Spaces:
Paused
Paused
| // @ts-nocheck | |
| // oxlint-disable eslint/no-unused-vars, typescript/no-explicit-any | |
| import type { DatabaseSync } from "node:sqlite"; | |
| import chokidar, { type FSWatcher } from "chokidar"; | |
| import { randomUUID } from "node:crypto"; | |
| import fsSync from "node:fs"; | |
| import fs from "node:fs/promises"; | |
| import path from "node:path"; | |
| import type { MemorySource, MemorySyncProgressUpdate } from "./types.js"; | |
| import { resolveSessionTranscriptsDirForAgent } from "../config/sessions/paths.js"; | |
| import { createSubsystemLogger } from "../logging/subsystem.js"; | |
| import { onSessionTranscriptUpdate } from "../sessions/transcript-events.js"; | |
| import { resolveUserPath } from "../utils.js"; | |
| import { | |
| buildFileEntry, | |
| ensureDir, | |
| isMemoryPath, | |
| listMemoryFiles, | |
| normalizeExtraMemoryPaths, | |
| parseEmbedding, | |
| remapChunkLines, | |
| runWithConcurrency, | |
| type MemoryFileEntry, | |
| } from "./internal.js"; | |
| import { ensureMemoryIndexSchema } from "./memory-schema.js"; | |
| import { | |
| buildSessionEntry, | |
| listSessionFilesForAgent, | |
| sessionPathForFile, | |
| type SessionFileEntry, | |
| } from "./session-files.js"; | |
| import { loadSqliteVecExtension } from "./sqlite-vec.js"; | |
| import { requireNodeSqlite } from "./sqlite.js"; | |
| type MemoryIndexMeta = { | |
| model: string; | |
| provider: string; | |
| providerKey?: string; | |
| chunkTokens: number; | |
| chunkOverlap: number; | |
| vectorDims?: number; | |
| }; | |
| type MemorySyncProgressState = { | |
| completed: number; | |
| total: number; | |
| label?: string; | |
| report: (update: MemorySyncProgressUpdate) => void; | |
| }; | |
| const META_KEY = "memory_index_meta_v1"; | |
| const VECTOR_TABLE = "chunks_vec"; | |
| const FTS_TABLE = "chunks_fts"; | |
| const EMBEDDING_CACHE_TABLE = "embedding_cache"; | |
| const SESSION_DIRTY_DEBOUNCE_MS = 5000; | |
| const SESSION_DELTA_READ_CHUNK_BYTES = 64 * 1024; | |
| const VECTOR_LOAD_TIMEOUT_MS = 30_000; | |
| const log = createSubsystemLogger("memory"); | |
| class MemoryManagerSyncOps { | |
| [key: string]: any; | |
| private async ensureVectorReady(dimensions?: number): Promise<boolean> { | |
| if (!this.vector.enabled) { | |
| return false; | |
| } | |
| if (!this.vectorReady) { | |
| this.vectorReady = this.withTimeout( | |
| this.loadVectorExtension(), | |
| VECTOR_LOAD_TIMEOUT_MS, | |
| `sqlite-vec load timed out after ${Math.round(VECTOR_LOAD_TIMEOUT_MS / 1000)}s`, | |
| ); | |
| } | |
| let ready = false; | |
| try { | |
| ready = await this.vectorReady; | |
| } catch (err) { | |
| const message = err instanceof Error ? err.message : String(err); | |
| this.vector.available = false; | |
| this.vector.loadError = message; | |
| this.vectorReady = null; | |
| log.warn(`sqlite-vec unavailable: ${message}`); | |
| return false; | |
| } | |
| if (ready && typeof dimensions === "number" && dimensions > 0) { | |
| this.ensureVectorTable(dimensions); | |
| } | |
| return ready; | |
| } | |
| private async loadVectorExtension(): Promise<boolean> { | |
| if (this.vector.available !== null) { | |
| return this.vector.available; | |
| } | |
| if (!this.vector.enabled) { | |
| this.vector.available = false; | |
| return false; | |
| } | |
| try { | |
| const resolvedPath = this.vector.extensionPath?.trim() | |
| ? resolveUserPath(this.vector.extensionPath) | |
| : undefined; | |
| const loaded = await loadSqliteVecExtension({ db: this.db, extensionPath: resolvedPath }); | |
| if (!loaded.ok) { | |
| throw new Error(loaded.error ?? "unknown sqlite-vec load error"); | |
| } | |
| this.vector.extensionPath = loaded.extensionPath; | |
| this.vector.available = true; | |
| return true; | |
| } catch (err) { | |
| const message = err instanceof Error ? err.message : String(err); | |
| this.vector.available = false; | |
| this.vector.loadError = message; | |
| log.warn(`sqlite-vec unavailable: ${message}`); | |
| return false; | |
