@codeyam/codeyam-cli 0.1.0-staging.d0ad4ae → 0.1.0-staging.d3e886e

package/analyzer-template/project/reconcileMockDataKeys.ts

@@ -7,6 +7,9 @@
  *
  * Keys are now original call signatures (e.g., "useFetcher<User>()", "db.select(query)").
  *
+ * Also handles flat dotted keys (e.g., "trpc.fastener.getAttributeValues.useQuery(...)")
+ * by converting them to nested structure.
+ *
  * @param mockData The mock data object to reconcile
  * @param expectedKeys The expected keys from the structure
  */
@@ -29,6 +32,26 @@ export function reconcileMockDataKeys(
       continue;
     }
 
+    // Check if this is a flat dotted key that should be converted to nested structure
+    // e.g., "trpc.fastener.getAttributeValues.useQuery({ ... })"
+    const flatKeyParts = parseFlatDottedKey(key);
+    if (flatKeyParts && flatKeyParts.length > 1) {
+      // Check if the parts match expected keys (indicating this should be nested)
+      const partsMatchExpected = flatKeyParts.some((part) =>
+        expectedKeys.includes(part),
+      );
+      if (partsMatchExpected) {
+        // Convert flat key to nested structure and merge
+        deepMergeIntoResult(
+          result,
+          flatKeyParts,
+          reconciledValue,
+          expectedKeys,
+        );
+        continue;
+      }
+    }
+
     // Try to find closest match using existing logic
     const closestKey = findClosestKey(key, expectedKeys);
     if (closestKey) {
@@ -42,6 +65,94 @@ export function reconcileMockDataKeys(
   return result;
 }
 
+/**
+ * Parse a flat dotted key into path segments.
+ * Handles keys like "trpc.fastener.getAttributeValues.useQuery({ ... })"
+ * where the last part might contain parentheses with function arguments.
+ *
+ * Returns null if the key doesn't look like a dotted path.
+ */
+function parseFlatDottedKey(key: string): string[] | null {
+  // Don't treat keys that start with a function call as dotted paths
+  // e.g., "useQuery({ id: 123 })" should not be parsed
+  if (key.match(/^[a-zA-Z_][a-zA-Z0-9_]*\s*\(/)) {
+    // Check if there's a dot BEFORE the first parenthesis
+    const parenIndex = key.indexOf('(');
+    const dotBeforeParen = key.slice(0, parenIndex).includes('.');
+    if (!dotBeforeParen) {
+      return null;
+    }
+  }
+
+  // Must contain a dot to be a dotted path
+  if (!key.includes('.')) {
+    return null;
+  }
+
+  const parts: string[] = [];
+  let current = '';
+  let depth = 0;
+
+  for (let i = 0; i < key.length; i++) {
+    const char = key[i];
+
+    if (char === '(' || char === '{' || char === '[') {
+      depth++;
+      current += char;
+    } else if (char === ')' || char === '}' || char === ']') {
+      depth--;
+      current += char;
+    } else if (char === '.' && depth === 0) {
+      if (current) {
+        parts.push(current);
+      }
+      current = '';
+    } else {
+      current += char;
+    }
+  }
+
+  if (current) {
+    parts.push(current);
+  }
+
+  // Need at least 2 parts for it to be a meaningful dotted path
+  return parts.length >= 2 ? parts : null;
+}
+
+/**
+ * Deep merge a value into the result object following the path segments.
+ * Reconciles the leaf key against expected keys.
+ */
+function deepMergeIntoResult(
+  result: Record<string, unknown>,
+  parts: string[],
+  value: unknown,
+  expectedKeys: string[],
+): void {
+  let current = result;
+
+  // Navigate/create path for all but the last part
+  for (let i = 0; i < parts.length - 1; i++) {
+    const part = parts[i];
+    if (!(part in current)) {
+      current[part] = {};
+    }
+    const next = current[part];
+    if (next && typeof next === 'object' && !Array.isArray(next)) {
+      current = next as Record<string, unknown>;
+    } else {
+      // Can't traverse further, structure mismatch
+      return;
+    }
+  }
+
+  // Handle the leaf key
+  const leafKey = parts[parts.length - 1];
+  const closestLeafKey = findClosestKey(leafKey, expectedKeys) || leafKey;
+  current[closestLeafKey] = value;
+}
+
 /**
  * Find the closest matching key from expectedKeys for a given key.
  *
@@ -51,10 +162,14 @@ export function reconcileMockDataKeys(
  * - Match based on the common prefix
  * - If key has no parentheses, try matching with "()" appended
  * - If key is "funcName()" (empty args), match expected key "funcName({...})" by function name
+ * - If key has object arguments (like useQuery({...})), match by function name
+ * - If key has single argument with literal value, match by function name
  *
  * e.g., "eq('user_id', 'user-12345')" matches "eq('user_id', user.id)"
  * e.g., "getAll" matches "getAll()"
  * e.g., "getSurveysAction()" matches "getSurveysAction({ environmentId, ... })"
+ * e.g., "useQuery({ id: 123 })" matches "useQuery({ id: Number(id) })"
+ * e.g., "useCustomSizes('example-slug')" matches "useCustomSizes(projectSlug)"
  */
 function findClosestKey(key: string, expectedKeys: string[]): string | null {
   const keyBase = extractKeyBase(key);
@@ -95,6 +210,105 @@ function findClosestKey(key: string, expectedKeys: string[]): string | null {
     }
   }
 
