@codeyam/codeyam-cli 0.1.0-staging.e38f7bd → 0.1.0-staging.fef152f

package/analyzer-template/packages/ai/src/lib/promptGenerators/collapseNullableObjects.ts

@@ -0,0 +1,118 @@
+import { JsonTypeDefinition } from '~codeyam/types';
+
+/**
+ * Check whether all non-`_nullable` children of an object are simple strings.
+ * "Leaf-like" nullable objects (all-string children) get collapsed to
+ * `"object | null"` so the LLM knows the field can be null.
+ * Objects with nested structure (object/array children) keep their shape
+ * so the LLM can still generate proper data when the field is truthy.
+ */
+function isLeafLikeObject(obj: Record<string, any>): boolean {
+  for (const key of Object.keys(obj)) {
+    if (key === '_nullable') continue;
+    const value = obj[key];
+    if (typeof value !== 'string') return false;
+  }
+  return true;
+}
+
+/**
+ * Collapse *leaf-like* nullable objects/arrays in the data structure to type strings.
+ *
+ * When convertDotNotation marks an object or array as `_nullable: true`
+ * (meaning the schema had `object | undefined`, `object | null`, etc.),
+ * we decide whether to collapse it:
+ *
+ * - **Leaf-like** (all children are simple strings): collapse to `"object | null"`.
+ *   Example: `{ _nullable: true, type: "string", path: "string" }` → `"object | null"`
+ *
+ * - **Rich** (has object/array children): strip `_nullable` and keep the structure
+ *   so the LLM can generate proper nested data when the field is truthy.
+ *   Example: `{ _nullable: true, items: [...], name: "string" }` → `{ items: [...], name: "string" }`
+ *
+ * Non-nullable objects are recursively processed but left as nested structures.
+ */
+export default function collapseNullableObjects(
+  data: JsonTypeDefinition,
+): JsonTypeDefinition {
+  if (typeof data !== 'object' || data === null) {
+    return data;
+  }
+
+  if (Array.isArray(data)) {
+    return data.map((item: any) =>
+      typeof item === 'object' && item !== null
+        ? collapseNullableObjects(item as JsonTypeDefinition)
+        : item,
+    ) as any as JsonTypeDefinition;
+  }
+
+  const result: JsonTypeDefinition = {};
+
+  for (const key of Object.keys(data)) {
+    // Skip the _nullable marker itself (shouldn't appear at top level, but be safe)
+    if (key === '_nullable') continue;
+
+    const value = data[key];
+
+    if (typeof value === 'object' && value !== null && !Array.isArray(value)) {
+      const hasNullable = (value as any)._nullable === true;
+      if (hasNullable && isLeafLikeObject(value as Record<string, any>)) {
+        // Leaf-like nullable object → collapse to type string
+        const childKeys = Object.keys(value as any).filter(
+          (k) => k !== '_nullable',
+        );
+        console.log(
+          `[collapseNullableObjects] collapsed "${key}" → "object | null" (leaf-like, children: [${childKeys.join(', ')}])`,
+        );
+        result[key] = 'object | null';
+      } else if (hasNullable) {
+        // Rich nullable object → strip _nullable, keep structure
+        const childKeys = Object.keys(value as any).filter(
+          (k) => k !== '_nullable',
+        );
+        console.log(
+          `[collapseNullableObjects] kept structure for nullable "${key}" (rich, children: [${childKeys.join(', ')}])`,
+        );
+        result[key] = collapseNullableObjects(value as JsonTypeDefinition);
+      } else {
+        // Non-nullable → recurse
+        result[key] = collapseNullableObjects(value as JsonTypeDefinition);
+      }
+    } else if (Array.isArray(value)) {
+      const hasNullable = (value as any)._nullable === true;
+      if (
+        hasNullable &&
+        value.length <= 1 &&
+        value.every((v) => typeof v === 'string')
+      ) {
+        // Leaf-like nullable array → collapse to type string
+        console.log(
+          `[collapseNullableObjects] collapsed "${key}" → "array | null" (leaf-like)`,
+        );
+        result[key] = 'array | null';
+      } else if (hasNullable) {
+        // Rich nullable array → strip _nullable, keep structure
+        console.log(
+          `[collapseNullableObjects] kept structure for nullable array "${key}" (${value.length} elements)`,
+        );
+        result[key] = value.map((item) =>
+          typeof item === 'object' && item !== null
+            ? collapseNullableObjects(item as JsonTypeDefinition)
+            : item,
+        ) as JsonTypeDefinition[];
+      } else {
+        // Non-nullable → recurse elements
+        result[key] = value.map((item) =>
+          typeof item === 'object' && item !== null
+            ? collapseNullableObjects(item as JsonTypeDefinition)
+            : item,
+        ) as JsonTypeDefinition[];
+      }
+    } else {
+      result[key] = value;
+    }
+  }
+
+  return result;
+}