| } | |
| } | |
| private ensureVectorTable(dimensions: number): void { | |
| if (this.vector.dims === dimensions) { | |
| return; | |
| } | |
| if (this.vector.dims && this.vector.dims !== dimensions) { | |
| this.dropVectorTable(); | |
| } | |
| this.db.exec( | |
| `CREATE VIRTUAL TABLE IF NOT EXISTS ${VECTOR_TABLE} USING vec0(\n` + | |
| ` id TEXT PRIMARY KEY,\n` + | |
| ` embedding FLOAT[${dimensions}]\n` + | |
| `)`, | |
| ); | |
| this.vector.dims = dimensions; | |
| } | |
| private dropVectorTable(): void { | |
| try { | |
| this.db.exec(`DROP TABLE IF EXISTS ${VECTOR_TABLE}`); | |
| } catch (err) { | |
| const message = err instanceof Error ? err.message : String(err); | |
| log.debug(`Failed to drop ${VECTOR_TABLE}: ${message}`); | |
| } | |
| } | |
| private buildSourceFilter(alias?: string): { sql: string; params: MemorySource[] } { | |
| const sources = Array.from(this.sources); | |
| if (sources.length === 0) { | |
| return { sql: "", params: [] }; | |
| } | |
| const column = alias ? `${alias}.source` : "source"; | |
| const placeholders = sources.map(() => "?").join(", "); | |
| return { sql: ` AND ${column} IN (${placeholders})`, params: sources }; | |
| } | |
| private openDatabase(): DatabaseSync { | |
| const dbPath = resolveUserPath(this.settings.store.path); | |
| return this.openDatabaseAtPath(dbPath); | |
| } | |
| private openDatabaseAtPath(dbPath: string): DatabaseSync { | |
| const dir = path.dirname(dbPath); | |
| ensureDir(dir); | |
| const { DatabaseSync } = requireNodeSqlite(); | |
| return new DatabaseSync(dbPath, { allowExtension: this.settings.store.vector.enabled }); | |
| } | |
| private seedEmbeddingCache(sourceDb: DatabaseSync): void { | |
| if (!this.cache.enabled) { | |
| return; | |
| } | |
| try { | |
| const rows = sourceDb | |
| .prepare( | |
| `SELECT provider, model, provider_key, hash, embedding, dims, updated_at FROM ${EMBEDDING_CACHE_TABLE}`, | |
| ) | |
| .all() as Array<{ | |
| provider: string; | |
| model: string; | |
| provider_key: string; | |
| hash: string; | |
| embedding: string; | |
| dims: number | null; | |
| updated_at: number; | |
| }>; | |
| if (!rows.length) { | |
| return; | |
| } | |
| const insert = this.db.prepare( | |
| `INSERT INTO ${EMBEDDING_CACHE_TABLE} (provider, model, provider_key, hash, embedding, dims, updated_at) | |
| VALUES (?, ?, ?, ?, ?, ?, ?) | |
| ON CONFLICT(provider, model, provider_key, hash) DO UPDATE SET | |
| embedding=excluded.embedding, | |
| dims=excluded.dims, | |
| updated_at=excluded.updated_at`, | |
| ); | |
| this.db.exec("BEGIN"); | |
| for (const row of rows) { | |
| insert.run( | |
| row.provider, | |
| row.model, | |
| row.provider_key, | |
| row.hash, | |
| row.embedding, | |
| row.dims, | |
| row.updated_at, | |
| ); | |
| } | |
| this.db.exec("COMMIT"); | |
| } catch (err) { | |
| try { | |
| this.db.exec("ROLLBACK"); | |
| } catch {} | |
| throw err; | |
| } | |
| } | |
| private async swapIndexFiles(targetPath: string, tempPath: string): Promise<void> { | |
| const backupPath = `${targetPath}.backup-${randomUUID()}`; | |
| await this.moveIndexFiles(targetPath, backupPath); | |
| try { | |
| await this.moveIndexFiles(tempPath, targetPath); | |
| } catch (err) { | |
| await this.moveIndexFiles(backupPath, targetPath); | |
| throw err; | |
| } | |
| await this.removeIndexFiles(backupPath); | |
| } | |
| private async moveIndexFiles(sourceBase: string, targetBase: string): Promise<void> { | |
| const suffixes = ["", "-wal", "-shm"]; | |
| for (const suffix of suffixes) { | |
| const source = `${sourceBase}${suffix}`; | |
| const target = `${targetBase}${suffix}`; | |
| try { | |
| await fs.rename(source, target); | |
| } catch (err) { | |
| if ((err as NodeJS.ErrnoException).code !== "ENOENT") { | |
| throw err; | |
| } | |
| } | |
| } | |
| } | |
| private async removeIndexFiles(basePath: string): Promise<void> { | |
| const suffixes = ["", "-wal", "-shm"]; | |
| await Promise.all(suffixes.map((suffix) => fs.rm(`${basePath}${suffix}`, { force: true }))); | |
