fix(wiki-worker): read current session from a local cache, not a full DB re-scan

src/hooks/capture.ts

@@ -24,6 +24,7 @@ import {
   ensureSessionOwner,
 } from "./summary-state.js";
 import { bundleDirFromImportMeta, spawnWikiWorker, wikiLog } from "./spawn-wiki-worker.js";
+import { appendSessionEvent } from "./session-event-cache.js";
 import { tryStopCounterTrigger } from "../skillify/triggers.js";
 import { reactSkillOpt } from "./shared/skillopt-hook.js";
 import { EmbedClient } from "../embeddings/client.js";
@@ -184,6 +185,13 @@ async function main(): Promise<void> {
 
   log("capture ok → cloud");
 
+  // Mirror the event into the local per-session cache (row-for-row identical
+  // to the `message` column just INSERTed). The wiki-worker reads this instead
+  // of re-scanning the entire fat `message` column for the current session on
+  // every periodic / session-end summary trigger. Best-effort; DB stays the
+  // source of truth. Only reached after a successful INSERT above.
+  appendSessionEvent(input.session_id, line);
+
   // Commit-driven KPI auto-extract is disabled for now — the
   // fire-and-forget sub-agent spawned per `git commit` (see
   // src/hooks/commit-kpi-extract.ts) consumed a high amount of tokens

src/hooks/cursor/capture.ts

@@ -34,6 +34,7 @@ import {
   releaseLock,
 } from "../summary-state.js";
 import { bundleDirFromImportMeta, spawnCursorWikiWorker, wikiLog } from "./spawn-wiki-worker.js";
+import { appendSessionEvent } from "../session-event-cache.js";
 import { tryStopCounterTrigger } from "../../skillify/triggers.js";
 import type { Config } from "../../config.js";
 import { getInstalledVersion } from "../../utils/version-check.js";
@@ -180,6 +181,12 @@ async function main(): Promise<void> {
 
   log("capture ok → cloud");
 
+  // Mirror the event into the local per-session cache (row-for-row identical
+  // to the `message` column just INSERTed) so the wiki-worker reads it instead
+  // of re-scanning the fat `message` column. Best-effort; only after a
+  // successful INSERT above.
+  appendSessionEvent(sessionId, line);
+
   maybeTriggerPeriodicSummary(sessionId, cwd, config);
 
   // Skillify Stop counter — afterAgentResponse is the assistant-complete event.

src/hooks/cursor/spawn-wiki-worker.ts

@@ -104,6 +104,7 @@ export function spawnCursorWikiWorker(opts: SpawnOptions): void {
     sessionsTable: config.sessionsTableName,
     sessionId,
     userName: config.userName,
+    orgName: config.orgName,
     project: projectName,
     pluginVersion,
     tmpDir,

src/hooks/cursor/wiki-worker.ts

@@ -18,6 +18,8 @@ import { buildTrailingPromptInvocation } from "../wiki-worker-spawn.js";
 import { dirname, join } from "node:path";
 import { fileURLToPath } from "node:url";
 import { finalizeSummary, releaseLock, readState } from "../summary-state.js";
+import { readSessionEventCache } from "../session-event-cache.js";
+import { buildSessionPath } from "../../utils/session-path.js";
 import { capLinesByBytes, stampOffset, WIKI_JSONL_MAX_BYTES } from "../wiki-offset.js";
 import { uploadSummary } from "../upload-summary.js";
 import { log as _log } from "../../utils/debug.js";
@@ -36,6 +38,7 @@ interface WorkerConfig {
   sessionsTable: string;
   sessionId: string;
   userName: string;
+  orgName: string;
   project: string;
   pluginVersion?: string;
   tmpDir: string;
@@ -114,27 +117,55 @@ function cleanup(): void {
 
 async function main(): Promise<void> {
   try {
-    // 1. Fetch session events from sessions table
-    wlog("fetching session events");
-    const rows = await query(
+    // 1. Load session events. Prefer the local per-session event cache the
+    // capture hook appends to as the session runs — it is row-for-row
+    // identical to the sessions-table `message` column but avoids re-scanning
+    // the entire fat `message` column on the backend for THIS session on every
+    // periodic / session-end trigger (the dominant cold-start cost on long
+    // mega-sessions). Falls back to the DB whenever the cache is absent
+    // (session resumed on another machine), empty, or — once the offset is
+    // known — shorter than the offset already summarized.
+    const dbFetch = () => query(
       `SELECT message, creation_date FROM "${cfg.sessionsTable}" ` +
       `WHERE path LIKE E'${esc(`/sessions/%${cfg.sessionId}%`)}' ORDER BY creation_date ASC`
     );
 
+    let usedLocalCache = false;
+    let rows: Record<string, unknown>[];
+    const cachedLines = readSessionEventCache(cfg.sessionId);
+    if (cachedLines && cachedLines.length > 0) {
+      rows = cachedLines.map(message => ({ message }));
+      usedLocalCache = true;
+      wlog(`loaded ${rows.length} events from local cache`);
+    } else {
+      wlog("fetching session events");
+      rows = await dbFetch();
+    }
+
     if (rows.length === 0) {
       wlog("no session events found — exiting");
       return;
     }
 
