catence 0.2.1-beta.15 → 0.2.1-beta.17

package/dist/contracts/release.d.ts

@@ -3,7 +3,7 @@
  * Keep these values synchronized with release/manifest.json; release checks
  * deliberately reject drift before a package can be published.
  */
-export declare const CATENCE_RUNTIME_VERSION = "0.2.1-beta.15";
+export declare const CATENCE_RUNTIME_VERSION = "0.2.1-beta.17";
 export declare const CATENCE_PROTOCOL_VERSION = 1;
 export type CatenceRuntimeCapabilities = {
     mcp: true;

package/dist/contracts/release.js

@@ -3,7 +3,7 @@
  * Keep these values synchronized with release/manifest.json; release checks
  * deliberately reject drift before a package can be published.
  */
-export const CATENCE_RUNTIME_VERSION = '0.2.1-beta.15';
+export const CATENCE_RUNTIME_VERSION = '0.2.1-beta.17';
 export const CATENCE_PROTOCOL_VERSION = 1;
 export function catenceRuntimeHealth() {
     return {

package/dist/core/retrieval/index.js

@@ -128,12 +128,30 @@ export async function buildRetrievalIndex(database) {
     await database.run('BEGIN TRANSACTION');
     try {
         await database.run('DELETE FROM retrieval_documents');
-        for (const document of deduplicated.values()) {
-            const id = documentId(document);
-            await database.run(`INSERT INTO retrieval_documents (document_id, entity_type, entity_id, activity_source_id, provider, occurred_on, document_text, metadata_json, content_hash)
-        VALUES ($id, $entityType, $entityId, $activitySourceId, $provider, $occurredOn, $text, $metadata, $hash)`, {
-                id, entityType: document.entityType, entityId: document.entityId, activitySourceId: document.activitySourceId ?? null, provider: document.provider ?? null, occurredOn: document.occurredOn || null, text: document.text, metadata: JSON.stringify(document.metadata), hash: id,
+        // Multi-row INSERT batches: one statement per 500 documents instead of
+        // one round trip per document keeps a full 125k-document rebuild from
+        // being dominated by per-statement overhead.
+        const BATCH_SIZE = 500;
+        const allDocuments = [...deduplicated.values()];
+        for (let offset = 0; offset < allDocuments.length; offset += BATCH_SIZE) {
+            const batch = allDocuments.slice(offset, offset + BATCH_SIZE);
+            const values = {};
+            const tuples = batch.map((document, index) => {
+                const id = documentId(document);
+                const prefix = `d${index}`;
+                values[`${prefix}id`] = id;
+                values[`${prefix}entityType`] = document.entityType;
+                values[`${prefix}entityId`] = document.entityId;
+                values[`${prefix}activitySourceId`] = document.activitySourceId ?? null;
+                values[`${prefix}provider`] = document.provider ?? null;
+                values[`${prefix}occurredOn`] = document.occurredOn || null;
+                values[`${prefix}text`] = document.text;
+                values[`${prefix}metadata`] = JSON.stringify(document.metadata);
+                values[`${prefix}hash`] = id;
+                return `($${prefix}id, $${prefix}entityType, $${prefix}entityId, $${prefix}activitySourceId, $${prefix}provider, $${prefix}occurredOn, $${prefix}text, $${prefix}metadata, $${prefix}hash)`;
             });
+            await database.run(`INSERT INTO retrieval_documents (document_id, entity_type, entity_id, activity_source_id, provider, occurred_on, document_text, metadata_json, content_hash)
+        VALUES ${tuples.join(', ')}`, values);
         }
         await database.run(`INSERT INTO retrieval_index_state (index_name, status, mode, source_watermark, built_at, detail_json)
       VALUES ('context', 'ready', $mode, $watermark, now(), $detail)

package/dist/elt/application/autosync.d.ts

@@ -0,0 +1,71 @@
+import { z } from 'zod';
+import { type CatalogPaths } from '../../core/runtime/catalog.js';
+import type { CatencePaths } from '../../contracts/runtime.js';
+import { startDetachedSync, type DetachedSyncHandle } from './detached-sync.js';
+/** The sync cadences the scheduler accepts, in minutes. */
+export declare const AUTO_SYNC_INTERVAL_MINUTES: readonly [15, 30, 60, 120, 360, 720, 1440];
+export type AutoSyncIntervalMinutes = (typeof AUTO_SYNC_INTERVAL_MINUTES)[number];
+/** Auto-sync is on by default at 2h: fresh data several times a day, far inside provider limits. */
+export declare const DEFAULT_AUTO_SYNC_INTERVAL_MINUTES: AutoSyncIntervalMinutes;
+declare const autoSyncFileSchema: z.ZodObject<{
+    enabled: z.ZodBoolean;
+    intervalMinutes: z.ZodNumber & z.ZodType<15 | 60 | 30 | 360 | 120 | 720 | 1440, number, z.core.$ZodTypeInternals<15 | 60 | 30 | 360 | 120 | 720 | 1440, number>>;
+    lastTriggeredAt: z.ZodDefault<z.ZodNullable<z.ZodString>>;
+    lastLogFile: z.ZodDefault<z.ZodNullable<z.ZodString>>;
+}, z.core.$strict>;
+export type AutoSyncConfig = z.infer<typeof autoSyncFileSchema>;
+export type AutoSyncSchedule = {
+    athleteId: string;
+    enabled: boolean;
+    intervalMinutes: AutoSyncIntervalMinutes;
+    intervalOptions: ReadonlyArray<{
+        minutes: AutoSyncIntervalMinutes;
+        label: string;
+    }>;
+    lastTriggeredAt: string | null;
+    lastLogFile: string | null;
+    nextRunAt: string | null;
+};
+export declare function autoSyncFilePath(paths: CatencePaths): string;
+/**
+ * Parse one of the interval labels the UI and CLI accept (`15m`, `2h`, `24h`)
+ * or a bare minute count into a valid cadence. Anything else is rejected at
+ * the boundary so stored state can never hold an unsupported interval.
+ */
+export declare function parseAutoSyncInterval(value: string): AutoSyncIntervalMinutes;
+/**
+ * Read the athlete's schedule. A missing file means "on by default": the
+ * effective schedule is reported without writing anything, so a fresh store
+ * stays pristine until the first explicit change (the scheduler does write
+ * trigger bookkeeping once it actually fires).
+ */
+export declare function readAutoSyncSchedule(paths: CatencePaths, athleteId: string): Promise<AutoSyncSchedule>;
+/**
+ * Apply an explicit schedule change (UI or CLI). Only the requested fields
+ * change; trigger bookkeeping survives an interval switch or a pause.
+ */
+export declare function updateAutoSyncSchedule(paths: CatencePaths, athleteId: string, change: {
+    enabled?: boolean;
+    intervalMinutes?: AutoSyncIntervalMinutes;
+}): Promise<AutoSyncSchedule>;
+export type AutoSyncSchedulerOptions = {
+    catalogPaths: CatalogPaths;
+    /** Internal compatibility hook for tests. */
+    startSync?: (request: Parameters<typeof startDetachedSync>[0]) => Promise<DetachedSyncHandle>;
+    now?: () => Date;
+    /** How often due schedules are checked; default every 30 seconds. */
+    tickMs?: number;
+};
+export type AutoSyncScheduler = {
+    stop(): void;
+};
+/**
+ * The auto-sync scheduler: a periodic timer that triggers a detached sync for
+ * every athlete whose schedule is due. It opens no database connection and
+ * takes no locks — the detached child and the single-writer DuckDB lock stay
+ * the only sync executors, so an already-running sync simply makes the tick a
+ * no-op (DetachedSyncBusyError is skipped quietly and retried on a later
+ * tick, once the active run has finished).
+ */
+export declare function startAutoSyncScheduler(options: AutoSyncSchedulerOptions): AutoSyncScheduler;
+export {};