| } | |
| private ensureSchema() { | |
| const result = ensureMemoryIndexSchema({ | |
| db: this.db, | |
| embeddingCacheTable: EMBEDDING_CACHE_TABLE, | |
| ftsTable: FTS_TABLE, | |
| ftsEnabled: this.fts.enabled, | |
| }); | |
| this.fts.available = result.ftsAvailable; | |
| if (result.ftsError) { | |
| this.fts.loadError = result.ftsError; | |
| log.warn(`fts unavailable: ${result.ftsError}`); | |
| } | |
| } | |
| private ensureWatcher() { | |
| if (!this.sources.has("memory") || !this.settings.sync.watch || this.watcher) { | |
| return; | |
| } | |
| const additionalPaths = normalizeExtraMemoryPaths(this.workspaceDir, this.settings.extraPaths) | |
| .map((entry) => { | |
| try { | |
| const stat = fsSync.lstatSync(entry); | |
| return stat.isSymbolicLink() ? null : entry; | |
| } catch { | |
| return null; | |
| } | |
| }) | |
| .filter((entry): entry is string => Boolean(entry)); | |
| const watchPaths = new Set<string>([ | |
| path.join(this.workspaceDir, "MEMORY.md"), | |
| path.join(this.workspaceDir, "memory.md"), | |
| path.join(this.workspaceDir, "memory"), | |
| ...additionalPaths, | |
| ]); | |
| this.watcher = chokidar.watch(Array.from(watchPaths), { | |
| ignoreInitial: true, | |
| awaitWriteFinish: { | |
| stabilityThreshold: this.settings.sync.watchDebounceMs, | |
| pollInterval: 100, | |
| }, | |
| }); | |
| const markDirty = () => { | |
| this.dirty = true; | |
| this.scheduleWatchSync(); | |
| }; | |
| this.watcher.on("add", markDirty); | |
| this.watcher.on("change", markDirty); | |
| this.watcher.on("unlink", markDirty); | |
| } | |
| private ensureSessionListener() { | |
| if (!this.sources.has("sessions") || this.sessionUnsubscribe) { | |
| return; | |
| } | |
| this.sessionUnsubscribe = onSessionTranscriptUpdate((update) => { | |
| if (this.closed) { | |
| return; | |
| } | |
| const sessionFile = update.sessionFile; | |
| if (!this.isSessionFileForAgent(sessionFile)) { | |
| return; | |
| } | |
| this.scheduleSessionDirty(sessionFile); | |
| }); | |
| } | |
| private scheduleSessionDirty(sessionFile: string) { | |
| this.sessionPendingFiles.add(sessionFile); | |
| if (this.sessionWatchTimer) { | |
| return; | |
| } | |
| this.sessionWatchTimer = setTimeout(() => { | |
| this.sessionWatchTimer = null; | |
| void this.processSessionDeltaBatch().catch((err) => { | |
| log.warn(`memory session delta failed: ${String(err)}`); | |
| }); | |
| }, SESSION_DIRTY_DEBOUNCE_MS); | |
| } | |
| private async processSessionDeltaBatch(): Promise<void> { | |
| if (this.sessionPendingFiles.size === 0) { | |
| return; | |
| } | |
| const pending = Array.from(this.sessionPendingFiles); | |
| this.sessionPendingFiles.clear(); | |
| let shouldSync = false; | |
| for (const sessionFile of pending) { | |
| const delta = await this.updateSessionDelta(sessionFile); | |
| if (!delta) { | |
| continue; | |
| } | |
| const bytesThreshold = delta.deltaBytes; | |
| const messagesThreshold = delta.deltaMessages; | |
| const bytesHit = | |
| bytesThreshold <= 0 ? delta.pendingBytes > 0 : delta.pendingBytes >= bytesThreshold; | |
| const messagesHit = | |
| messagesThreshold <= 0 | |
| ? delta.pendingMessages > 0 | |
| : delta.pendingMessages >= messagesThreshold; | |
| if (!bytesHit && !messagesHit) { | |
| continue; | |
| } | |
| this.sessionsDirtyFiles.add(sessionFile); | |
| this.sessionsDirty = true; | |
| delta.pendingBytes = | |
| bytesThreshold > 0 ? Math.max(0, delta.pendingBytes - bytesThreshold) : 0; | |
| delta.pendingMessages = | |
| messagesThreshold > 0 ? Math.max(0, delta.pendingMessages - messagesThreshold) : 0; | |
| shouldSync = true; | |
| } | |
| if (shouldSync) { | |
| void this.sync({ reason: "session-delta" }).catch((err) => { | |
| log.warn(`memory sync failed (session-delta): ${String(err)}`); | |
| }); | |
| } | |
| } | |
| private async updateSessionDelta(sessionFile: string): Promise<{ | |
| deltaBytes: number; | |
| deltaMessages: number; | |
| pendingBytes: number; | |
| pendingMessages: number; | |
| } | null> { | |
