@codeyam/codeyam-cli 0.1.0-staging.b8a55ba → 0.1.0-staging.bd4126b

package/analyzer-template/project/orchestrateCapture.ts

@@ -50,6 +50,9 @@ interface OrchestrateCaptureArgs {
   taskRunner: CaptureTaskRunner;
   analysisLoader?: AnalysisLoader;
   pollingInterval?: number;
+  // Optional commit for metadata/progress tracking only (not filtering).
+  // Use when full=true but you still want capture progress updates on a commit.
+  metadataCommit?: Commit | null;
 }
 
 export default async function orchestrateCapture({
@@ -62,6 +65,7 @@ export default async function orchestrateCapture({
   taskRunner,
   analysisLoader,
   pollingInterval = POLLING_INTERVAL,
+  metadataCommit,
 }: OrchestrateCaptureArgs) {
   if (!full && !analysisIds && !commit && !entityShas) {
     throw new Error(
@@ -95,15 +99,20 @@ export default async function orchestrateCapture({
   const attemptCount = new Map<string, number>();
   const maxAttempts = 2;
 
+  // Track which analyses had a worker report success — only these can be "captured"
+  const confirmedSuccessAnalyses = new Set<string>();
+
   // Register batch completion callback with task runner
-  taskRunner.onBatchCompleted((analysisIds: string[]) => {
+  taskRunner.onBatchCompleted((analysisIds: string[], success: boolean) => {
     console.log(
-      `CodeYam Capture: Received batch completion for ${analysisIds.length} analyses: ${analysisIds.join(', ')}`,
+      `CodeYam Capture: Received batch completion (success=${success}) for ${analysisIds.length} analyses: ${analysisIds.join(', ')}`,
     );
-    // Remove from inFlight and increment attempt count for each analysis
     for (const analysisId of analysisIds) {
       inFlight.delete(analysisId);
       attemptCount.set(analysisId, (attemptCount.get(analysisId) || 0) + 1);
+      if (success) {
+        confirmedSuccessAnalyses.add(analysisId);
+      }
     }
   });
 
@@ -191,9 +200,13 @@ export default async function orchestrateCapture({
         knownAnalyses.add(id);
       }
 
-      // an analysis we've seen, but is no longer "ready to be captured", is captured!
+      // An analysis is "captured" only if:
+      // 1. A worker reported success for its batch (in confirmedSuccessAnalyses)
+      // 2. It no longer appears in the ready-to-be-captured list (DB confirms)
+      // This prevents transient empty DB results from inflating the captured count.
       const capturedAnalysisIds = Array.from(knownAnalyses).filter(
-        (id) => !allReadyAnalysisIds.includes(id),
+        (id) =>
+          confirmedSuccessAnalyses.has(id) && !allReadyAnalysisIds.includes(id),
       );
 
       // Remove captured analyses from inFlight tracking
@@ -318,9 +331,12 @@ export default async function orchestrateCapture({
         // console.log(
         //   `Codeyam Debug: All known analyses: ${Array.from(knownAnalyses).join(', ')}`,
         // );
-        if (commit) {
-          commit.metadata = await updateCommitMetadata({
-            commitId: commit.id,
+        // Use metadataCommit if provided, otherwise fall back to commit.
+        // This allows progress tracking even when commit is null (e.g., full=true mode).
+        const trackingCommit = metadataCommit ?? commit;
+        if (trackingCommit) {
+          trackingCommit.metadata = await updateCommitMetadata({
+            commitId: trackingCommit.id,
             runStatusUpdate: {
               readyToBeCaptured: knownAnalyses.size,
               capturesAttempted: capturedAnalysisIds.length + abandoned.length,
@@ -348,6 +364,32 @@ export default async function orchestrateCapture({
         }
       }
 