package/dist/elt/application/autosync.js

@@ -0,0 +1,173 @@
+import { mkdir, readFile, rename, writeFile } from 'node:fs/promises';
+import path from 'node:path';
+import { z } from 'zod';
+import { athleteStorePaths, loadCatalog } from '../../core/runtime/catalog.js';
+import { createSyncLogger } from '../../core/logging.js';
+import { DetachedSyncBusyError, startDetachedSync } from './detached-sync.js';
+/** The sync cadences the scheduler accepts, in minutes. */
+export const AUTO_SYNC_INTERVAL_MINUTES = [15, 30, 60, 120, 360, 720, 1440];
+/** Auto-sync is on by default at 2h: fresh data several times a day, far inside provider limits. */
+export const DEFAULT_AUTO_SYNC_INTERVAL_MINUTES = 120;
+const INTERVAL_CHOICE_LABELS = new Map([
+    [15, '15m'], [30, '30m'], [60, '1h'], [120, '2h'], [360, '6h'], [720, '12h'], [1440, '24h'],
+]);
+const autoSyncFileSchema = z.object({
+    enabled: z.boolean(),
+    intervalMinutes: z.number().int().refine((value) => AUTO_SYNC_INTERVAL_MINUTES.includes(value), 'intervalMinutes must be one of 15, 30, 60, 120, 360, 720, 1440.'),
+    lastTriggeredAt: z.string().datetime().nullable().default(null),
+    lastLogFile: z.string().nullable().default(null),
+}).strict();
+export function autoSyncFilePath(paths) {
+    return path.join(paths.root, 'autosync.json');
+}
+/**
+ * Parse one of the interval labels the UI and CLI accept (`15m`, `2h`, `24h`)
+ * or a bare minute count into a valid cadence. Anything else is rejected at
+ * the boundary so stored state can never hold an unsupported interval.
+ */
+export function parseAutoSyncInterval(value) {
+    const trimmed = value.trim().toLowerCase();
+    const match = /^(\d+)(m|h)$/.exec(trimmed);
+    const minutes = match ? (match[2] === 'h' ? Number(match[1]) * 60 : Number(match[1])) : Number(trimmed);
+    if (Number.isInteger(minutes) && AUTO_SYNC_INTERVAL_MINUTES.includes(minutes)) {
+        return minutes;
+    }
+    throw new Error(`interval must be one of: ${[...INTERVAL_CHOICE_LABELS.values()].join(', ')}.`);
+}
+function defaultConfig() {
+    return { enabled: true, intervalMinutes: DEFAULT_AUTO_SYNC_INTERVAL_MINUTES, lastTriggeredAt: null, lastLogFile: null };
+}
+function nextRunAt(config) {
+    if (!config.enabled || !config.lastTriggeredAt)
+        return null;
+    return new Date(Date.parse(config.lastTriggeredAt) + config.intervalMinutes * 60_000).toISOString();
+}
+function toSchedule(athleteId, config) {
+    return {
+        athleteId,
+        enabled: config.enabled,
+        intervalMinutes: config.intervalMinutes,
+        intervalOptions: AUTO_SYNC_INTERVAL_MINUTES.map((minutes) => ({ minutes, label: INTERVAL_CHOICE_LABELS.get(minutes) ?? `${minutes}m` })),
+        lastTriggeredAt: config.lastTriggeredAt,
+        lastLogFile: config.lastLogFile,
+        nextRunAt: nextRunAt(config),
+    };
+}
+function isMissingFile(error) {
+    return typeof error === 'object' && error !== null && error.code === 'ENOENT';
+}
+async function readEffectiveConfig(paths) {
+    let raw;
+    try {
+        raw = await readFile(autoSyncFilePath(paths), 'utf8');
+    }
+    catch (error) {
+        if (!isMissingFile(error))
+            throw error;
+        return defaultConfig();
+    }
+    const parsed = autoSyncFileSchema.safeParse(JSON.parse(raw));
+    if (!parsed.success) {
+        throw new Error(`autosync.json is not a valid auto-sync schedule (${parsed.error.issues[0]?.message ?? 'unknown issue'}).`);
+    }
+    return parsed.data;
+}
+/**
+ * Read the athlete's schedule. A missing file means "on by default": the
+ * effective schedule is reported without writing anything, so a fresh store
+ * stays pristine until the first explicit change (the scheduler does write
+ * trigger bookkeeping once it actually fires).
+ */
+export async function readAutoSyncSchedule(paths, athleteId) {
+    return toSchedule(athleteId, await readEffectiveConfig(paths));
+}
+async function writeAutoSyncConfig(paths, config) {
+    const file = autoSyncFilePath(paths);
+    // A freshly added athlete has no store directory yet; the schedule file is
+    // often its first artifact.
+    await mkdir(path.dirname(file), { recursive: true });
+    const temporary = `${file}.${process.pid}.tmp`;
+    await writeFile(temporary, `${JSON.stringify(config, null, 2)}\n`, 'utf8');
+    await rename(temporary, file);
+}
+/**
+ * Apply an explicit schedule change (UI or CLI). Only the requested fields
+ * change; trigger bookkeeping survives an interval switch or a pause.
+ */
+export async function updateAutoSyncSchedule(paths, athleteId, change) {
+    const current = await readEffectiveConfig(paths);
+    const updated = {
+        ...current,
+        ...(change.enabled === undefined ? {} : { enabled: change.enabled }),