| const thresholds = this.settings.sync.sessions; | |
| if (!thresholds) { | |
| return null; | |
| } | |
| let stat: { size: number }; | |
| try { | |
| stat = await fs.stat(sessionFile); | |
| } catch { | |
| return null; | |
| } | |
| const size = stat.size; | |
| let state = this.sessionDeltas.get(sessionFile); | |
| if (!state) { | |
| state = { lastSize: 0, pendingBytes: 0, pendingMessages: 0 }; | |
| this.sessionDeltas.set(sessionFile, state); | |
| } | |
| const deltaBytes = Math.max(0, size - state.lastSize); | |
| if (deltaBytes === 0 && size === state.lastSize) { | |
| return { | |
| deltaBytes: thresholds.deltaBytes, | |
| deltaMessages: thresholds.deltaMessages, | |
| pendingBytes: state.pendingBytes, | |
| pendingMessages: state.pendingMessages, | |
| }; | |
| } | |
| if (size < state.lastSize) { | |
| state.lastSize = size; | |
| state.pendingBytes += size; | |
| const shouldCountMessages = | |
| thresholds.deltaMessages > 0 && | |
| (thresholds.deltaBytes <= 0 || state.pendingBytes < thresholds.deltaBytes); | |
| if (shouldCountMessages) { | |
| state.pendingMessages += await this.countNewlines(sessionFile, 0, size); | |
| } | |
| } else { | |
| state.pendingBytes += deltaBytes; | |
| const shouldCountMessages = | |
| thresholds.deltaMessages > 0 && | |
| (thresholds.deltaBytes <= 0 || state.pendingBytes < thresholds.deltaBytes); | |
| if (shouldCountMessages) { | |
| state.pendingMessages += await this.countNewlines(sessionFile, state.lastSize, size); | |
| } | |
| state.lastSize = size; | |
| } | |
| this.sessionDeltas.set(sessionFile, state); | |
| return { | |
| deltaBytes: thresholds.deltaBytes, | |
| deltaMessages: thresholds.deltaMessages, | |
| pendingBytes: state.pendingBytes, | |
| pendingMessages: state.pendingMessages, | |
| }; | |
| } | |
| private async countNewlines(absPath: string, start: number, end: number): Promise<number> { | |
| if (end <= start) { | |
| return 0; | |
| } | |
| const handle = await fs.open(absPath, "r"); | |
| try { | |
| let offset = start; | |
| let count = 0; | |
| const buffer = Buffer.alloc(SESSION_DELTA_READ_CHUNK_BYTES); | |
| while (offset < end) { | |
| const toRead = Math.min(buffer.length, end - offset); | |
| const { bytesRead } = await handle.read(buffer, 0, toRead, offset); | |
| if (bytesRead <= 0) { | |
| break; | |
| } | |
| for (let i = 0; i < bytesRead; i += 1) { | |
| if (buffer[i] === 10) { | |
| count += 1; | |
| } | |
| } | |
| offset += bytesRead; | |
| } | |
| return count; | |
| } finally { | |
| await handle.close(); | |
| } | |
| } | |
| private resetSessionDelta(absPath: string, size: number): void { | |
| const state = this.sessionDeltas.get(absPath); | |
| if (!state) { | |
| return; | |
| } | |
| state.lastSize = size; | |
| state.pendingBytes = 0; | |
| state.pendingMessages = 0; | |
| } | |
| private isSessionFileForAgent(sessionFile: string): boolean { | |
| if (!sessionFile) { | |
| return false; | |
| } | |
| const sessionsDir = resolveSessionTranscriptsDirForAgent(this.agentId); | |
| const resolvedFile = path.resolve(sessionFile); | |
| const resolvedDir = path.resolve(sessionsDir); | |
| return resolvedFile.startsWith(`${resolvedDir}${path.sep}`); | |
| } | |
| private ensureIntervalSync() { | |
| const minutes = this.settings.sync.intervalMinutes; | |
| if (!minutes || minutes <= 0 || this.intervalTimer) { | |
| return; | |
| } | |
| const ms = minutes * 60 * 1000; | |
| this.intervalTimer = setInterval(() => { | |
| void this.sync({ reason: "interval" }).catch((err) => { | |
| log.warn(`memory sync failed (interval): ${String(err)}`); | |
| }); | |
| }, ms); | |
| } | |
| private scheduleWatchSync() { | |
| if (!this.sources.has("memory") || !this.settings.sync.watch) { | |
| return; | |
| } | |
| if (this.watchTimer) { | |
| clearTimeout(this.watchTimer); | |
| } | |
| this.watchTimer = setTimeout(() => { | |
| this.watchTimer = null; | |
| void this.sync({ reason: "watch" }).catch((err) => { | |
| log.warn(`memory sync failed (watch): ${String(err)}`); | |
| }); | |