-    const jsonlLines = rows.length;
+    let jsonlLines = rows.length;
 
-    const pathRows = await query(
-      `SELECT DISTINCT path FROM "${cfg.sessionsTable}" ` +
-      `WHERE path LIKE '${esc(`/sessions/%${cfg.sessionId}%`)}' LIMIT 1`
-    );
-    const jsonlServerPath = pathRows.length > 0
-      ? pathRows[0].path as string
-      : `/sessions/unknown/${cfg.sessionId}.jsonl`;
+    // Derive the server path locally when using the cache (avoids a second
+    // self-session `SELECT DISTINCT path` scan); the DB branch keeps its lookup.
+    let jsonlServerPath: string;
+    if (usedLocalCache) {
+      jsonlServerPath = buildSessionPath(
+        { userName: cfg.userName, orgName: cfg.orgName, workspaceId: cfg.workspaceId },
+        cfg.sessionId,
+      );
+    } else {
+      const pathRows = await query(
+        `SELECT DISTINCT path FROM "${cfg.sessionsTable}" ` +
+        `WHERE path LIKE '${esc(`/sessions/%${cfg.sessionId}%`)}' LIMIT 1`
+      );
+      jsonlServerPath = pathRows.length > 0
+        ? pathRows[0].path as string
+        : `/sessions/unknown/${cfg.sessionId}.jsonl`;
+    }
 
     // 2. Determine how many rows were already summarized (resumed session).
     // The sidecar count is authoritative: finalizeSummary writes it after every
@@ -176,6 +207,15 @@ async function main(): Promise<void> {
       if (sidecarCount > prevOffset) prevOffset = sidecarCount;
     }
 
+    // Safety net: a local cache shorter than the summarized offset is an
+    // incomplete copy (session resumed on another machine) — re-load the full
+    // session from the DB so no genuinely-new rows get sliced to nothing.
+    if (usedLocalCache && rows.length < prevOffset) {
+      wlog(`local cache (${rows.length}) < summarized offset (${prevOffset}) — refetching from DB`);
+      rows = await dbFetch();
+      jsonlLines = rows.length;
+    }
+
     // Feed the agent only the rows added since the last summary. Reprocessing
     // the full session on every run is what drives ENOBUFS / 120s-timeout
     // failures on long (4000+ event) sessions — a stuck offset re-summarizes

src/hooks/hermes/capture.ts

@@ -35,6 +35,7 @@ import {
   releaseLock,
 } from "../summary-state.js";
 import { bundleDirFromImportMeta, spawnHermesWikiWorker, wikiLog } from "./spawn-wiki-worker.js";
+import { appendSessionEvent } from "../session-event-cache.js";
 import { tryStopCounterTrigger } from "../../skillify/triggers.js";
 import type { Config } from "../../config.js";
 import { getInstalledVersion } from "../../utils/version-check.js";
@@ -163,6 +164,12 @@ async function main(): Promise<void> {
 
   log("capture ok → cloud");
 
+  // Mirror the event into the local per-session cache (row-for-row identical
+  // to the `message` column just INSERTed) so the wiki-worker reads it instead
+  // of re-scanning the fat `message` column. Best-effort; only after a
+  // successful INSERT above.
+  appendSessionEvent(sessionId, line);
+
   // SkillOpt: a pre_llm_call prompt is the user's reaction to a recently-used org skill.
   // Swallowed; no-op unless a judgment window is open for this session.
   reactSkillOpt(sessionId, reactPrompt, "hermes");

src/hooks/hermes/spawn-wiki-worker.ts

@@ -105,6 +105,7 @@ export function spawnHermesWikiWorker(opts: SpawnOptions): void {
     sessionsTable: config.sessionsTableName,
     sessionId,
     userName: config.userName,
+    orgName: config.orgName,
     project: projectName,
     pluginVersion,
     tmpDir,

src/hooks/hermes/wiki-worker.ts

@@ -17,6 +17,8 @@ import { execFileSync } from "node:child_process";
 import { dirname, join } from "node:path";
 import { fileURLToPath } from "node:url";
 import { finalizeSummary, releaseLock, readState } from "../summary-state.js";
+import { readSessionEventCache } from "../session-event-cache.js";
+import { buildSessionPath } from "../../utils/session-path.js";
 import { capLinesByBytes, stampOffset, WIKI_JSONL_MAX_BYTES } from "../wiki-offset.js";
 import { uploadSummary } from "../upload-summary.js";
 import { log as _log } from "../../utils/debug.js";
@@ -35,6 +37,7 @@ interface WorkerConfig {
   sessionsTable: string;
   sessionId: string;
   userName: string;
+  orgName: string;
   project: string;
   pluginVersion?: string;
   tmpDir: string;
@@ -114,27 +117,55 @@ function cleanup(): void {
 