+        ...(change.intervalMinutes === undefined ? {} : { intervalMinutes: change.intervalMinutes }),
+    };
+    await writeAutoSyncConfig(paths, updated);
+    return toSchedule(athleteId, updated);
+}
+/** Record that the scheduler fired (or attempted to fire) for this athlete. */
+async function recordTrigger(paths, logFile, at) {
+    const config = await readEffectiveConfig(paths);
+    await writeAutoSyncConfig(paths, { ...config, lastTriggeredAt: at.toISOString(), lastLogFile: logFile });
+}
+/**
+ * The auto-sync scheduler: a periodic timer that triggers a detached sync for
+ * every athlete whose schedule is due. It opens no database connection and
+ * takes no locks — the detached child and the single-writer DuckDB lock stay
+ * the only sync executors, so an already-running sync simply makes the tick a
+ * no-op (DetachedSyncBusyError is skipped quietly and retried on a later
+ * tick, once the active run has finished).
+ */
+export function startAutoSyncScheduler(options) {
+    const startSync = options.startSync ?? startDetachedSync;
+    const now = options.now ?? (() => new Date());
+    const tickMs = options.tickMs ?? 30_000;
+    let ticking = false;
+    const tick = async () => {
+        if (ticking)
+            return; // Never overlap ticks; the in-flight one wins.
+        ticking = true;
+        try {
+            const catalog = await loadCatalog(options.catalogPaths);
+            await Promise.allSettled(catalog.athletes.map(async (athlete) => {
+                const paths = athleteStorePaths(options.catalogPaths, athlete.id);
+                const log = createSyncLogger(paths);
+                try {
+                    const config = await readEffectiveConfig(paths);
+                    if (!config.enabled)
+                        return;
+                    const dueAt = config.lastTriggeredAt ? Date.parse(config.lastTriggeredAt) + config.intervalMinutes * 60_000 : 0;
+                    if (now().getTime() < dueAt)
+                        return;
+                    const handle = await startSync({ paths, athleteId: athlete.id, provider: 'all', cliHome: options.catalogPaths.root });
+                    await recordTrigger(paths, handle.logFile, now());
+                    log.info('Auto-sync triggered', { athleteId: athlete.id, intervalMinutes: config.intervalMinutes, logFile: handle.logFile });
+                }
+                catch (error) {
+                    if (error instanceof DetachedSyncBusyError)
+                        return; // A run is active; retry on a later tick.
+                    // Any other failure is recorded as a trigger attempt so a broken
+                    // store cannot be hammered every tick; the next cadence retries.
+                    await recordTrigger(paths, null, now()).catch(() => undefined);
+                    log.error('Auto-sync trigger failed', { athleteId: athlete.id, error: error instanceof Error ? error.message : String(error) });
+                }
+            }));
+        }
+        catch (error) {
+            process.stderr.write(`[catence] Auto-sync tick failed: ${error instanceof Error ? error.message : String(error)}\n`);
+        }
+        finally {
+            ticking = false;
+        }
+    };
+    const timer = setInterval(() => { void tick(); }, tickMs);
+    timer.unref?.();
+    // The first tick is immediate: a restart should heal stale data within a
+    // minute instead of waiting a full interval.
+    void tick();
+    return {
+        stop() {
+            clearInterval(timer);
+        },
+    };
+}
+//# sourceMappingURL=autosync.js.map

package/dist/elt/application/management.js

@@ -114,9 +114,16 @@ async function backfillGarminSwimLengths(database, paths, parentRunId, log) {
     }
     // The swim list is queried here and handed to the Python worker as a file:
     // the child must never open DuckDB itself because the parent holds the
-    // write lock for the whole run.
+    // write lock for the whole run. Only activities without per-length rows are
+    // handed over: re-parsing every archived FIT on every sync cost ~23s/run
+    // with nothing new to stage.
     const activitiesFile = path.join(paths.staging, 'garmin', `${parentRunId}.swim_activities.json`);
-    await writeFile(activitiesFile, JSON.stringify(await database.garminSwimActivitySources()));
+    const pendingSwims = await database.garminSwimActivitySourcesWithoutLengths();
+    if (pendingSwims.length === 0) {
+        log.info('Garmin swim backfill skipped: every swim activity already has lengths', { parentRunId });
+        return;
+    }
+    await writeFile(activitiesFile, JSON.stringify(pendingSwims));
     const arguments_ = [
         'run',
         'python',

