feat(compression): capture per-engine analytics (#3960) + Lite schema fix (#3952)

config/quality/file-size-baseline.json

@@ -20,6 +20,7 @@
   "_rebaseline_2026_06_16_3972_logs_autorefresh": "Issue #3972 own growth: RequestLoggerV2.tsx 1282->1287 (+5 = the auto-refresh interval now reads the live document.visibilityState each tick instead of a stale mount-time ref, plus a 3-line comment explaining the hidden-tab trap). Cohesive one-spot fix; structural shrink of this component tracked in #3501.",
   "_rebaseline_2026_06_16_3976_llm7_byteplus_models": "Issue #3976 own growth: models/route.ts 2489->2494 (+5 = add llm7 + byteplus to NAMED_OPENAI_STYLE_PROVIDERS with an explanatory comment so the import route does a live <baseUrl>/models fetch instead of serving the stale hardcoded registry catalog). Structural shrink of this route tracked in #3789.",
   "_rebaseline_2026_06_16_3954_cooldown_epoch": "Issue #3954 own growth: accountFallback.ts 1708->1727 (+19 = a shared cooldownUntilMs() normalizer + its use in isAccountUnavailable/getEarliestRateLimitedUntil/filterAvailableAccounts so a rate_limited_until persisted as a numeric-epoch string is honored, not parsed to NaN) and auth.ts 2216->2219 (+3 = parseFutureDateMs reuses cooldownUntilMs). Cohesive cooldown read-path hardening at the existing chokepoints; one helper, not extractable.",
+  "_rebaseline_2026_06_16_3960_engine_breakdown": "PR #3960 capture own growth: chatCore.ts 5851->5868 (+17 = persist result.stats.engineBreakdown into the new compression_engine_breakdown table after a stacked compression run, so getPerEngineAnalytics is accurate historically and not live-only). Cohesive analytics-persistence at the existing compression chokepoint; structural shrink of chatCore.ts tracked in #3501.",
   "_rebaseline_2026_06_15_3938_perplexity_v218": "PR #3938 own growth: perplexity-web.ts 868->939 (+71 = rebuild buildPplxRequestBody to mirror the current www.perplexity.ai schematized request body — version 2.18, use_schematized_api + the full supported_block_use_cases list, dsl_query, shared requestId for frontend_uuid/client_search_results_cache_key, last_backend_uuid only on follow-ups — plus the x-perplexity-request-* / x-request-id headers replacing the stale X-App-ApiVersion pair that triggered HTTP 400). Cohesive upstream-schema sync in a single executor; not extractable.",
   "cap": 800,
   "frozen": {
@@ -36,7 +37,7 @@
     "open-sse/executors/muse-spark-web.ts": 1284,
     "open-sse/executors/perplexity-web.ts": 1013,
     "open-sse/handlers/audioSpeech.ts": 965,
-    "open-sse/handlers/chatCore.ts": 5851,
+    "open-sse/handlers/chatCore.ts": 5868,
     "open-sse/handlers/imageGeneration.ts": 3777,
     "open-sse/handlers/responseSanitizer.ts": 1103,
     "open-sse/handlers/search.ts": 1442,

open-sse/handlers/chatCore.ts

@@ -2705,7 +2705,7 @@ export async function handleChatCore({
             compressionAnalyticsRecorded = true;
             compressionAnalyticsWritePromise = (async () => {
               try {
-                const { insertCompressionAnalyticsRow } =
+                const { insertCompressionAnalyticsRow, insertCompressionEngineBreakdown } =
                   await import("../../src/lib/db/compressionAnalytics.ts");
                 const { calculateCost } = await import("../../src/lib/usage/costCalculator.ts");
                 const tokensSaved = Math.max(
@@ -2746,6 +2746,23 @@ export async function handleChatCore({
                     ? rtkPointers.reduce((total, pointer) => total + pointer.bytes, 0)
                     : null,
                 });
+                // Persist the per-engine breakdown of a stacked run so per-engine
+                // analytics (getPerEngineAnalytics) is accurate historically, not just
+                // in the live `compression.completed` event.
+                const engineBreakdown = result.stats.engineBreakdown ?? [];
+                if (engineBreakdown.length > 0) {
+                  insertCompressionEngineBreakdown(
+                    engineBreakdown.map((b) => ({
+                      timestamp: new Date().toISOString(),
+                      request_id: skillRequestId,
+                      engine: b.engine,
+                      original_tokens: b.originalTokens,
+                      compressed_tokens: b.compressedTokens,
+                      tokens_saved: Math.max(0, b.originalTokens - b.compressedTokens),
+                      duration_ms: b.durationMs ?? null,
+                    }))
+                  );
+                }
               } catch (err) {
                 log?.debug?.(
                   "COMPRESSION",

open-sse/services/compression/engines/cavemanAdapter.ts

@@ -211,6 +211,29 @@ function validateUltraConfig(config: Record<string, unknown>): EngineValidationR
   return { valid: errors.length === 0, errors };
 }
 