+  // If key has object literal arguments (like "useQuery({ id: 123, ... })"),
+  // try matching expected keys with the same function path AND first object property.
+  // This handles cases where LLM generates literal values but expected keys have variable references.
+  // e.g., "useQuery({ attributes: ['a', 'b'] })" matches "useQuery({ attributes: items || [] })"
+  // e.g., "trpc.users.getById.useQuery({ userId: 'user-123' })" matches "trpc.users.getById.useQuery({ userId })"
+  // Uses first property matching when multiple candidates exist for disambiguation.
+  const objectArgMatch = key.match(/^([a-zA-Z_][a-zA-Z0-9_.]*)\s*\(\s*\{/);
+  if (objectArgMatch) {
+    const funcPath = objectArgMatch[1]; // Can be "funcName" or "path.to.funcName"
+    // Extract the first property name from the object argument
+    const firstPropMatch = key.match(
+      /\(\s*\{\s*\n?\s*([a-zA-Z_][a-zA-Z0-9_]*)\s*:/,
+    );
+    const firstProp = firstPropMatch ? firstPropMatch[1] : null;
+
+    // Collect all matching candidates
+    const candidates: string[] = [];
+    for (const expected of expectedKeys) {
+      // Match if expected key has the same function path and starts with object argument
+      const expectedObjArgMatch = expected.match(
+        new RegExp(`^${escapeRegExp(funcPath)}\\s*\\(\\s*\\{`),
+      );
+      if (expectedObjArgMatch) {
+        candidates.push(expected);
+      }
+    }
+
+    // If we have a first property, prefer matching by property name
+    if (firstProp && candidates.length > 1) {
+      // Look for a candidate with the same first property
+      const propRegex = new RegExp(
+        `\\(\\s*\\{[\\s\\n]*${escapeRegExp(firstProp)}\\s*:`,
+      );
+      const propMatch = candidates.find((c) => propRegex.test(c));
+      if (propMatch) {
+        return propMatch;
+      }
+    }
+
+    // Fall back to first candidate if no property match found
+    if (candidates.length > 0) {
+      return candidates[0];
+    }
+  }
+
+  // If key has a single argument (literal value OR variable reference),
+  // try matching expected keys with the same function name but different argument.
+  // This handles cases where:
+  // - LLM generates "useCustomSizes('example-slug')" but expected key is "useCustomSizes(projectSlug)"
+  // - LLM generates "useCustomSizes(someProjectSlug)" but expected key is "useCustomSizes(projectSlug)"
+  //
+  // The second case is important because when LLM generates multiple keys with different
+  // variable names (e.g., "useCustomSizes(someProjectSlug)" and "useCustomSizes(projectSlug)"),
+  // without reconciliation, normalizeMockData merges them into "useCustomSizes()" which
+  // doesn't match the expected key "useCustomSizes(projectSlug)", causing undefined data access.
+  const singleArgMatch = key.match(
+    /^([a-zA-Z_][a-zA-Z0-9_.]*)\s*\(([^,)]+)\)$/,
+  );
+  if (singleArgMatch) {
+    const funcPath = singleArgMatch[1];
+    const arg = singleArgMatch[2].trim();
+    // Apply this matching if the argument is a literal OR a simple variable reference
+    const isLiteralArg = /^['"`].*['"`]$/.test(arg) || /^\d+$/.test(arg);
+    const isVariableRef = /^[a-zA-Z_][a-zA-Z0-9_]*$/.test(arg);
+    if (isLiteralArg || isVariableRef) {
+      for (const expected of expectedKeys) {
+        // Match if expected key has the same function path with a single argument
+        const expectedSingleArgMatch = expected.match(
+          new RegExp(`^${escapeRegExp(funcPath)}\\s*\\([^,)]+\\)$`),
+        );
+        if (expectedSingleArgMatch) {
+          return expected;
+        }
+      }
+    }
+  }
+
+  // If key has multiple arguments where one contains a literal value,
+  // try matching expected keys with the same function name and argument count.
+  // This handles cases like "getScenarioDisplayState(scenario, undefined, undefined, 'entitysha123', undefined)"
+  // matching "getScenarioDisplayState(scenario, undefined, undefined, entitySha, undefined)"
+  const multiArgWithLiteralMatch = key.match(
+    /^([a-zA-Z_][a-zA-Z0-9_.]*)\s*\((.+)\)$/,
+  );
+  if (multiArgWithLiteralMatch) {
+    const funcPath = multiArgWithLiteralMatch[1];
+    const argsStr = multiArgWithLiteralMatch[2];
+    // Check if any argument is a string literal
+    const hasLiteralArg = /'[^']*'|"[^"]*"/.test(argsStr);
+    if (hasLiteralArg) {
+      for (const expected of expectedKeys) {
+        // Match if expected key starts with the same function path
+        if (expected.startsWith(funcPath + '(')) {
+          return expected;
+        }
+      }
+    }
+  }
+
   return null;
 }
 