package/dist/elt/ingestion/importer.d.ts

@@ -1,7 +1,17 @@
 import type { CatenceDatabase } from '../storage/database.js';
 import { type StagingRecord } from '../../contracts/staging.js';
 import type { SyncLogger } from '../../core/logging.js';
-export declare function importRecord(database: CatenceDatabase, runId: string, record: StagingRecord, log?: SyncLogger): Promise<void>;
+/**
+ * Per-pass import state. When present, activity-quality refreshes are
+ * deferred to the end of the pass: a staging file lists an activity before
+ * its detail records, so refreshing inline would compute flags from partial
+ * data, and one refresh per activity is enough regardless of how many
+ * detail entity types follow it.
+ */
+export type ImportPassContext = {
+    deferredQualityActivityIds: Set<string>;
+};
+export declare function importRecord(database: CatenceDatabase, runId: string, record: StagingRecord, log?: SyncLogger, pass?: ImportPassContext): Promise<void>;
 export declare function importJsonl(database: CatenceDatabase, runId: string, filePath: string, log?: SyncLogger): Promise<number>;
 export declare function importSingleStagingFile(database: CatenceDatabase, runId: string, filePath: string): Promise<number>;
 export declare function queueWorkItem(database: CatenceDatabase, runId: string, provider: 'intervals' | 'garmin' | 'strava', endpoint: string, remoteId: string | null, scope: unknown): Promise<string>;

package/dist/elt/ingestion/importer.js

@@ -7,7 +7,12 @@ import { importSourceEntity, refreshActivityQuality } from '../normalization/nor
 import { normalizeStravaEntity } from '../normalization/segments/strava-normalizer.js';
 import { reconcileActivityLink } from '../normalization/activities/linking.js';
 import { json } from '../storage/sql.js';
-export async function importRecord(database, runId, record, log) {
+const GARMIN_ACTIVITY_DETAIL_ENTITY_TYPES = ['activity_detail', 'activity_interval', 'activity_exercise_set', 'activity_swim_lengths'];
+async function activityIdForSource(database, activitySourceId) {
+    const activity = await database.rows('SELECT activity_id FROM activity_sources WHERE activity_source_id = $activitySourceId', { activitySourceId });
+    return activity[0]?.activity_id ?? null;
+}
+export async function importRecord(database, runId, record, log, pass) {
     if (record.kind === 'run_manifest') {
         if (record.runId !== runId)
             throw new Error(`Staging manifest run ${record.runId} does not match importer run ${runId}.`);
@@ -29,14 +34,14 @@ export async function importRecord(database, runId, record, log) {
             await importSourceEntity(database, record);
         if (record.entityType === 'activity') {
             await reconcileActivityLink(database, record.provider + ':' + record.remoteId);
-            const activity = await database.rows('SELECT activity_id FROM activity_sources WHERE activity_source_id = $activitySourceId', { activitySourceId: record.provider + ':' + record.remoteId });
-            if (activity[0]?.activity_id)
-                await refreshActivityQuality(database, activity[0].activity_id);
+            const activityId = await activityIdForSource(database, record.provider + ':' + record.remoteId);
+            if (activityId)
+                await queueActivityQualityRefresh(database, activityId, pass);
         }
-        if (record.provider === 'garmin' && ['activity_detail', 'activity_interval', 'activity_exercise_set', 'activity_swim_lengths'].includes(record.entityType) && record.parentRemoteId) {
-            const activity = await database.rows('SELECT activity_id FROM activity_sources WHERE activity_source_id = $activitySourceId', { activitySourceId: `garmin:${record.parentRemoteId}` });
-            if (activity[0]?.activity_id)
-                await refreshActivityQuality(database, activity[0].activity_id);
+        if (record.provider === 'garmin' && GARMIN_ACTIVITY_DETAIL_ENTITY_TYPES.includes(record.entityType) && record.parentRemoteId) {
+            const activityId = await activityIdForSource(database, `garmin:${record.parentRemoteId}`);
+            if (activityId)
+                await queueActivityQualityRefresh(database, activityId, pass);
         }
         return;
     }
@@ -56,17 +61,61 @@ export async function importRecord(database, runId, record, log) {
     await database.addError(runId, record.provider, record.endpoint, record.remoteId, record.message, record.retryable);
     log?.warn('Staging record reported an extraction error', { runId, provider: record.provider, endpoint: record.endpoint, remoteId: record.remoteId, error: record.message });
 }
+/**
+ * Refresh now (single-record callers) or defer to the end of the pass
+ * (bulk importJsonl callers) — see ImportPassContext.
+ */
+async function queueActivityQualityRefresh(database, activityId, pass) {
+    if (pass) {
+        pass.deferredQualityActivityIds.add(activityId);
+        return;
+    }
+    await refreshActivityQuality(database, activityId);
+}
+/** Records per transaction: large enough to amortize DuckDB commits, small enough to bound a mid-file failure. */
+const IMPORT_CHUNK_SIZE = 200;
 export async function importJsonl(database, runId, filePath, log) {
     let count = 0;
+    const pass = { deferredQualityActivityIds: new Set() };
     const file = createReadStream(filePath, { encoding: 'utf8' });
     const lines = createInterface({ input: file, crlfDelay: Infinity });
-    for await (const line of lines) {
-        if (!line.trim())
-            continue;
-        const record = stagingRecordSchema.parse(JSON.parse(line));
-        await importRecord(database, runId, record, log);
-        count += 1;
+    let inTransaction = false;
+    const startedAt = Date.now();
+    try {
+        for await (const line of lines) {
+            if (!line.trim())
+                continue;
+            if (count % IMPORT_CHUNK_SIZE === 0) {
+                await database.run('BEGIN TRANSACTION');
+                inTransaction = true;
+            }
+            const record = stagingRecordSchema.parse(JSON.parse(line));
+            await importRecord(database, runId, record, log, pass);
+            count += 1;
+            if (count % IMPORT_CHUNK_SIZE === 0) {
+                await database.run('COMMIT');
+                inTransaction = false;
+            }
+        }
+    }
+    catch (error) {
+        if (inTransaction)
+            await database.run('ROLLBACK');
+        throw error;
+    }
+    if (inTransaction) {
+        await database.run('COMMIT');
+        inTransaction = false;
+    }
+    const qualityStartedAt = Date.now();
+    for (const activityId of pass.deferredQualityActivityIds) {
+        await refreshActivityQuality(database, activityId);
     }
+    log?.info('Staging import finished with stage timings', {
+        runId, records: count, durationMs: Date.now() - startedAt,
+        qualityRefreshes: pass.deferredQualityActivityIds.size,
+        qualityRefreshMs: Date.now() - qualityStartedAt,
+    });
     return count;
 }
 export async function importSingleStagingFile(database, runId, filePath) {