+      // Early exit when all enqueued work is settled (captured or abandoned)
+      // This avoids waiting for the 20-minute ANALYSIS_TIMEOUT when capture is complete
+      // IMPORTANT: Only check this when isDraining (analysis phase complete) to avoid
+      // exiting while slow-running analyses are still in progress
+      if (
+        isDraining &&
+        !firstLoop &&
+        enqueuedAnalyses.size > 0 &&
+        inFlight.size === 0 &&
+        remaining.length === 0 &&
+        toRetry.length === 0
+      ) {
+        const allAnalysesSettled = Array.from(enqueuedAnalyses).every((id) => {
+          const captured = !allReadyAnalysisIds.includes(id);
+          const abandoned = (attemptCount.get(id) || 0) >= maxAttempts;
+          return captured || abandoned;
+        });
+
+        if (allAnalysesSettled) {
+          console.log(
+            'CodeYam Capture: All enqueued analyses settled (captured or abandoned), exiting early',
+          );
+          break;
+        }
+      }
+
       if (isDraining) {
         // Exit if analysis complete but no entities ready for capture
         // Only check this after at least one poll (not firstLoop) to ensure we've
@@ -364,7 +406,9 @@ export default async function orchestrateCapture({
           // check if all enqueued analyses have been either captured or abandoned
           const allAnalysesSettled = Array.from(enqueuedAnalyses).every(
             (id) => {
-              const captured = !allReadyAnalysisIds.includes(id);
+              const captured =
+                confirmedSuccessAnalyses.has(id) &&
+                !allReadyAnalysisIds.includes(id);
               const abandoned = (attemptCount.get(id) || 0) >= maxAttempts;
               return captured || abandoned;
             },
@@ -404,5 +448,33 @@ export default async function orchestrateCapture({
   } finally {
     // Always clean up task runner resources in finally block
     await taskRunner.cleanup();
+
+    const trackingCommit = metadataCommit ?? commit;
+    if (trackingCommit?.id) {
+      const captureCompletedAt = new Date().toISOString();
+
+      await updateCommitMetadata({
+        commitId: trackingCommit.id,
+        runStatusUpdate: {
+          captureCompletedAt,
+          capturePid: undefined,
+        },
+        // completedAt = max(analysisCompletedAt, captureCompletedAt)
+        // Normally capture finishes after analysis, but using max means illogical
+        // orderings leave a detectable signal (completedAt !== captureCompletedAt).
+        updateCallback(metadata) {
+          const run = metadata.currentRun!;
+          run.completedAt =
+            run.analysisCompletedAt &&
+            run.analysisCompletedAt > captureCompletedAt
+              ? run.analysisCompletedAt
+              : captureCompletedAt;
+        },
+      }).catch((error) => {
+        console.log(
+          `Warning: Failed to update capture completion metadata: ${error}`,
+        );
+      });
+    }
   }
 }

package/analyzer-template/project/reconcileMockDataKeys.ts

@@ -5,51 +5,49 @@
  * (e.g., "eq('user_id', 'user-12345')") when the mock code expects variable
  * references (e.g., "eq('user_id', user.id)").
  *
- * This function also handles canonical key format (EntityName::hookName::index)
- * and remaps old variable-based keys (variableName <- functionName) to canonical keys.
+ * 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
- * @param entityName Optional entity name for canonical key remapping
  */
 export function reconcileMockDataKeys(
   mockData: Record<string, unknown>,
   expectedKeys: string[],
-  entityName?: string,
 ): Record<string, unknown> {
   const result: Record<string, unknown> = {};
-  // Track which canonical keys have been used for multiple calls to same hook
-  const usedCanonicalKeys = new Set<string>();
 
   for (const [key, value] of Object.entries(mockData)) {
     // Recursively reconcile nested objects
     const reconciledValue =
       value && typeof value === 'object' && !Array.isArray(value)
-        ? reconcileMockDataKeys(
-            value as Record<string, unknown>,
-            expectedKeys,
-            entityName,
-          )
+        ? reconcileMockDataKeys(value as Record<string, unknown>, expectedKeys)
         : value;
 
     // Check if key matches an expected key exactly
     if (expectedKeys.includes(key)) {
       result[key] = reconciledValue;
-      usedCanonicalKeys.add(key);
       continue;
     }
 
-    // Try to remap variable-based key to canonical key
-    if (entityName) {
-      const canonicalKey = remapToCanonicalKey(
-        key,
-        expectedKeys,
-        entityName,
-        usedCanonicalKeys,
+    // 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 (canonicalKey) {
-        result[canonicalKey] = reconciledValue;
-        usedCanonicalKeys.add(canonicalKey);
+      if (partsMatchExpected) {
+        // Convert flat key to nested structure and merge
+        deepMergeIntoResult(
+          result,
+          flatKeyParts,
+          reconciledValue,
+          expectedKeys,
+        );
         continue;
       }
     }
@@ -68,76 +66,91 @@ export function reconcileMockDataKeys(
 }
 
