trickle-cli 0.1.196 → 0.1.198

package/dist/commands/cost-report.js

@@ -262,6 +262,52 @@ function costReportCommand(opts) {
             console.log(`  ${chalk_1.default.cyan(name.padEnd(30))} $${data.cost.toFixed(4).padEnd(10)} ${chalk_1.default.gray(pct + '%')}  ${data.calls} calls  ${formatTokens(data.tokens)} tokens`);
         }
     }
+    // Cache hit/miss analysis — detect from latency bimodality
+    if (calls.length >= 4) {
+        // Group by model, find bimodal latency distribution
+        const modelLatencies = {};
+        for (const c of calls) {
+            if (!c.durationMs || c.error)
+                continue;
+            const key = c.model || 'unknown';
+            if (!modelLatencies[key])
+                modelLatencies[key] = [];
+            modelLatencies[key].push(c.durationMs);
+        }
+        let cacheDetected = false;
+        const cacheAnalysis = [];
+        for (const [model, latencies] of Object.entries(modelLatencies)) {
+            if (latencies.length < 3)
+                continue;
+            latencies.sort((a, b) => a - b);
+            const median = latencies[Math.floor(latencies.length / 2)];
+            // Split into fast (< 30% of median) and slow (>= 30% of median)
+            const threshold = median * 0.3;
+            const fast = latencies.filter(l => l < threshold);
+            const slow = latencies.filter(l => l >= threshold);
+            if (fast.length >= 1 && slow.length >= 1 && fast.length / latencies.length >= 0.1) {
+                const fastAvg = fast.reduce((s, l) => s + l, 0) / fast.length;
+                const slowAvg = slow.reduce((s, l) => s + l, 0) / slow.length;
+                // Only report if there's a significant speed difference (5x+)
+                if (slowAvg / Math.max(1, fastAvg) >= 5) {
+                    cacheDetected = true;
+                    cacheAnalysis.push({
+                        model, fastCalls: fast.length, slowCalls: slow.length,
+                        fastAvg: Math.round(fastAvg), slowAvg: Math.round(slowAvg),
+                        hitRate: Math.round((fast.length / latencies.length) * 100),
+                    });
+                }
+            }
+        }
+        if (cacheDetected) {
+            console.log(chalk_1.default.gray('\n  ' + '─'.repeat(60)));
+            console.log(chalk_1.default.bold('  Cache Analysis') + chalk_1.default.gray(' (detected from latency bimodality)'));
+            for (const ca of cacheAnalysis) {
+                const speedup = (ca.slowAvg / Math.max(1, ca.fastAvg)).toFixed(0);
+                console.log(`  ${chalk_1.default.cyan(ca.model.padEnd(25))} hit rate: ${chalk_1.default.green(ca.hitRate + '%')} (${ca.fastCalls} fast, ${ca.slowCalls} slow)  ${speedup}x speedup  fast=${ca.fastAvg}ms slow=${ca.slowAvg}ms`);
+            }
+        }
+    }
     if (costlyCalls.length > 0) {
         console.log(chalk_1.default.gray('\n  ' + '─'.repeat(60)));
         console.log(chalk_1.default.bold('  Most Expensive Calls'));

package/dist/commands/run.js

@@ -279,8 +279,7 @@ async function runCommand(command, opts) {
         if (!backendProc) {
             // Fall back to local/offline mode instead of exiting
             localMode = true;
-            console.log(chalk_1.default.yellow(`\n  Backend not available — using local mode (offline)`));
-            console.log(chalk_1.default.gray("  Observations will be saved to .trickle/observations.jsonl"));
+            // Silent for first-time users — local mode is the default experience
         }
     }
     // Detect language and inject instrumentation
@@ -400,13 +399,9 @@ async function executeSingleRun(instrumentedCommand, env, opts, singleFile, loca
         await autoCloudPush();
         // Generate post-run summary for AI agents
         (0, summary_1.writeRunSummary)({ exitCode, command: instrumentedCommand });
-        // Next steps hint
         console.log("");
-        console.log(chalk_1.default.bold("  Next steps:"));
-        console.log(chalk_1.default.gray("    trickle summary        ") + "full analysis (errors, queries, root causes)");
-        console.log(chalk_1.default.gray("    trickle explain <file>  ") + "understand a file (functions, call graph, data flow)");
-        console.log(chalk_1.default.gray("    trickle flamegraph      ") + "performance hotspots");
-        console.log(chalk_1.default.gray("    trickle test            ") + "run tests with observability");
+        console.log(chalk_1.default.gray("  trickle summary      ") + "full analysis");
+        console.log(chalk_1.default.gray("  trickle why          ") + "trace any error to root cause");
         console.log("");
         return exitCode;
     }
@@ -562,8 +557,19 @@ async function autoCloudPush() {
     try {
         const config = JSON.parse(fs.readFileSync(configPath, "utf-8"));
         if (config.url && config.token) {
-            const { cloudPush } = await Promise.resolve().then(() => __importStar(require("./cloud")));
-            await cloudPush();
+            // Suppress console output for auto-push to avoid noise
+            const origLog = console.log;
+            const origErr = console.error;
+            console.log = () => { };
+            console.error = () => { };
+            try {
+                const { cloudPush } = await Promise.resolve().then(() => __importStar(require("./cloud")));
+                await cloudPush();
+            }
+            finally {
+                console.log = origLog;
+                console.error = origErr;
+            }
         }
     }
     catch { }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "trickle-cli",
-  "version": "0.1.196",
+  "version": "0.1.198",
   "description": "CLI for trickle runtime type observability",
   "bin": {
     "trickle": "dist/index.js"

package/src/commands/cost-report.ts

@@ -241,6 +241,54 @@ export function costReportCommand(opts: { json?: boolean; budget?: string }): vo
     }
   }
 
