perf: reduce analytics page load — DB filter, parallel costs, merged queries (#1038)

Performance: reduces analytics page load via DB date filter, parallel cost calculation, and merged COUNT queries. Integrated into release/v3.5.4.

src/app/api/usage/analytics/route.ts

@@ -37,7 +37,11 @@ export async function GET(request) {
     const { searchParams } = new URL(request.url);
     const range = searchParams.get("range") || "30d";
 
-    const db = await getUsageDb();
+    // Cap history load to last 365 days — the heatmap never looks beyond that,
+    // and all named ranges (1d/7d/30d/90d/ytd) fall within this window.
+    const heatmapSince = new Date();
+    heatmapSince.setDate(heatmapSince.getDate() - 365);
+    const db = await getUsageDb(heatmapSince.toISOString());
     const history = db.data.history || [];
 
     // Build connection map for account names

src/app/api/v1/search/analytics/route.ts

@@ -16,39 +16,38 @@ export async function GET(req: Request) {
   try {
     const db = getDbInstance();
 
-    // Total search requests
-    const totalRow = db
-      .prepare(`SELECT COUNT(*) as cnt FROM call_logs WHERE request_type = 'search'`)
-      .get() as { cnt: number };
-    const total = totalRow?.cnt ?? 0;
-
-    // Today's searches (UTC date)
+    // Single aggregated query for all scalar metrics — replaces 5 separate round-trips
     const todayStart = new Date();
     todayStart.setUTCHours(0, 0, 0, 0);
-    const todayRow = db
-      .prepare(
-        `SELECT COUNT(*) as cnt FROM call_logs WHERE request_type = 'search' AND timestamp >= ?`
-      )
-      .get(todayStart.toISOString()) as { cnt: number };
-    const today = todayRow?.cnt ?? 0;
+    const todayIso = todayStart.toISOString();
 
-    // Errors
-    const errRow = db
+    type StatsRow = {
+      total: number;
+      today: number;
+      errors: number;
+      avg_duration: number | null;
+      cached: number;
+    };
+    const statsRow = db
       .prepare(
-        `SELECT COUNT(*) as cnt FROM call_logs WHERE request_type = 'search' AND (status >= 400 OR error IS NOT NULL)`
+        `SELECT
+          COUNT(*) as total,
+          COALESCE(SUM(CASE WHEN timestamp >= ? THEN 1 ELSE 0 END), 0) as today,
+          COALESCE(SUM(CASE WHEN status >= 400 OR error IS NOT NULL THEN 1 ELSE 0 END), 0) as errors,
+          AVG(CASE WHEN duration > 0 THEN duration END) as avg_duration,
+          COALESCE(SUM(CASE WHEN duration > 0 AND duration < 5 THEN 1 ELSE 0 END), 0) as cached
+         FROM call_logs
+         WHERE request_type = 'search'`
       )
-      .get() as { cnt: number };
-    const errors = errRow?.cnt ?? 0;
+      .get(todayIso) as StatsRow | undefined;
 
-    // Avg duration
-    const durRow = db
-      .prepare(
-        `SELECT AVG(duration) as avg FROM call_logs WHERE request_type = 'search' AND duration > 0`
-      )
-      .get() as { avg: number | null };
-    const avgDurationMs = Math.round(durRow?.avg ?? 0);
+    const total = statsRow?.total ?? 0;
+    const today = statsRow?.today ?? 0;
+    const errors = statsRow?.errors ?? 0;
+    const avgDurationMs = Math.round(statsRow?.avg_duration ?? 0);
+    const cached = statsRow?.cached ?? 0;
 
-    // Per-provider breakdown (provider column stores search provider id)
+    // Per-provider breakdown
     const provRows = db
       .prepare(
         `SELECT provider, COUNT(*) as cnt
@@ -74,14 +73,6 @@ export async function GET(req: Request) {
       totalCostUsd += cost;
     }
 
-    // Cached: very fast responses (< 5ms) indicate cache hits
-    const cachedRow = db
-      .prepare(
-        `SELECT COUNT(*) as cnt FROM call_logs 
-         WHERE request_type = 'search' AND duration > 0 AND duration < 5`
-      )
-      .get() as { cnt: number };
-    const cached = cachedRow?.cnt ?? 0;
     const cacheHitRate = total > 0 ? Math.round((cached / total) * 100) : 0;
 