| }, this.settings.sync.watchDebounceMs); | |
| } | |
| private shouldSyncSessions( | |
| params?: { reason?: string; force?: boolean }, | |
| needsFullReindex = false, | |
| ) { | |
| if (!this.sources.has("sessions")) { | |
| return false; | |
| } | |
| if (params?.force) { | |
| return true; | |
| } | |
| const reason = params?.reason; | |
| if (reason === "session-start" || reason === "watch") { | |
| return false; | |
| } | |
| if (needsFullReindex) { | |
| return true; | |
| } | |
| return this.sessionsDirty && this.sessionsDirtyFiles.size > 0; | |
| } | |
| private async syncMemoryFiles(params: { | |
| needsFullReindex: boolean; | |
| progress?: MemorySyncProgressState; | |
| }) { | |
| const files = await listMemoryFiles(this.workspaceDir, this.settings.extraPaths); | |
| const fileEntries = await Promise.all( | |
| files.map(async (file) => buildFileEntry(file, this.workspaceDir)), | |
| ); | |
| log.debug("memory sync: indexing memory files", { | |
| files: fileEntries.length, | |
| needsFullReindex: params.needsFullReindex, | |
| batch: this.batch.enabled, | |
| concurrency: this.getIndexConcurrency(), | |
| }); | |
| const activePaths = new Set(fileEntries.map((entry) => entry.path)); | |
| if (params.progress) { | |
| params.progress.total += fileEntries.length; | |
| params.progress.report({ | |
| completed: params.progress.completed, | |
| total: params.progress.total, | |
| label: this.batch.enabled ? "Indexing memory files (batch)..." : "Indexing memory files…", | |
| }); | |
| } | |
| const tasks = fileEntries.map((entry) => async () => { | |
| const record = this.db | |
| .prepare(`SELECT hash FROM files WHERE path = ? AND source = ?`) | |
| .get(entry.path, "memory") as { hash: string } | undefined; | |
| if (!params.needsFullReindex && record?.hash === entry.hash) { | |
| if (params.progress) { | |
| params.progress.completed += 1; | |
| params.progress.report({ | |
| completed: params.progress.completed, | |
| total: params.progress.total, | |
| }); | |
| } | |
| return; | |
| } | |
| await this.indexFile(entry, { source: "memory" }); | |
| if (params.progress) { | |
| params.progress.completed += 1; | |
| params.progress.report({ | |
| completed: params.progress.completed, | |
| total: params.progress.total, | |
| }); | |
| } | |
| }); | |
| await runWithConcurrency(tasks, this.getIndexConcurrency()); | |
| const staleRows = this.db | |
| .prepare(`SELECT path FROM files WHERE source = ?`) | |
| .all("memory") as Array<{ path: string }>; | |
| for (const stale of staleRows) { | |
| if (activePaths.has(stale.path)) { | |
| continue; | |
| } | |
| this.db.prepare(`DELETE FROM files WHERE path = ? AND source = ?`).run(stale.path, "memory"); | |
| try { | |
| this.db | |
| .prepare( | |
| `DELETE FROM ${VECTOR_TABLE} WHERE id IN (SELECT id FROM chunks WHERE path = ? AND source = ?)`, | |
| ) | |
| .run(stale.path, "memory"); | |
| } catch {} | |
| this.db.prepare(`DELETE FROM chunks WHERE path = ? AND source = ?`).run(stale.path, "memory"); | |
| if (this.fts.enabled && this.fts.available) { | |
| try { | |
| this.db | |
| .prepare(`DELETE FROM ${FTS_TABLE} WHERE path = ? AND source = ? AND model = ?`) | |
| .run(stale.path, "memory", this.provider.model); | |
| } catch {} | |
| } | |
| } | |
| } | |
| private async syncSessionFiles(params: { | |
| needsFullReindex: boolean; | |
| progress?: MemorySyncProgressState; | |
| }) { | |
| const files = await listSessionFilesForAgent(this.agentId); | |
| const activePaths = new Set(files.map((file) => sessionPathForFile(file))); | |
| const indexAll = params.needsFullReindex || this.sessionsDirtyFiles.size === 0; | |
| log.debug("memory sync: indexing session files", { | |
| files: files.length, | |
| indexAll, | |
| dirtyFiles: this.sessionsDirtyFiles.size, | |
| batch: this.batch.enabled, | |
| concurrency: this.getIndexConcurrency(), | |
| }); | |
| if (params.progress) { | |
| params.progress.total += files.length; | |
| params.progress.report({ | |
| completed: params.progress.completed, | |