package/dist/elt/ingestion/providers/intervals.d.ts

@@ -27,6 +27,13 @@ export declare class IntervalsExtractor {
     private captureActivity;
     private capture;
     private importEntities;
+    /**
+     * Intervals.icu enforces 2500 requests per rolling 15 minutes and 5000 per
+     * day; over-limit calls answer 429 with a Retry-After header. Back off
+     * cooperatively and retry a bounded number of times so one auto-sync run
+     * degrades instead of recording a wall of retryable errors.
+     */
+    private request;
     private getJson;
     private getBinary;
 }

package/dist/elt/ingestion/providers/intervals.js

@@ -243,14 +243,33 @@ export class IntervalsExtractor {
             }
         }
     }
+    /**
+     * Intervals.icu enforces 2500 requests per rolling 15 minutes and 5000 per
+     * day; over-limit calls answer 429 with a Retry-After header. Back off
+     * cooperatively and retry a bounded number of times so one auto-sync run
+     * degrades instead of recording a wall of retryable errors.
+     */
+    async request(route, accept) {
+        for (let attempt = 0;; attempt++) {
+            const headers = { Authorization: basicAuth(this.apiKey) };
+            if (accept)
+                headers.Accept = accept;
+            const response = await fetch(`${apiBaseUrl}${route}`, { headers });
+            if (response.status !== 429 || attempt >= 2)
+                return response;
+            const retryAfterSeconds = Math.min(Math.max(Number(response.headers.get('retry-after')) || 30, 1), 300);
+            this.log.warn('Intervals rate limit hit; backing off', { route, retryAfterSeconds, attempt: attempt + 1 });
+            await new Promise((resolve) => setTimeout(resolve, retryAfterSeconds * 1_000));
+        }
+    }
     async getJson(route) {
-        const response = await fetch(`${apiBaseUrl}${route}`, { headers: { Authorization: basicAuth(this.apiKey), Accept: 'application/json' } });
+        const response = await this.request(route, 'application/json');
         if (!response.ok)
             throw new Error(`Intervals ${route} failed with HTTP ${response.status}`);
         return response.json();
     }
     async getBinary(route) {
-        const response = await fetch(`${apiBaseUrl}${route}`, { headers: { Authorization: basicAuth(this.apiKey) } });
+        const response = await this.request(route, null);
         if (!response.ok)
             throw new Error(`Intervals ${route} failed with HTTP ${response.status}`);
         return new Uint8Array(await response.arrayBuffer());

package/dist/elt/storage/database.d.ts

@@ -60,7 +60,12 @@ export declare class CatenceDatabase implements WriteDataStore {
     knownActivitySummaryHashes(provider: Provider): Promise<Record<string, string>>;
     hasGarminLactateThresholdHistory(): Promise<boolean>;
     earliestActivityStartDate(provider: Provider): Promise<string | null>;
-    garminSwimActivitySources(): Promise<Array<{
+    /**
+     * Swim activities that still lack per-length rows: the FIT backfill only
+     * has work to do for these, so syncs skip re-parsing already-derived
+     * archives entirely.
+     */
+    garminSwimActivitySourcesWithoutLengths(): Promise<Array<{
         remote_activity_id: string;
         started_on: string | null;
     }>>;

package/dist/elt/storage/database.js

@@ -319,13 +319,19 @@ export class CatenceDatabase {
          JOIN activity_sources source USING (activity_id) WHERE source.provider = $provider`, { provider });
         return toIsoDate(rows[0]?.earliest ?? null);
     }
-    async garminSwimActivitySources() {
+    /**
+     * Swim activities that still lack per-length rows: the FIT backfill only
+     * has work to do for these, so syncs skip re-parsing already-derived
+     * archives entirely.
+     */
+    async garminSwimActivitySourcesWithoutLengths() {
         const sports = ['lap_swimming', 'swimming', 'open_water_swimming', 'openwaterswimming'];
         return this.rows(`SELECT source.remote_activity_id, cast(activity.started_at_utc AS VARCHAR) AS started_on
          FROM activity_sources AS source
          JOIN activities AS activity USING (activity_id)
         WHERE source.provider = 'garmin'
           AND lower(coalesce(activity.sport, '')) IN (${sports.map((sport) => `'${sport}'`).join(', ')})
+          AND NOT EXISTS (SELECT 1 FROM swim_lengths WHERE swim_lengths.activity_source_id = source.activity_source_id)
         ORDER BY activity.started_at_utc`);
     }
     async latestSourceDate(provider, cursorName) {