package/analyzer-template/project/runAnalysis.ts

@@ -313,6 +313,10 @@ export default async function runAnalysis({
   // Route to appropriate analyzer
   try {
     if (runReadyToBeCapturedPath) {
+      // Signal that analysis can produce captures immediately since analyses are already ready.
+      // This is needed for recapture to work - orchestration waits for the 'analysisProducing' event.
+      controller?.setAnalysisProducingPhase();
+
       return await analyzeReadyToBeCaptured({
         project,
         projectAnalyzer,
@@ -335,6 +339,7 @@ export default async function runAnalysis({
         desiredAnalyses,
         force,
         fileStore,
+        onAnalysisProducing: () => controller?.setAnalysisProducingPhase(),
       });
     }
 
@@ -355,6 +360,7 @@ export default async function runAnalysis({
       backgroundJob,
       desiredAnalyses,
       fileStore,
+      onAnalysisProducing: () => controller?.setAnalysisProducingPhase(),
     });
   } catch (e) {
     awsLog('CodeYam Error: Analysis failed', {

package/analyzer-template/project/start.ts

@@ -1,4 +1,5 @@
 import 'dotenv/config';
+import { once } from 'events';
 import yargs from 'yargs';
 import { hideBin } from 'yargs/helpers';
 import {
@@ -16,7 +17,8 @@ import { startScenarioCapture } from './startScenarioCapture';
 import { createController } from './controller/startController';
 import { awsLog, awsLogDebugLevel } from '~codeyam/utils';
 import * as v8 from 'v8';
-import { initializeAIServiceMode } from '~codeyam/ai';
+import * as os from 'os';
+import { initializeAIServiceMode, destroyWorkerPool } from '~codeyam/ai';
 import runAnalysis, { EnvConfig, SharedContext } from './runAnalysis';
 import {
   loadOrCreateCommit,
@@ -118,28 +120,31 @@ async function main({
     // and attempted to access project/commit/branch data before it was loaded.
     const sharedContext = await loadSharedContext(envConfig);
 
-    // Heap monitoring: pulse every 1s, log every 30s when normal, warn every 1s when high
+    // Memory monitoring: tracks both V8 heap and RSS (which includes worker threads)
+    const containerMemoryMB = process.env.ECS_CONTAINER_MEMORY_MB
+      ? parseInt(process.env.ECS_CONTAINER_MEMORY_MB, 10)
+      : Math.round(os.totalmem() / 1024 / 1024);
     let lastLogTime = 0;
     setInterval(() => {
       const mem = process.memoryUsage();
       const heap = v8.getHeapStatistics();
-      const usagePercent = (mem.heapUsed / heap.heap_size_limit) * 100;
+      const heapMB = Math.round(mem.heapUsed / 1024 / 1024);
+      const rssMB = Math.round(mem.rss / 1024 / 1024);
+      const heapLimitMB = Math.round(heap.heap_size_limit / 1024 / 1024);
+      const rssPercent = (rssMB / containerMemoryMB) * 100;
       const now = Date.now();
 