| total: params.progress.total, | |
| label: this.batch.enabled ? "Indexing session files (batch)..." : "Indexing session files…", | |
| }); | |
| } | |
| const tasks = files.map((absPath) => async () => { | |
| if (!indexAll && !this.sessionsDirtyFiles.has(absPath)) { | |
| if (params.progress) { | |
| params.progress.completed += 1; | |
| params.progress.report({ | |
| completed: params.progress.completed, | |
| total: params.progress.total, | |
| }); | |
| } | |
| return; | |
| } | |
| const entry = await buildSessionEntry(absPath); | |
| if (!entry) { | |
| if (params.progress) { | |
| params.progress.completed += 1; | |
| params.progress.report({ | |
| completed: params.progress.completed, | |
| total: params.progress.total, | |
| }); | |
| } | |
| return; | |
| } | |
| const record = this.db | |
| .prepare(`SELECT hash FROM files WHERE path = ? AND source = ?`) | |
| .get(entry.path, "sessions") as { hash: string } | undefined; | |
| if (!params.needsFullReindex && record?.hash === entry.hash) { | |
| if (params.progress) { | |
| params.progress.completed += 1; | |
| params.progress.report({ | |
| completed: params.progress.completed, | |
| total: params.progress.total, | |
| }); | |
| } | |
| this.resetSessionDelta(absPath, entry.size); | |
| return; | |
| } | |
| await this.indexFile(entry, { source: "sessions", content: entry.content }); | |
| this.resetSessionDelta(absPath, entry.size); | |
| if (params.progress) { | |
| params.progress.completed += 1; | |
| params.progress.report({ | |
| completed: params.progress.completed, | |
| total: params.progress.total, | |
| }); | |
| } | |
| }); | |
| await runWithConcurrency(tasks, this.getIndexConcurrency()); | |
| const staleRows = this.db | |
| .prepare(`SELECT path FROM files WHERE source = ?`) | |
| .all("sessions") as Array<{ path: string }>; | |
| for (const stale of staleRows) { | |
| if (activePaths.has(stale.path)) { | |
| continue; | |
| } | |
| this.db | |
| .prepare(`DELETE FROM files WHERE path = ? AND source = ?`) | |
| .run(stale.path, "sessions"); | |
| try { | |
| this.db | |
| .prepare( | |
| `DELETE FROM ${VECTOR_TABLE} WHERE id IN (SELECT id FROM chunks WHERE path = ? AND source = ?)`, | |
| ) | |
| .run(stale.path, "sessions"); | |
| } catch {} | |
| this.db | |
| .prepare(`DELETE FROM chunks WHERE path = ? AND source = ?`) | |
| .run(stale.path, "sessions"); | |
| if (this.fts.enabled && this.fts.available) { | |
| try { | |
| this.db | |
| .prepare(`DELETE FROM ${FTS_TABLE} WHERE path = ? AND source = ? AND model = ?`) | |
| .run(stale.path, "sessions", this.provider.model); | |
| } catch {} | |
| } | |
| } | |
| } | |
| private createSyncProgress( | |
| onProgress: (update: MemorySyncProgressUpdate) => void, | |
| ): MemorySyncProgressState { | |
| const state: MemorySyncProgressState = { | |
| completed: 0, | |
| total: 0, | |
| label: undefined, | |
| report: (update) => { | |
| if (update.label) { | |
| state.label = update.label; | |
| } | |
| const label = | |
| update.total > 0 && state.label | |
| ? `${state.label} ${update.completed}/${update.total}` | |
| : state.label; | |
| onProgress({ | |
| completed: update.completed, | |
| total: update.total, | |
| label, | |
| }); | |
| }, | |
| }; | |
| return state; | |
| } | |
| private async runSync(params?: { | |
| reason?: string; | |
| force?: boolean; | |
| progress?: (update: MemorySyncProgressUpdate) => void; | |
| }) { | |
| const progress = params?.progress ? this.createSyncProgress(params.progress) : undefined; | |
| if (progress) { | |
| progress.report({ | |
| completed: progress.completed, | |
| total: progress.total, | |
| label: "Loading vector extension…", | |
| }); | |
| } | |
| const vectorReady = await this.ensureVectorReady(); | |
| const meta = this.readMeta(); | |
| const needsFullReindex = | |
| params?.force || | |
| !meta || | |
| meta.model !== this.provider.model || | |
| meta.provider !== this.provider.id || | |
| meta.providerKey !== this.providerKey || | |
| meta.chunkTokens !== this.settings.chunking.tokens || | |