package/dist/interfaces/cli/commands.js

@@ -3,7 +3,7 @@ import { createInterface } from 'node:readline/promises';
 import { stdin as input, stdout as output } from 'node:process';
 import { Command } from 'commander';
 import { defaultBearerTokenPath, disableBearerToken, ensureBearerToken, mcpExposureStatus, readBearerToken, } from '../http/cli-options.js';
-import { addAthlete, athleteStorePaths, CATENCE_RUNTIME_VERSION, CatenceDatabase, channelForVersion, createDemoStore, dataStatus, deleteAthleteSecret, defaultCatalogHome, disconnectStravaAccount, fetchNpmDistTags, fetchPypiReleases, hasCommand, importStagedFile, initializeCatalog, isGlobalNpmInstall, linkActivity, loadCatalog, npmGlobalPrefix, planSelfUpdate, readInstalledConsoleVersion, rebuildRetrievalIndex, reimportNutrition, reimportWeight, resolveExtractionErrors, resolveAthlete, resolveCatalogPaths, retryDataSync, runSelfUpdateCommand, setAthleteSecret, syncData, syncProgress, connectStrava, connectStravaWithCallback, unlinkActivity, } from '../../runtime/index.js';
+import { addAthlete, athleteStorePaths, CATENCE_RUNTIME_VERSION, CatenceDatabase, channelForVersion, createDemoStore, dataStatus, deleteAthleteSecret, defaultCatalogHome, disconnectStravaAccount, fetchNpmDistTags, fetchPypiReleases, hasCommand, importStagedFile, initializeCatalog, isGlobalNpmInstall, linkActivity, loadCatalog, npmGlobalPrefix, parseAutoSyncInterval, planSelfUpdate, readAutoSyncSchedule, readInstalledConsoleVersion, rebuildRetrievalIndex, reimportNutrition, reimportWeight, resolveExtractionErrors, resolveAthlete, resolveCatalogPaths, retryDataSync, runSelfUpdateCommand, setAthleteSecret, syncData, syncProgress, connectStrava, connectStravaWithCallback, unlinkActivity, updateAutoSyncSchedule, } from '../../runtime/index.js';
 const program = new Command()
     .name('catence-data')
     .description('Athlete-scoped Catence data setup, extraction, and normalization.')
@@ -166,6 +166,27 @@ program.command('retry')
     const { paths } = await currentAthlete();
     process.stdout.write(`${JSON.stringify(await retryDataSync(paths, options.run), null, 2)}\n`);
 });
+const autoSyncCommand = program.command('auto-sync').description('Manage the automatic sync schedule for one athlete. The scheduler runs inside the Catence server (default: enabled, every 2 hours).');
+autoSyncCommand.command('status')
+    .description('Report the athlete’s automatic sync schedule: enabled flag, interval, and next run.')
+    .action(async () => {
+    const { athlete, paths } = await currentAthlete();
+    process.stdout.write(`${JSON.stringify(await readAutoSyncSchedule(paths, athlete.id), null, 2)}\n`);
+});
+autoSyncCommand.command('enable')
+    .description('Turn automatic syncing on, optionally changing the interval (15m, 30m, 1h, 2h, 6h, 12h, 24h).')
+    .option('--interval <interval>', 'sync interval label or minutes', '2h')
+    .action(async (options) => {
+    const { athlete, paths } = await currentAthlete();
+    const intervalMinutes = parseAutoSyncInterval(options.interval);
+    process.stdout.write(`${JSON.stringify(await updateAutoSyncSchedule(paths, athlete.id, { enabled: true, intervalMinutes }), null, 2)}\n`);
+});
+autoSyncCommand.command('disable')
+    .description('Turn automatic syncing off. Manual sync keeps working.')
+    .action(async () => {
+    const { athlete, paths } = await currentAthlete();
+    process.stdout.write(`${JSON.stringify(await updateAutoSyncSchedule(paths, athlete.id, { enabled: false }), null, 2)}\n`);
+});
 program.command('auth').description('Connect a source account using one athlete’s local credentials.')
     .command('strava')
     .option('--code <authorization-code>', 'OAuth authorization code returned by Strava')