-      if (usagePercent >= 80) {
-        // High usage: warn every second
+      if (rssPercent >= 80) {
         console.warn(
-          `CodeYam Heap Warning: ${Math.round(mem.heapUsed / 1024 / 1024)}MB used / ${Math.round(heap.heap_size_limit / 1024 / 1024)}MB limit (${Math.round(usagePercent)}%) ⚠️ HIGH USAGE`,
+          `CodeYam Memory Warning: RSS ${rssMB}MB/${containerMemoryMB}MB (${Math.round(rssPercent)}%), heap ${heapMB}MB/${heapLimitMB}MB ⚠️ HIGH`,
         );
-      } else if (usagePercent >= 50 && now - lastLogTime >= 5000) {
-        // Elevated usage: log every 5 seconds
+      } else if (rssPercent >= 50 && now - lastLogTime >= 5000) {
         console.log(
-          `CodeYam Heap Notice: ${Math.round(mem.heapUsed / 1024 / 1024)}MB used / ${Math.round(heap.heap_size_limit / 1024 / 1024)}MB limit (${Math.round(usagePercent)}%) ⚠️ Elevated`,
+          `CodeYam Memory Notice: RSS ${rssMB}MB/${containerMemoryMB}MB (${Math.round(rssPercent)}%), heap ${heapMB}MB/${heapLimitMB}MB ⚠️ Elevated`,
         );
       } else if (now - lastLogTime >= 30000) {
-        // Normal usage: log every 30 seconds
         console.log(
-          `CodeYam Heap Stats: ${Math.round(mem.heapUsed / 1024 / 1024)}MB used / ${Math.round(heap.heap_size_limit / 1024 / 1024)}MB limit (${Math.round(usagePercent)}%) 📊 OK`,
+          `CodeYam Memory Stats: RSS ${rssMB}MB/${containerMemoryMB}MB (${Math.round(rssPercent)}%), heap ${heapMB}MB/${heapLimitMB}MB 📊 OK`,
         );
       }
 
@@ -186,6 +191,11 @@ async function main({
               analyzerPid: undefined,
               capturePid: process.pid,
             },
+            updateCallback(metadata) {
+              if (metadata.currentRun?.captureCompletedAt) {
+                metadata.currentRun.completedAt = new Date().toISOString();
+              }
+            },
           }).catch((error) => {
             awsLog(`Warning: Failed to update commit metadata: ${error}`);
           });
@@ -207,7 +217,20 @@ async function main({
         while (!controller?.hasCompletedRepoCheckout()) {
           await new Promise((resolve) => setTimeout(resolve, 1000));
         }
-        awsLog('CodeYam: Repo checkout complete, starting orchestration...');
+        awsLog('CodeYam: Repo checkout complete.');
+
+        // Wait for analysis to enter producing phase (after prepareDataStructures)
+        // This prevents the 20-minute orchestration timeout from triggering before
+        // any analyses become ready to capture
+        awsLog('CodeYam: Waiting for analysis to start producing...');
+        if (!controller.hasStartedProducing()) {
+          await once(controller, 'analysisProducing');
+        }
+        awsLog('CodeYam: Analysis producing, starting orchestration...');
+
+        // Worker pool is no longer needed after data structure prep phase.
+        // Freeing it reclaims ~2GB before capture workers ramp up.
+        await destroyWorkerPool();
 
         // Create task runner based on orchestration mode
         let taskRunner;
@@ -219,6 +242,17 @@ async function main({
             framework: framework,
             packageManager: packageManager,
           });
+
+          const captureCommit =
+            sharedContext.commit ?? sharedContext.branchCommit;
+          if (captureCommit) {
+            await updateCommitMetadata({
+              commitId: captureCommit.id,
+              runStatusUpdate: {
+                captureTaskArns: taskRunner.getTaskArns(),
+              },
+            });
+          }
         } else if (orchestrateCapture === 'local-sequential') {
           taskRunner = await SequentialCaptureTaskRunner.create(
             {