| meta.chunkOverlap !== this.settings.chunking.overlap || | |
| (vectorReady && !meta?.vectorDims); | |
| try { | |
| if (needsFullReindex) { | |
| await this.runSafeReindex({ | |
| reason: params?.reason, | |
| force: params?.force, | |
| progress: progress ?? undefined, | |
| }); | |
| return; | |
| } | |
| const shouldSyncMemory = | |
| this.sources.has("memory") && (params?.force || needsFullReindex || this.dirty); | |
| const shouldSyncSessions = this.shouldSyncSessions(params, needsFullReindex); | |
| if (shouldSyncMemory) { | |
| await this.syncMemoryFiles({ needsFullReindex, progress: progress ?? undefined }); | |
| this.dirty = false; | |
| } | |
| if (shouldSyncSessions) { | |
| await this.syncSessionFiles({ needsFullReindex, progress: progress ?? undefined }); | |
| this.sessionsDirty = false; | |
| this.sessionsDirtyFiles.clear(); | |
| } else if (this.sessionsDirtyFiles.size > 0) { | |
| this.sessionsDirty = true; | |
| } else { | |
| this.sessionsDirty = false; | |
| } | |
| } catch (err) { | |
| const reason = err instanceof Error ? err.message : String(err); | |
| const activated = | |
| this.shouldFallbackOnError(reason) && (await this.activateFallbackProvider(reason)); | |
| if (activated) { | |
| await this.runSafeReindex({ | |
| reason: params?.reason ?? "fallback", | |
| force: true, | |
| progress: progress ?? undefined, | |
| }); | |
| return; | |
| } | |
| throw err; | |
| } | |
| } | |
| private shouldFallbackOnError(message: string): boolean { | |
| return /embedding|embeddings|batch/i.test(message); | |
| } | |
| private resolveBatchConfig(): { | |
| enabled: boolean; | |
| wait: boolean; | |
| concurrency: number; | |
| pollIntervalMs: number; | |
| timeoutMs: number; | |
| } { | |
| const batch = this.settings.remote?.batch; | |
| const enabled = Boolean( | |
| batch?.enabled && | |
| ((this.openAi && this.provider.id === "openai") || | |
| (this.gemini && this.provider.id === "gemini") || | |
| (this.voyage && this.provider.id === "voyage")), | |
| ); | |
| return { | |
| enabled, | |
| wait: batch?.wait ?? true, | |
| concurrency: Math.max(1, batch?.concurrency ?? 2), | |
| pollIntervalMs: batch?.pollIntervalMs ?? 2000, | |
| timeoutMs: (batch?.timeoutMinutes ?? 60) * 60 * 1000, | |
| }; | |
| } | |
| private async activateFallbackProvider(reason: string): Promise<boolean> { | |
| const fallback = this.settings.fallback; | |
| if (!fallback || fallback === "none" || fallback === this.provider.id) { | |
| return false; | |
| } | |
| if (this.fallbackFrom) { | |
| return false; | |
| } | |
| const fallbackFrom = this.provider.id as "openai" | "gemini" | "local" | "voyage"; | |
| const fallbackModel = | |
| fallback === "gemini" | |
| ? DEFAULT_GEMINI_EMBEDDING_MODEL | |
| : fallback === "openai" | |
| ? DEFAULT_OPENAI_EMBEDDING_MODEL | |
| : fallback === "voyage" | |
| ? DEFAULT_VOYAGE_EMBEDDING_MODEL | |
| : this.settings.model; | |
| const fallbackResult = await createEmbeddingProvider({ | |
| config: this.cfg, | |
| agentDir: resolveAgentDir(this.cfg, this.agentId), | |
| provider: fallback, | |
| remote: this.settings.remote, | |
| model: fallbackModel, | |
| fallback: "none", | |
| local: this.settings.local, | |
| }); | |
| this.fallbackFrom = fallbackFrom; | |
| this.fallbackReason = reason; | |
| this.provider = fallbackResult.provider; | |
| this.openAi = fallbackResult.openAi; | |
| this.gemini = fallbackResult.gemini; | |
| this.voyage = fallbackResult.voyage; | |
| this.providerKey = this.computeProviderKey(); | |
| this.batch = this.resolveBatchConfig(); | |
| log.warn(`memory embeddings: switched to fallback provider (${fallback})`, { reason }); | |
| return true; | |
| } | |
| private async runSafeReindex(params: { | |
| reason?: string; | |
| force?: boolean; | |
| progress?: MemorySyncProgressState; | |
| }): Promise<void> { | |
| const dbPath = resolveUserPath(this.settings.store.path); | |
| const tempDbPath = `${dbPath}.tmp-${randomUUID()}`; | |
| const tempDb = this.openDatabaseAtPath(tempDbPath); | |