package/dist/interfaces/http/server.js

@@ -2,7 +2,7 @@ import { timingSafeEqual } from 'node:crypto';
 import { createServer } from 'node:http';
 import path from 'node:path';
 import { StreamableHTTPServerTransport } from '@modelcontextprotocol/sdk/server/streamableHttp.js';
-import { addAthlete, AthleteFileConflictError, AthleteFileTooLargeError, AthleteFileValidationError, catenceRuntimeHealth, DashboardSnapshotService, deleteAthleteSecret, describeAthleteSecrets, DetachedSyncBusyError, DETACHED_SYNC_PROVIDERS, jsonSafe, listAthleteFileRevisions, loadCatalog, mergeOpenCodeGoConsoleProfiles, openReadOnlyRepository, readAthleteFile, readAthleteFileRevision, resolveAthlete, resolveCatalogPaths, setAthleteSecret, startDetachedSync, syncProgress, updateAthleteFile, } from '../../runtime/index.js';
+import { addAthlete, AthleteFileConflictError, AthleteFileTooLargeError, AthleteFileValidationError, catenceRuntimeHealth, DashboardSnapshotService, deleteAthleteSecret, describeAthleteSecrets, DetachedSyncBusyError, DETACHED_SYNC_PROVIDERS, jsonSafe, listAthleteFileRevisions, loadCatalog, mergeOpenCodeGoConsoleProfiles, openReadOnlyRepository, parseAutoSyncInterval, readAutoSyncSchedule, readAthleteFile, readAthleteFileRevision, resolveAthlete, resolveCatalogPaths, setAthleteSecret, startDetachedSync, syncProgress, updateAthleteFile, updateAutoSyncSchedule, } from '../../runtime/index.js';
 import { activeBearerToken, ensureBearerToken, disableBearerToken, mcpExposureStatus } from './cli-options.js';
 import { createCatenceMcpServer } from '../mcp/server.js';
 import { generationStatus } from './generation-status.js';
@@ -103,6 +103,34 @@ async function dashboardPaths(catalogPaths, staticPaths, url) {
         return staticPaths;
     throw new Error('No Catence data store is configured.');
 }
+function parseSyncScheduleBody(body) {
+    if (body === undefined)
+        return {};
+    if (typeof body !== 'object' || body === null || Array.isArray(body))
+        throw new Error('Request body must be a JSON object.');
+    const raw = body;
+    for (const key of Object.keys(raw)) {
+        if (!['athleteId', 'enabled', 'intervalMinutes'].includes(key)) {
+            throw new Error(`Unknown auto-sync field: ${key}.`);
+        }
+    }
+    if (raw.athleteId !== undefined && typeof raw.athleteId !== 'string')
+        throw new Error('athleteId must be a string.');
+    if (raw.enabled !== undefined && typeof raw.enabled !== 'boolean')
+        throw new Error('enabled must be a boolean.');
+    if (raw.intervalMinutes !== undefined) {
+        if (typeof raw.intervalMinutes === 'string') {
+            raw.intervalMinutes = parseAutoSyncInterval(raw.intervalMinutes);
+        }
+        else if (typeof raw.intervalMinutes !== 'number') {
+            throw new Error('intervalMinutes must be a number or a label like 15m, 2h, 24h.');
+        }
+    }
+    if (raw.enabled === undefined && raw.intervalMinutes === undefined) {
+        throw new Error('Provide enabled, intervalMinutes, or both.');
+    }
+    return raw;
+}
 function parseSyncBody(body) {
     if (body === undefined)
         return {};
@@ -548,6 +576,34 @@ export function createCatenceHttpServer(options = {}) {
                 }
                 return;
             }
+            if (pathname === '/api/v1/sync/schedule' && request.method === 'GET') {
+                try {
+                    const url = new URL(request.url ?? '/', 'http://localhost');
+                    const paths = await dashboardPaths(catalogPaths, staticPaths, url);
+                    const athleteId = catalogPaths ? (url.searchParams.get('athleteId') ?? '') : 'local';
+                    json(response, 200, await readAutoSyncSchedule(paths, athleteId));
+                }
+                catch (error) {
+                    json(response, 400, { error: { code: 'schedule_unavailable', message: error instanceof Error ? error.message : String(error) } });
+                }
+                return;
+            }
+            if (pathname === '/api/v1/sync/schedule' && request.method === 'PUT') {
+                try {
+                    const body = parseSyncScheduleBody(await readJsonBody(request));
+                    const url = new URL(request.url ?? '/', 'http://localhost');
+                    const paths = await resolveSyncPaths(catalogPaths, staticPaths, { athleteId: body.athleteId });
+                    const athleteId = catalogPaths ? (body.athleteId ?? '') : 'local';
+                    json(response, 200, await updateAutoSyncSchedule(paths, athleteId, {
+                        enabled: body.enabled,
+                        ...(body.intervalMinutes === undefined ? {} : { intervalMinutes: parseAutoSyncInterval(String(body.intervalMinutes)) }),
+                    }));
+                }
+                catch (error) {
+                    json(response, 400, { error: { code: 'schedule_update_failed', message: error instanceof Error ? error.message : String(error) } });
+                }
+                return;
+            }
             if (pathname === '/api/v1/models/discover' && request.method === 'POST') {
                 // Refresh the OpenCode Go profiles without starting a data sync. The
                 // Console's Models page calls this behind its authenticated proxy.

package/dist/interfaces/mcp/main.js

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
 import { existsSync } from 'node:fs';
-import { addAthlete, athleteStorePaths, CatenceDatabase, createDemoStore, defaultCatalogHome, initializeCatalog, loadCatalog, resolveCatalogPaths } from '../../runtime/index.js';
+import { addAthlete, athleteStorePaths, CatenceDatabase, createDemoStore, defaultCatalogHome, initializeCatalog, loadCatalog, resolveCatalogPaths, startAutoSyncScheduler } from '../../runtime/index.js';
 import { defaultBearerTokenPath, parseServeCliOptions, SERVE_USAGE } from '../http/cli-options.js';
 import { createCatenceHttpServer } from '../http/server.js';
 import { MCP_USAGE, parseMcpCliOptions } from './cli-options.js';
@@ -63,7 +63,14 @@ async function run() {
         });
         console.error(`Catence Streamable HTTP server is listening at http://${options.host}:${options.port}/mcp`);
         console.error(`Bearer-token authentication is armed at ${options.authTokenFile ?? defaultBearerTokenPath(catalogPaths.root)}; enable it with 'catence-data mcp enable'. Health endpoints stay open for readiness probes.`);
+        // Auto-sync runs while the server is up: each athlete's autosync.json
+        // (default: enabled, every 2 hours) triggers a detached sync. The
+        // detached child and the single-writer DuckDB lock stay the only
+        // executors, so concurrent ticks are no-ops by design.
+        const scheduler = startAutoSyncScheduler({ catalogPaths });
+        console.error('Auto-sync scheduler started (default: enabled, every 2 hours per athlete; configure with auto-sync enable --interval or Console Settings).');
         const close = () => {
+            scheduler.stop();
             httpServer.close(() => process.exit(0));
         };
         process.once('SIGINT', close);

package/dist/interfaces/mcp/server.js

@@ -312,7 +312,17 @@ export function createCatenceMcpServer(paths = resolveCatalogPaths(), dependenci
     }, tool('start_detached_sync', async (input) => {
         const paths = activePaths();
         const athleteId = activeAthlete()?.id ?? 'local';
-        const handle = await startSync({ paths, athleteId, provider: input.provider, from: input.from, refresh: input.refresh });
+        const handle = await startSync({
+            paths,
+            athleteId,
+            provider: input.provider,
+            from: input.from,
+            refresh: input.refresh,
+            // The child resolves the athlete within the catalog, not within the
+            // athlete store this tool resolved for locks and logs: passing the
+            // store itself trips the legacy-store guard in catalog deployments.
+            ...(catalogPaths ? { cliHome: catalogPaths.root } : {}),
+        });
         return {
             data: { ...handle, progressTool: 'catence_sync_progress' },
             provenance: { operation: 'detached_sync_spawn' },