 async function main(): Promise<void> {
   try {
-    // 1. Fetch session events from sessions table
-    wlog("fetching session events");
-    const rows = await query(
+    // 1. Load session events. Prefer the local per-session event cache the
+    // capture hook appends to as the session runs — it is row-for-row
+    // identical to the sessions-table `message` column but avoids re-scanning
+    // the entire fat `message` column on the backend for THIS session on every
+    // periodic / session-end trigger (the dominant cold-start cost on long
+    // mega-sessions). Falls back to the DB whenever the cache is absent
+    // (session resumed on another machine), empty, or — once the offset is
+    // known — shorter than the offset already summarized.
+    const dbFetch = () => query(
       `SELECT message, creation_date FROM "${cfg.sessionsTable}" ` +
       `WHERE path LIKE E'${esc(`/sessions/%${cfg.sessionId}%`)}' ORDER BY creation_date ASC`
     );
 
+    let usedLocalCache = false;
+    let rows: Record<string, unknown>[];
+    const cachedLines = readSessionEventCache(cfg.sessionId);
+    if (cachedLines && cachedLines.length > 0) {
+      rows = cachedLines.map(message => ({ message }));
+      usedLocalCache = true;
+      wlog(`loaded ${rows.length} events from local cache`);
+    } else {
+      wlog("fetching session events");
+      rows = await dbFetch();
+    }
+
     if (rows.length === 0) {
       wlog("no session events found — exiting");
       return;
     }
 
-    const jsonlLines = rows.length;
+    let jsonlLines = rows.length;
 
-    const pathRows = await query(
-      `SELECT DISTINCT path FROM "${cfg.sessionsTable}" ` +
-      `WHERE path LIKE '${esc(`/sessions/%${cfg.sessionId}%`)}' LIMIT 1`
-    );
-    const jsonlServerPath = pathRows.length > 0
-      ? pathRows[0].path as string
-      : `/sessions/unknown/${cfg.sessionId}.jsonl`;
+    // Derive the server path locally when using the cache (avoids a second
+    // self-session `SELECT DISTINCT path` scan); the DB branch keeps its lookup.
+    let jsonlServerPath: string;
+    if (usedLocalCache) {
+      jsonlServerPath = buildSessionPath(
+        { userName: cfg.userName, orgName: cfg.orgName, workspaceId: cfg.workspaceId },
+        cfg.sessionId,
+      );
+    } else {
+      const pathRows = await query(
+        `SELECT DISTINCT path FROM "${cfg.sessionsTable}" ` +
+        `WHERE path LIKE '${esc(`/sessions/%${cfg.sessionId}%`)}' LIMIT 1`
+      );
+      jsonlServerPath = pathRows.length > 0
+        ? pathRows[0].path as string
+        : `/sessions/unknown/${cfg.sessionId}.jsonl`;
+    }
 
     // 2. Determine how many rows were already summarized (resumed session).
     // The sidecar count is authoritative: finalizeSummary writes it after every
@@ -176,6 +207,15 @@ async function main(): Promise<void> {
       if (sidecarCount > prevOffset) prevOffset = sidecarCount;
     }
 
+    // Safety net: a local cache shorter than the summarized offset is an
+    // incomplete copy (session resumed on another machine) — re-load the full
+    // session from the DB so no genuinely-new rows get sliced to nothing.
+    if (usedLocalCache && rows.length < prevOffset) {
+      wlog(`local cache (${rows.length}) < summarized offset (${prevOffset}) — refetching from DB`);
+      rows = await dbFetch();
+      jsonlLines = rows.length;
+    }
+
     // Feed the agent only the rows added since the last summary. Reprocessing
     // the full session on every run is what drives ENOBUFS / 120s-timeout
     // failures on long (4000+ event) sessions — a stuck offset re-summarizes

src/hooks/session-end.ts

@@ -14,6 +14,7 @@ import { resolveDirConfig } from "../dir-config.js";
 import { log as _log } from "../utils/debug.js";
 import { bundleDirFromImportMeta, spawnWikiWorker, wikiLog } from "./spawn-wiki-worker.js";
 import { tryAcquireLock, releaseLock, markSessionEnded } from "./summary-state.js";
+import { pruneStaleSessionEventCaches } from "./session-event-cache.js";
 import { forceSessionEndTrigger } from "../skillify/triggers.js";
 import { parseTranscript } from "../notifications/transcript-parser.js";
 import { appendUsageRecord } from "../notifications/usage-tracker.js";
@@ -67,6 +68,10 @@ async function main(): Promise<void> {
   // the activity window to lapse). Independent of the wiki-worker lock below.
   markSessionEnded(sessionId);
 
+  // Housekeeping: drop long-dead per-session event caches (14-day TTL). This
+  // session's own cache has a fresh mtime and is left intact for the worker.
+  try { pruneStaleSessionEventCaches(); } catch { /* best-effort */ }
+
   const base = loadConfig();
   if (!base) { log("no config"); return; }