package/analyzer-template/packages/ai/src/lib/promptGenerators/gatherAttributesMap.ts

@@ -139,16 +139,19 @@ export function gatherAttributesMap(
       // Use merged type if available, otherwise fall back to isolated type
       const typeValue =
         mergedSignatureSchema[key] ?? isolatedSignatureSchema[key];
-      const equivalentSignatureVariable =
-        equivalentSignatureVariables[keyParts[0]];
-      const equivalentSignatureVariableParts = splitOutsideParenthesesAndArrays(
-        equivalentSignatureVariable,
-      );
+      // Handle array case (OR expressions) - use first element if array
+      const rawEquivalent = equivalentSignatureVariables[keyParts[0]];
+      const equivalentSignatureVariable = Array.isArray(rawEquivalent)
+        ? rawEquivalent[0]
+        : rawEquivalent;
+      const equivalentSignatureVariableParts = equivalentSignatureVariable
+        ? splitOutsideParenthesesAndArrays(equivalentSignatureVariable)
+        : [];
       if (
         equivalentSignatureVariable &&
-        !equivalentSignatureVariableParts[0].includes('(')
+        !equivalentSignatureVariableParts[0]?.includes('(')
       ) {
-        const equivalentKey = equivalentSignatureVariables[keyParts[0]];
+        const equivalentKey = equivalentSignatureVariable;
         const equivalentPath = joinParenthesesAndArrays([
           equivalentKey,
           ...keyParts.slice(1),
@@ -198,6 +201,13 @@ export function gatherAttributesMap(
         const pathParts = splitOutsideParenthesesAndArrays(path);
         if (pathParts.length < 2) continue;
 
+        // Filter out paths that contain signature[N] anywhere
+        // These represent function input parameters, not data from data sources
+        // Key attributes must reference data sources, not function inputs
+        if (pathParts.some((part) => part.startsWith('signature['))) {
+          continue;
+        }
+
         if (mergedReturnValueSchema[path] === 'function') {
           continue;
         }

package/analyzer-template/packages/ai/src/lib/promptGenerators/generateChangesEntityScenarioDataGenerator.ts

@@ -66,7 +66,7 @@ ${incompleteResponse}
     {
       name: scenario.name,
       description: scenario.description,
-      keyAttributeInstructions: scenario.metadata.keyAttributeInstructions,
+      coveredFlows: scenario.metadata?.coveredFlows || [],
     },
     null,
     2,

package/analyzer-template/packages/ai/src/lib/promptGenerators/generateChangesEntityScenariosGenerator.ts

@@ -1,19 +1,11 @@
-import {
-  DataStructure,
-  Entity,
-  Scenario,
-  ReadonlyAnalysisMap,
-  Analysis,
-} from '~codeyam/types';
-import gatherRelevantDependentKeyAttributes from '../gatherRelevantDependentKeyAttributes';
+import { DataStructure, Entity, ExecutionFlow, Scenario } from '~codeyam/types';
 
 interface GenerateChangesEntityScenariosGeneratorArgs {
   entity: Pick<Entity, 'name' | 'code' | 'metadata'>;
   mergedDataStructure: Omit<DataStructure, 'equivalentSignatureVariables'>;
-  keyAttributes: Analysis['metadata']['keyAttributes'];
-  addedKeyAttributes: Analysis['metadata']['keyAttributes'];
-  removedKeyAttributes: Analysis['metadata']['keyAttributes'];
-  dependentAnalyses: ReadonlyAnalysisMap;
+  executionFlows?: ExecutionFlow[];
+  addedFlows?: ExecutionFlow[];
+  removedFlows?: ExecutionFlow[];
   existingScenarios: Scenario[];
   commitDiff: string;
   error: boolean;
@@ -22,62 +14,26 @@ interface GenerateChangesEntityScenariosGeneratorArgs {
 