+// Lite only honors `preserveSystemPrompt` (model/vision are runtime, not user config).
+// Previously this engine wrongly exposed AGGRESSIVE_SCHEMA, surfacing irrelevant
+// summarizer/threshold fields in the per-engine config UI.
+const LITE_SCHEMA: EngineConfigField[] = [
+  {
+    key: "preserveSystemPrompt",
+    type: "boolean",
+    label: "Preserve system prompt",
+    defaultValue: true,
+  },
+];
+
+function validateLiteConfig(config: Record<string, unknown>): EngineValidationResult {
+  const errors: string[] = [];
+  if (
+    config.preserveSystemPrompt !== undefined &&
+    typeof config.preserveSystemPrompt !== "boolean"
+  ) {
+    errors.push("preserveSystemPrompt must be a boolean");
+  }
+  return { valid: errors.length === 0, errors };
+}
+
 export const liteEngine: CompressionEngine = {
   id: "lite",
   name: "Lite",
@@ -240,10 +263,10 @@ export const liteEngine: CompressionEngine = {
     return this.apply(body, { stepConfig: config });
   },
   getConfigSchema() {
-    return AGGRESSIVE_SCHEMA;
+    return LITE_SCHEMA;
   },
   validateConfig(config) {
-    return validateAggressiveConfig(config);
+    return validateLiteConfig(config);
   },
 };
 

src/lib/db/compressionAnalytics.ts

@@ -30,6 +30,21 @@ export interface CompressionAnalyticsRow {
   rtk_raw_output_total_bytes?: number | null;
 }
 
+/**
+ * One row per engine that ran inside a stacked compression pipeline. A stacked
+ * request writes a single aggregate `compression_analytics` row (engine = mode) plus
+ * N of these — so per-engine savings are queryable historically, not just live.
+ */
+export interface CompressionEngineBreakdownRow {
+  timestamp: string;
+  request_id?: string | null;
+  engine: string;
+  original_tokens: number;
+  compressed_tokens: number;
+  tokens_saved: number;
+  duration_ms?: number | null;
+}
+
 export interface CompressionAnalyticsSummary {
   totalRequests: number;
   totalTokensSaved: number;
@@ -139,6 +154,53 @@ export function insertCompressionAnalyticsRow(row: CompressionAnalyticsRow): voi
   );
 }
 
+let breakdownTableEnsuredForDb: unknown = null;
+
+function ensureCompressionEngineBreakdownTable(): void {
+  const db = getDbInstance();
+  if (breakdownTableEnsuredForDb === db) return;
+  db.exec(`
+    CREATE TABLE IF NOT EXISTS compression_engine_breakdown (
+      id INTEGER PRIMARY KEY AUTOINCREMENT,
+      timestamp TEXT NOT NULL,
+      request_id TEXT,
+      engine TEXT NOT NULL,
+      original_tokens INTEGER NOT NULL DEFAULT 0,
+      compressed_tokens INTEGER NOT NULL DEFAULT 0,
+      tokens_saved INTEGER NOT NULL DEFAULT 0,
+      duration_ms INTEGER
+    );
+    CREATE INDEX IF NOT EXISTS idx_ceb_engine_ts ON compression_engine_breakdown(engine, timestamp);
+    CREATE INDEX IF NOT EXISTS idx_ceb_request ON compression_engine_breakdown(request_id);
+  `);
+  breakdownTableEnsuredForDb = db;
+}
+
+export function insertCompressionEngineBreakdown(rows: CompressionEngineBreakdownRow[]): void {
+  if (!rows.length) return;
+  const db = getDbInstance();
+  ensureCompressionEngineBreakdownTable();
+  const stmt = db.prepare(
+    `INSERT INTO compression_engine_breakdown
+       (timestamp, request_id, engine, original_tokens, compressed_tokens, tokens_saved, duration_ms)
+     VALUES (?, ?, ?, ?, ?, ?, ?)`
+  );
+  const insertAll = db.transaction((items: CompressionEngineBreakdownRow[]) => {
+    for (const r of items) {
+      stmt.run(
+        r.timestamp,
+        r.request_id ?? null,
+        r.engine,
+        r.original_tokens,
+        r.compressed_tokens,
+        r.tokens_saved,
+        r.duration_ms ?? null
+      );
+    }
+  });
+  insertAll(rows);
+}
+
 export function attachCompressionUsageReceipt(
   requestId: string | null | undefined,
   usage: Record<string, unknown> | null | undefined,
@@ -205,24 +267,53 @@ function appendCondition(whereClause: string, condition: string): string {
   return whereClause ? `${whereClause} AND ${condition}` : `WHERE ${condition}`;
 }
 