| const originalDb = this.db; | |
| let originalDbClosed = false; | |
| const originalState = { | |
| ftsAvailable: this.fts.available, | |
| ftsError: this.fts.loadError, | |
| vectorAvailable: this.vector.available, | |
| vectorLoadError: this.vector.loadError, | |
| vectorDims: this.vector.dims, | |
| vectorReady: this.vectorReady, | |
| }; | |
| const restoreOriginalState = () => { | |
| if (originalDbClosed) { | |
| this.db = this.openDatabaseAtPath(dbPath); | |
| } else { | |
| this.db = originalDb; | |
| } | |
| this.fts.available = originalState.ftsAvailable; | |
| this.fts.loadError = originalState.ftsError; | |
| this.vector.available = originalDbClosed ? null : originalState.vectorAvailable; | |
| this.vector.loadError = originalState.vectorLoadError; | |
| this.vector.dims = originalState.vectorDims; | |
| this.vectorReady = originalDbClosed ? null : originalState.vectorReady; | |
| }; | |
| this.db = tempDb; | |
| this.vectorReady = null; | |
| this.vector.available = null; | |
| this.vector.loadError = undefined; | |
| this.vector.dims = undefined; | |
| this.fts.available = false; | |
| this.fts.loadError = undefined; | |
| this.ensureSchema(); | |
| let nextMeta: MemoryIndexMeta | null = null; | |
| try { | |
| this.seedEmbeddingCache(originalDb); | |
| const shouldSyncMemory = this.sources.has("memory"); | |
| const shouldSyncSessions = this.shouldSyncSessions( | |
| { reason: params.reason, force: params.force }, | |
| true, | |
| ); | |
| if (shouldSyncMemory) { | |
| await this.syncMemoryFiles({ needsFullReindex: true, progress: params.progress }); | |
| this.dirty = false; | |
| } | |
| if (shouldSyncSessions) { | |
| await this.syncSessionFiles({ needsFullReindex: true, progress: params.progress }); | |
| this.sessionsDirty = false; | |
| this.sessionsDirtyFiles.clear(); | |
| } else if (this.sessionsDirtyFiles.size > 0) { | |
| this.sessionsDirty = true; | |
| } else { | |
| this.sessionsDirty = false; | |
| } | |
| nextMeta = { | |
| model: this.provider.model, | |
| provider: this.provider.id, | |
| providerKey: this.providerKey, | |
| chunkTokens: this.settings.chunking.tokens, | |
| chunkOverlap: this.settings.chunking.overlap, | |
| }; | |
| if (this.vector.available && this.vector.dims) { | |
| nextMeta.vectorDims = this.vector.dims; | |
| } | |
| this.writeMeta(nextMeta); | |
| this.pruneEmbeddingCacheIfNeeded(); | |
| this.db.close(); | |
| originalDb.close(); | |
| originalDbClosed = true; | |
| await this.swapIndexFiles(dbPath, tempDbPath); | |
| this.db = this.openDatabaseAtPath(dbPath); | |
| this.vectorReady = null; | |
| this.vector.available = null; | |
| this.vector.loadError = undefined; | |
| this.ensureSchema(); | |
| this.vector.dims = nextMeta.vectorDims; | |
| } catch (err) { | |
| try { | |
| this.db.close(); | |
| } catch {} | |
| await this.removeIndexFiles(tempDbPath); | |
| restoreOriginalState(); | |
| throw err; | |
| } | |
| } | |
| private resetIndex() { | |
| this.db.exec(`DELETE FROM files`); | |
| this.db.exec(`DELETE FROM chunks`); | |
| if (this.fts.enabled && this.fts.available) { | |
| try { | |
| this.db.exec(`DELETE FROM ${FTS_TABLE}`); | |
| } catch {} | |
| } | |
| this.dropVectorTable(); | |
| this.vector.dims = undefined; | |
| this.sessionsDirtyFiles.clear(); | |
| } | |
| private readMeta(): MemoryIndexMeta | null { | |
| const row = this.db.prepare(`SELECT value FROM meta WHERE key = ?`).get(META_KEY) as | |
| | { value: string } | |
| | undefined; | |
| if (!row?.value) { | |
| return null; | |
| } | |
| try { | |
| return JSON.parse(row.value) as MemoryIndexMeta; | |
| } catch { | |
| return null; | |
| } | |
| } | |
| private writeMeta(meta: MemoryIndexMeta) { | |
| const value = JSON.stringify(meta); | |
| this.db | |
| .prepare( | |
| `INSERT INTO meta (key, value) VALUES (?, ?) ON CONFLICT(key) DO UPDATE SET value=excluded.value`, | |
| ) | |
| .run(META_KEY, value); | |
| } | |
| } | |
| export const memoryManagerSyncOps = MemoryManagerSyncOps.prototype; | |