graph } of subdirData) {
+    for (const e of graph.entities) {
+      if (!entityMap.has(e.name)) {
+        entityMap.set(e.name, e);
+      } else {
+        // Merge observations lists
+        const existing = entityMap.get(e.name);
+        const obsSet = new Set([...(existing.observations || []), ...(e.observations || [])]);
+        existing.observations = [...obsSet];
+      }
+    }
+  }
+
+  const relationSet = new Set<string>();
+  const mergedRelations: any[] = [];
+  for (const rel of [...baseGraph.relations, ...subdirData.flatMap((d) => d.graph.relations)]) {
+    const key = `${rel.from}|${rel.to}|${rel.relationType}`;
+    if (!relationSet.has(key)) {
+      relationSet.add(key);
+      mergedRelations.push(rel);
+    }
+  }
+
+  // Write merged data to base directory
+  await fs.writeFile(path.join(base, 'observations.json'), JSON.stringify(allObs, null, 2), 'utf-8');
+  await fs.writeFile(
+    path.join(base, 'counter.json'),
+    JSON.stringify({ nextId: allObs.length + 1 }),
+    'utf-8',
+  );
+
+  // Write merged graph
+  const graphLines = [
+    ...[...entityMap.values()].map((e) => JSON.stringify({ type: 'entity', name: e.name, entityType: e.entityType, observations: e.observations })),
+    ...mergedRelations.map((r) => JSON.stringify({ type: 'relation', from: r.from, to: r.to, relationType: r.relationType })),
+  ];
+  if (graphLines.length > 0) {
+    await fs.writeFile(path.join(base, 'graph.jsonl'), graphLines.join('\n'), 'utf-8');
+  }
+
+  // Also merge sessions if present
+  let allSessions: any[] = [];
+  try {
+    const data = await fs.readFile(path.join(base, 'sessions.json'), 'utf-8');
+    allSessions = JSON.parse(data);
+    if (!Array.isArray(allSessions)) allSessions = [];
+  } catch { /* no sessions */ }
+  for (const { dir } of subdirData) {
+    try {
+      const data = await fs.readFile(path.join(dir, 'sessions.json'), 'utf-8');
+      const sessions = JSON.parse(data);
+      if (Array.isArray(sessions)) allSessions.push(...sessions);
+    } catch { /* no sessions */ }
+  }
+  if (allSessions.length > 0) {
+    await fs.writeFile(path.join(base, 'sessions.json'), JSON.stringify(allSessions, null, 2), 'utf-8');
+  }
+
+  // Move subdirectories to backup
+  const backupDir = path.join(base, '.migrated-subdirs');
+  await fs.mkdir(backupDir, { recursive: true });
+  for (const { dir } of subdirData) {
+    const dirName = path.basename(dir);
+    try {
+      await fs.rename(dir, path.join(backupDir, dirName));
+    } catch {
+      // If rename fails (cross-device), try to just leave it
+      // The important thing is the merged data is written
+    }
+  }
+
+  // Also move remaining empty subdirectories
+  for (const dir of dataDirs) {
+    const dirName = path.basename(dir);
+    try {
+      await fs.access(dir);
+      await fs.rename(dir, path.join(backupDir, dirName));
+    } catch { /* already moved or doesn't exist */ }
+  }
+
+  return true;
+}
+
+/**
+ * Get the file path for the Orama database file.
+ */
+export function getDbFilePath(projectDir: string): string {
+  return path.join(projectDir, 'memorix.msp');
+}
+
+/**
+ * Check if a database file exists for the given project.
+ */
+export async function hasExistingData(projectDir: string): Promise<boolean> {
+  try {
+    await fs.access(getDbFilePath(projectDir));
+    return true;
+  } catch {
+    return false;
+  }
+}