     return NextResponse.json({

src/lib/usage/costCalculator.ts

@@ -18,7 +18,7 @@
  * @param {string} model
  * @returns {string}
  */
-function normalizeModelName(model) {
+export function normalizeModelName(model) {
   if (!model || !model.includes("/")) return model;
   const parts = model.split("/");
   return parts[parts.length - 1];
@@ -41,6 +41,41 @@ function toNumber(value: unknown, fallback = 0): number {
  * @param {Object} tokens
  * @returns {Promise<number>} Cost in USD
  */
+/**
+ * Compute cost synchronously from a pre-fetched pricing record.
+ * Use this when pricing has already been loaded (e.g. in batch analytics).
+ */
+export function computeCostFromPricing(
+  pricing: Record<string, unknown> | null | undefined,
+  tokens: any
+): number {
+  if (!pricing || !tokens) return 0;
+  const inputPrice = toNumber(pricing.input, 0);
+  const cachedPrice = toNumber(pricing.cached, inputPrice);
+  const outputPrice = toNumber(pricing.output, 0);
+  const reasoningPrice = toNumber(pricing.reasoning, outputPrice);
+  const cacheCreationPrice = toNumber(pricing.cache_creation, inputPrice);
+
+  let cost = 0;
+  const inputTokens = tokens.input ?? tokens.prompt_tokens ?? tokens.input_tokens ?? 0;
+  const cachedTokens =
+    tokens.cacheRead ?? tokens.cached_tokens ?? tokens.cache_read_input_tokens ?? 0;
+  const nonCachedInput = Math.max(0, inputTokens - cachedTokens);
+  cost += nonCachedInput * (inputPrice / 1_000_000);
+  if (cachedTokens > 0) cost += cachedTokens * (cachedPrice / 1_000_000);
+
+  const outputTokens = tokens.output ?? tokens.completion_tokens ?? tokens.output_tokens ?? 0;
+  cost += outputTokens * (outputPrice / 1_000_000);
+
+  const reasoningTokens = tokens.reasoning ?? tokens.reasoning_tokens ?? 0;
+  if (reasoningTokens > 0) cost += reasoningTokens * (reasoningPrice / 1_000_000);
+
+  const cacheCreationTokens = tokens.cacheCreation ?? tokens.cache_creation_input_tokens ?? 0;
+  if (cacheCreationTokens > 0) cost += cacheCreationTokens * (cacheCreationPrice / 1_000_000);
+
+  return cost;
+}
+
 export async function calculateCost(provider, model, tokens) {
   if (!tokens || !provider || !model) return 0;
 

src/lib/usage/usageHistory.ts

@@ -108,9 +108,13 @@ export function getPendingRequests() {
  * Returns an object compatible with the old LowDB interface.
  * Only `api/usage/analytics/route.js` uses this — it reads `db.data.history`.
  */
-export async function getUsageDb() {
+export async function getUsageDb(sinceIso?: string | null) {
   const db = getDbInstance();
-  const rows = db.prepare("SELECT * FROM usage_history ORDER BY timestamp ASC").all();
+  const rows = sinceIso
+    ? db
+        .prepare("SELECT * FROM usage_history WHERE timestamp >= ? ORDER BY timestamp ASC")
+        .all(sinceIso)
+    : db.prepare("SELECT * FROM usage_history ORDER BY timestamp ASC").all();
 
   const history = rows.map((row) => {
     const r = asRecord(row);

src/lib/usageAnalytics.ts

@@ -5,7 +5,10 @@
  * summary cards, daily trends, activity heatmap, model breakdown, etc.
  */
 
-import { calculateCost } from "@/lib/usageDb";
+import {
+  normalizeModelName as normModel,
+  computeCostFromPricing,
+} from "@/lib/usage/costCalculator";
 
 /**
  * Compute date range boundaries
@@ -128,8 +131,31 @@ export async function computeAnalytics(
     }
   }
 
+  // Pre-fetch pricing for all unique (provider, model) pairs — one DB round-trip
+  // per unique pair instead of one per entry, then compute costs synchronously.
+  const { getPricingForModel } = await import("@/lib/localDb");
+  const pricingCache = new Map<string, Record<string, unknown> | null>();
+  const uniquePairs = new Set(entries.map((e) => `${e.provider}|||${e.model}`));
+  await Promise.all(
+    [...uniquePairs].map(async (key) => {
+      const [provider, model] = key.split("|||");
+      let pricing = (await getPricingForModel(provider, model)) as Record<string, unknown> | null;
+      if (!pricing) {
+        const normalized = normModel(model);
+        if (normalized !== model) {
+          pricing = (await getPricingForModel(provider, normalized)) as Record<
+            string,
+            unknown
+          > | null;
+        }
+      }
+      pricingCache.set(key, pricing ?? null);
+    })
+  );
+
   // ---- Single pass over filtered entries for everything else ----
-  for (const entry of entries) {
+  for (let i = 0; i < entries.length; i++) {
+    const entry = entries[i];
     const pt = entry.tokens?.input ?? entry.tokens?.prompt_tokens ?? 0;
     const ct = entry.tokens?.output ?? entry.tokens?.completion_tokens ?? 0;
     const totalTkns = pt + ct;
@@ -138,13 +164,8 @@ export async function computeAnalytics(
     const dayOfWeek = entryDate.getDay();
     const modelShort = shortModelName(entry.model);
 
-    // Cost
-    let cost = 0;
-    try {
-      cost = await calculateCost(entry.provider, entry.model, entry.tokens);
-    } catch {
-      /* ignore */
-    }
+    const pricingKey = `${entry.provider}|||${entry.model}`;
+    const cost = computeCostFromPricing(pricingCache.get(pricingKey), entry.tokens);
 
     // Summary
     summary.promptTokens += pt;