+type EngineAggRow = { runs: number; original: number; compressed: number; saved: number };
+
 export function getPerEngineAnalytics(engineId: string, days = 7) {
   const db = getDbInstance();
   ensureCompressionAnalyticsColumns();
+  ensureCompressionEngineBreakdownTable();
   const since = new Date(Date.now() - days * 86400_000).toISOString();
-  const row = db
+
+  // (1) Per-engine contributions from stacked runs (one breakdown row per engine).
+  const breakdown = db
+    .prepare(
+      `SELECT COUNT(*) AS runs,
+              COALESCE(SUM(original_tokens), 0) AS original,
+              COALESCE(SUM(compressed_tokens), 0) AS compressed,
+              COALESCE(SUM(tokens_saved), 0) AS saved
+       FROM compression_engine_breakdown
+       WHERE engine = ? AND timestamp >= ?`
+    )
+    .get(engineId, since) as EngineAggRow;
+
+  // (2) Legacy single-engine rows from compression_analytics, EXCLUDING any request
+  // that already has a per-engine breakdown — so a stacked run's aggregate row is not
+  // double-counted on top of its breakdown rows.
+  const legacy = db
     .prepare(
       `SELECT COUNT(*) AS runs,
               COALESCE(SUM(original_tokens), 0) AS original,
               COALESCE(SUM(compressed_tokens), 0) AS compressed,
               COALESCE(SUM(tokens_saved), 0) AS saved
        FROM compression_analytics
-       WHERE COALESCE(engine, mode) = ? AND timestamp >= ?`
+       WHERE COALESCE(engine, mode) = ? AND timestamp >= ?
+         AND (
+           request_id IS NULL
+           OR request_id NOT IN (
+             SELECT request_id FROM compression_engine_breakdown WHERE request_id IS NOT NULL
+           )
+         )`
     )
-    .get(engineId, since) as { runs: number; original: number; compressed: number; saved: number };
-  const tokensSaved = Math.max(0, row.saved);
+    .get(engineId, since) as EngineAggRow;
+
+  const runs = breakdown.runs + legacy.runs;
+  const original = breakdown.original + legacy.original;
+  const compressed = breakdown.compressed + legacy.compressed;
+  const tokensSaved = Math.max(0, breakdown.saved + legacy.saved);
   const avgSavingsPercent =
-    row.original > 0 ? Math.round(((row.original - row.compressed) / row.original) * 1000) / 10 : 0;
-  return { engineId, runs: row.runs, tokensSaved, avgSavingsPercent, days };
+    original > 0 ? Math.round(((original - compressed) / original) * 1000) / 10 : 0;
+  return { engineId, runs, tokensSaved, avgSavingsPercent, days };
 }
 
 export function getCompressionAnalyticsSummary(since?: string): CompressionAnalyticsSummary {

tests/unit/compression/engine-registry.test.ts

@@ -4,6 +4,7 @@ import assert from "node:assert/strict";
 import {
   aggressiveEngine,
   cavemanEngine,
+  liteEngine,
   ultraEngine,
 } from "../../../open-sse/services/compression/engines/cavemanAdapter.ts";
 import { rtkEngine as realRtkEngine } from "../../../open-sse/services/compression/engines/rtk/index.ts";
@@ -74,6 +75,15 @@ describe("compression engine registry contract", () => {
     assert.ok(rtkSchema.some((field) => field.key === "applyToCodeBlocks"));
     assert.ok(aggressiveSchema.some((field) => field.key === "maxTokensPerMessage"));
     assert.ok(ultraSchema.some((field) => field.key === "compressionRate"));
+
+    // Lite exposes its OWN minimal schema (preserveSystemPrompt), NOT the aggressive
+    // summarizer/threshold fields it previously leaked.
+    const liteSchema = liteEngine.getConfigSchema();
+    assert.ok(liteSchema.some((field) => field.key === "preserveSystemPrompt"));
+    assert.ok(!liteSchema.some((field) => field.key === "maxTokensPerMessage"));
+    assert.ok(!liteSchema.some((field) => field.key === "summarizerEnabled"));
+    assert.equal(liteEngine.validateConfig({ preserveSystemPrompt: true }).valid, true);
+    assert.equal(liteEngine.validateConfig({ preserveSystemPrompt: "yes" }).valid, false);
     assert.equal(cavemanEngine.validateConfig({ intensity: "full" }).valid, true);
     assert.equal(cavemanEngine.validateConfig({ intensity: "bad" }).valid, false);
     assert.equal(realRtkEngine.validateConfig({ maxLinesPerResult: 20 }).valid, true);