package/dist/runtime/index.d.ts

@@ -35,6 +35,8 @@ export { connectStrava, connectStravaWithCallback, dataStatus, disconnectStravaA
 export type { ProviderChoice } from '../elt/application/management.js';
 export { DETACHED_SYNC_PROVIDERS, DetachedSyncBusyError, startDetachedSync, } from '../elt/application/detached-sync.js';
 export type { DetachedSyncHandle, DetachedSyncProvider, DetachedSyncRequest, DetachedSyncSpawner } from '../elt/application/detached-sync.js';
+export { AUTO_SYNC_INTERVAL_MINUTES, DEFAULT_AUTO_SYNC_INTERVAL_MINUTES, parseAutoSyncInterval, readAutoSyncSchedule, startAutoSyncScheduler, updateAutoSyncSchedule, } from '../elt/application/autosync.js';
+export type { AutoSyncConfig, AutoSyncIntervalMinutes, AutoSyncSchedule, AutoSyncScheduler, AutoSyncSchedulerOptions } from '../elt/application/autosync.js';
 export { buildOpenCodeGoConsoleProfiles, classifyOpenCodeGoModel, fetchOpenCodeGoModelIds, mergeOpenCodeGoConsoleProfiles, OPENCODE_GO_API_BASE_ENV, OPENCODE_GO_API_KEY_ENV, OPENCODE_GO_DEFAULT_BASE_URL, OPENCODE_GO_MESSAGES_API_BASE_ENV, } from '../core/runtime/opencode-go.js';
 export type { OpenCodeGoConsoleProfiles, OpenCodeGoMergeResult, OpenCodeGoRoute } from '../core/runtime/opencode-go.js';
 export type { InterruptedRuns, SyncProgressSnapshot, SyncProgressState } from '../contracts/progress.js';

package/dist/runtime/index.js

@@ -19,6 +19,7 @@ export { queryReadOnlyData } from '../core/query/sql-guard.js';
 export { createDemoStore, demoStoreMetadata } from '../elt/application/demo.js';
 export { connectStrava, connectStravaWithCallback, dataStatus, disconnectStravaAccount, initializeDataStore, importStagedFile, linkActivity, rebuildRetrievalIndex, reimportNutrition, reimportWeight, resolveExtractionErrors, retryDataSync, syncData, syncProgress, unlinkActivity, } from '../elt/application/management.js';
 export { DETACHED_SYNC_PROVIDERS, DetachedSyncBusyError, startDetachedSync, } from '../elt/application/detached-sync.js';
+export { AUTO_SYNC_INTERVAL_MINUTES, DEFAULT_AUTO_SYNC_INTERVAL_MINUTES, parseAutoSyncInterval, readAutoSyncSchedule, startAutoSyncScheduler, updateAutoSyncSchedule, } from '../elt/application/autosync.js';
 export { buildOpenCodeGoConsoleProfiles, classifyOpenCodeGoModel, fetchOpenCodeGoModelIds, mergeOpenCodeGoConsoleProfiles, OPENCODE_GO_API_BASE_ENV, OPENCODE_GO_API_KEY_ENV, OPENCODE_GO_DEFAULT_BASE_URL, OPENCODE_GO_MESSAGES_API_BASE_ENV, } from '../core/runtime/opencode-go.js';
 export { CatenceDatabase, openReadOnlyRepository, ReadOnlyDatabaseError } from '../elt/storage/database.js';
 export { DataWriteBusyError } from '../elt/storage/write-lock.js';

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "catence",
-  "version": "0.2.1-beta.15",
+  "version": "0.2.1-beta.17",
   "description": "Local MCP access to normalized Garmin, Intervals.icu, and targeted Strava fitness data.",
   "license": "AGPL-3.0-or-later",
   "repository": {

package/python/catence/providers/garmin/cli.py

@@ -75,12 +75,32 @@ def main() -> None:
     email = os.environ.get("GARMIN_EMAIL")
     if not email:
         raise SystemExit("GARMIN_EMAIL is required.")
-    password = os.environ.get("GARMIN_PASSWORD") or getpass.getpass("Garmin password: ")
-    writer = StagingWriter(args.data_dir, args.output)
-    writer.manifest(args.run_id, args.from_date.isoformat())
     token_store = args.data_dir / "auth" / "garmin_tokens.json"
     token_store.parent.mkdir(parents=True, exist_ok=True)
-    client = Garmin(email, password, prompt_mfa=lambda: input("Garmin MFA code: "))
+    if not sys.stdin.isatty():
+        # Unattended auto-sync must never hang on a hidden prompt: fail fast
+        # so the run records a visible error instead.
+        password = os.environ.get("GARMIN_PASSWORD")
+        if not password and not token_store.exists():
+            raise SystemExit(
+                "Garmin login is required (no stored token, no GARMIN_PASSWORD, "
+                "and stdin is not interactive). Run an interactive sync once to "
+                "store the session token, or set GARMIN_PASSWORD."
+            )
+    else:
+        password = os.environ.get("GARMIN_PASSWORD") or getpass.getpass("Garmin password: ")
+
+    def prompt_mfa() -> str:
+        if not sys.stdin.isatty():
+            raise SystemExit(
+                "Garmin MFA is required but stdin is not interactive; the "
+                "session token expired. Run an interactive sync once to renew it."
+            )
+        return input("Garmin MFA code: ")
+
+    writer = StagingWriter(args.data_dir, args.output)
+    writer.manifest(args.run_id, args.from_date.isoformat())
+    client = Garmin(email, password, prompt_mfa=prompt_mfa)
     client.login(str(token_store))
     reporter = ProgressReporter(args.run_id)
     reporter.set_stage("login")