+  // Cache hit/miss analysis — detect from latency bimodality
+  if (calls.length >= 4) {
+    // Group by model, find bimodal latency distribution
+    const modelLatencies: Record<string, number[]> = {};
+    for (const c of calls) {
+      if (!c.durationMs || c.error) continue;
+      const key = c.model || 'unknown';
+      if (!modelLatencies[key]) modelLatencies[key] = [];
+      modelLatencies[key].push(c.durationMs);
+    }
+
+    let cacheDetected = false;
+    const cacheAnalysis: Array<{ model: string; fastCalls: number; slowCalls: number; fastAvg: number; slowAvg: number; hitRate: number }> = [];
+
+    for (const [model, latencies] of Object.entries(modelLatencies)) {
+      if (latencies.length < 3) continue;
+      latencies.sort((a, b) => a - b);
+      const median = latencies[Math.floor(latencies.length / 2)];
+      // Split into fast (< 30% of median) and slow (>= 30% of median)
+      const threshold = median * 0.3;
+      const fast = latencies.filter(l => l < threshold);
+      const slow = latencies.filter(l => l >= threshold);
+
+      if (fast.length >= 1 && slow.length >= 1 && fast.length / latencies.length >= 0.1) {
+        const fastAvg = fast.reduce((s, l) => s + l, 0) / fast.length;
+        const slowAvg = slow.reduce((s, l) => s + l, 0) / slow.length;
+        // Only report if there's a significant speed difference (5x+)
+        if (slowAvg / Math.max(1, fastAvg) >= 5) {
+          cacheDetected = true;
+          cacheAnalysis.push({
+            model, fastCalls: fast.length, slowCalls: slow.length,
+            fastAvg: Math.round(fastAvg), slowAvg: Math.round(slowAvg),
+            hitRate: Math.round((fast.length / latencies.length) * 100),
+          });
+        }
+      }
+    }
+
+    if (cacheDetected) {
+      console.log(chalk.gray('\n  ' + '─'.repeat(60)));
+      console.log(chalk.bold('  Cache Analysis') + chalk.gray(' (detected from latency bimodality)'));
+      for (const ca of cacheAnalysis) {
+        const speedup = (ca.slowAvg / Math.max(1, ca.fastAvg)).toFixed(0);
+        console.log(`  ${chalk.cyan(ca.model.padEnd(25))} hit rate: ${chalk.green(ca.hitRate + '%')} (${ca.fastCalls} fast, ${ca.slowCalls} slow)  ${speedup}x speedup  fast=${ca.fastAvg}ms slow=${ca.slowAvg}ms`);
+      }
+    }
+  }
+
   if (costlyCalls.length > 0) {
     console.log(chalk.gray('\n  ' + '─'.repeat(60)));
     console.log(chalk.bold('  Most Expensive Calls'));

package/src/commands/run.ts

@@ -290,16 +290,7 @@ export async function runCommand(
     if (!backendProc) {
       // Fall back to local/offline mode instead of exiting
       localMode = true;
-      console.log(
-        chalk.yellow(
-          `\n  Backend not available — using local mode (offline)`,
-        ),
-      );
-      console.log(
-        chalk.gray(
-          "  Observations will be saved to .trickle/observations.jsonl",
-        ),
-      );
+      // Silent for first-time users — local mode is the default experience
     }
   }
 
@@ -453,13 +444,9 @@ async function executeSingleRun(
     // Generate post-run summary for AI agents
     writeRunSummary({ exitCode, command: instrumentedCommand });
 
-    // Next steps hint
     console.log("");
-    console.log(chalk.bold("  Next steps:"));
-    console.log(chalk.gray("    trickle summary        ") + "full analysis (errors, queries, root causes)");
-    console.log(chalk.gray("    trickle explain <file>  ") + "understand a file (functions, call graph, data flow)");
-    console.log(chalk.gray("    trickle flamegraph      ") + "performance hotspots");
-    console.log(chalk.gray("    trickle test            ") + "run tests with observability");
+    console.log(chalk.gray("  trickle summary      ") + "full analysis");
+    console.log(chalk.gray("  trickle why          ") + "trace any error to root cause");
     console.log("");
 
     return exitCode;
@@ -632,8 +619,18 @@ async function autoCloudPush(): Promise<void> {
   try {
     const config = JSON.parse(fs.readFileSync(configPath, "utf-8"));
     if (config.url && config.token) {
-      const { cloudPush } = await import("./cloud");
-      await cloudPush();
+      // Suppress console output for auto-push to avoid noise
+      const origLog = console.log;
+      const origErr = console.error;
+      console.log = () => {};
+      console.error = () => {};
+      try {
+        const { cloudPush } = await import("./cloud");
+        await cloudPush();
+      } finally {
+        console.log = origLog;
+        console.error = origErr;
+      }
     }
   } catch {}
 }