 /**
- * Attempt to remap a variable-based key to a canonical key.
- *
- * Variable-based format: "variableName <- functionName"
- * Canonical format: "EntityName::hookName::index"
+ * 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.
  *
- * @example
- * remapToCanonicalKey("currentCommit <- useLoaderData", ["DashboardPage::useLoaderData::0"], "DashboardPage", new Set())
- * // returns "DashboardPage::useLoaderData::0"
+ * Returns null if the key doesn't look like a dotted path.
  */
-function remapToCanonicalKey(
-  key: string,
-  expectedCanonicalKeys: string[],
-  entityName: string,
-  usedKeys: Set<string>,
-): string | null {
-  // Check if key is in variable-based format: "varName <- functionName"
-  const arrowMatch = key.match(/<-\s*(\w+)/);
-  if (!arrowMatch) {
-    // Not a variable-based key, try other patterns
-    // Check if it's a simple function call like "useLoaderData()"
-    const funcCallMatch = key.match(/^(\w+)\(\)/);
-    if (funcCallMatch) {
-      const hookName = funcCallMatch[1];
-      return findMatchingCanonicalKey(
-        hookName,
-        entityName,
-        expectedCanonicalKeys,
-        usedKeys,
-      );
+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 hookName = arrowMatch[1];
+  const parts: string[] = [];
+  let current = '';
+  let depth = 0;
+
+  for (let i = 0; i < key.length; i++) {
+    const char = key[i];
 
-  return findMatchingCanonicalKey(
-    hookName,
-    entityName,
-    expectedCanonicalKeys,
-    usedKeys,
-  );
+    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;
 }
 
 /**
- * Find a matching canonical key for a hook name within an entity.
+ * Deep merge a value into the result object following the path segments.
+ * Reconciles the leaf key against expected keys.
  */
-function findMatchingCanonicalKey(
-  hookName: string,
-  entityName: string,
-  expectedCanonicalKeys: string[],
-  usedKeys: Set<string>,
-): string | null {
-  // Find all canonical keys for this entity and hook
-  const candidatePrefix = `${entityName}::${hookName}::`;
-  const candidates = expectedCanonicalKeys.filter(
-    (k) => k.startsWith(candidatePrefix) && !usedKeys.has(k),
-  );
+function deepMergeIntoResult(
+  result: Record<string, unknown>,
+  parts: string[],
+  value: unknown,
+  expectedKeys: string[],
+): void {
+  let current = result;
 
-  if (candidates.length === 0) {
-    return null;
+  // 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;
+    }
   }
 
-  // Sort by index to prefer lower indices
-  candidates.sort((a, b) => {
-    const aIndex = parseInt(a.split('::')[2], 10);
-    const bIndex = parseInt(b.split('::')[2], 10);
-    return aIndex - bIndex;
-  });
-
-  return candidates[0];
+  // Handle the leaf key
+  const leafKey = parts[parts.length - 1];
+  const closestLeafKey = findClosestKey(leafKey, expectedKeys) || leafKey;
+  current[closestLeafKey] = value;
 }
 
 /**
@@ -148,9 +161,15 @@ function findMatchingCanonicalKey(
  * - Ignore the last argument value (could be literal or variable)
  * - 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);
@@ -174,9 +193,132 @@ function findClosestKey(key: string, expectedKeys: string[]): string | null {
     }
   }
 
+  // If key is "funcName()" (empty args), try matching expected key "funcName({...})"
+  // This handles the case where LLM generates "getSurveysAction()" but expected key
+  // is "getSurveysAction({ environmentId, limit, ... })" with an object argument
+  const emptyArgsMatch = key.match(/^([a-zA-Z_][a-zA-Z0-9_]*)\(\)$/);
+  if (emptyArgsMatch) {
+    const funcName = emptyArgsMatch[1];
+    for (const expected of expectedKeys) {
+      // Match if expected key starts with the same function name and has args
+      const expectedMatch = expected.match(
+        new RegExp(`^${escapeRegExp(funcName)}\\([^)]+\\)$`),
+      );
+      if (expectedMatch) {
+        return expected;
+      }
+    }
+  }
+
+  // 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;
 }
 
+/**
+ * Escape special regex characters in a string.
+ */
+function escapeRegExp(str: string): string {
+  return str.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+}
+
 /**
  * Normalize quotes in a string: convert double quotes to single quotes.
  * This handles cases where the LLM generates keys with double quotes

package/analyzer-template/project/runAnalysis.ts

@@ -215,6 +215,11 @@ export default async function runAnalysis({
       entityShas,
       branchName: branch?.name,
     });
+
+    // Propagate the fallback branch to branchCommit so downstream code can use it
+    if (branchCommit && branch) {
+      branchCommit.branch = branch;
+    }
   }
 
   // After this block, 'commit' is the union of a regular or baseline commit
@@ -308,6 +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,
@@ -330,6 +339,7 @@ export default async function runAnalysis({
         desiredAnalyses,
         force,
         fileStore,
+        onAnalysisProducing: () => controller?.setAnalysisProducingPhase(),
       });
     }
 
@@ -350,6 +360,7 @@ export default async function runAnalysis({
       backgroundJob,
       desiredAnalyses,
       fileStore,
+      onAnalysisProducing: () => controller?.setAnalysisProducingPhase(),
     });
   } catch (e) {
     awsLog('CodeYam Error: Analysis failed', {