 export default function generateChangesEntityScenariosGenerator({
   entity,
-  mergedDataStructure,
-  keyAttributes,
-  addedKeyAttributes,
-  removedKeyAttributes,
-  dependentAnalyses,
+  executionFlows,
+  addedFlows,
+  removedFlows,
   existingScenarios,
   commitDiff,
   error,
   changedDataStructureFields = [],
 }: GenerateChangesEntityScenariosGeneratorArgs) {
-  const relevantDependentKeyAttributesFlat =
-    gatherRelevantDependentKeyAttributes(
-      entity,
-      mergedDataStructure,
-      dependentAnalyses,
-    );
-
-  // Restructure flat array into nested format: { [filePath]: { [entityName]: KeyAttribute[] } }
-  const relevantDependentKeyAttributes: Record<
-    string,
-    Record<string, any[]>
-  > = {};
-  for (const attr of relevantDependentKeyAttributesFlat) {
-    const filePath = (attr as any).filePath;
-    const entityName = (attr as any).originalEntityName;
-
-    if (!filePath || !entityName) continue;
-
-    if (!relevantDependentKeyAttributes[filePath]) {
-      relevantDependentKeyAttributes[filePath] = {};
-    }
-    if (!relevantDependentKeyAttributes[filePath][entityName]) {
-      relevantDependentKeyAttributes[filePath][entityName] = [];
-    }
-    relevantDependentKeyAttributes[filePath][entityName].push(attr);
-  }
-
-  const formatKeyAttrs = (attrs: Analysis['metadata']['keyAttributes']) =>
-    attrs.map((ka) => ({
-      dataStructurePath: ka.dataStructurePath,
-      localVariable: ka.internalPath,
-      description: ka.description,
-      validValueOptions: ka.validValueOptions,
-      // Include new fields when present
-      ...(ka.valueType && { valueType: ka.valueType }),
-      ...(ka.dependencies && { dependencies: ka.dependencies }),
-    }));
-
   const addedSection =
-    addedKeyAttributes?.length > 0
-      ? JSON.stringify(formatKeyAttrs(addedKeyAttributes), null, 2)
+    addedFlows && addedFlows.length > 0
+      ? JSON.stringify(addedFlows, null, 2)
       : 'None';
 
   const removedSection =
-    removedKeyAttributes?.length > 0
-      ? JSON.stringify(removedKeyAttributes, null, 2)
+    removedFlows && removedFlows.length > 0
+      ? JSON.stringify(
+          removedFlows.map((f) => ({ id: f.id, name: f.name })),
+          null,
+          2,
+        )
       : 'None';
 
   const changedFieldsSection =
@@ -89,18 +45,18 @@ Fields that changed type, were added, or removed:
 ${JSON.stringify(changedDataStructureFields, null, 2)}
 \`\`\`
 
-**Update keyAttributeInstructions for affected scenarios** (e.g., if "stats.visualEntities" changed from array to number).
+**Update scenarios if their coveredFlows depend on these changed fields.**
 `
       : '';
 
   const existingScenariosJson = JSON.stringify(
     existingScenarios
-      .filter((s) => (error ? !!s.metadata.error : !s.metadata.error))
+      .filter((s) => (error ? !!s.metadata?.error : !s.metadata?.error))
       .map((s) => ({
         name: s.name,
         description: s.description,
-        keyAttributeInstructions: s.metadata.keyAttributeInstructions,
-        playwrightInstructions: s.metadata.playwrightInstructions,
+        coveredFlows: s.metadata?.coveredFlows || [],
+        playwrightInstructions: s.metadata?.playwrightInstructions,
       })),
     null,
     2,
@@ -114,55 +70,29 @@ ${commitDiff}
 \`\`\``
     : '';
 
-  const dependencySection =
-    !error &&
-    relevantDependentKeyAttributes &&
-    Object.keys(relevantDependentKeyAttributes).length > 0
-      ? `
-## Dependency Key Attributes
+  const executionFlowsSection =
+    executionFlows && executionFlows.length > 0
+      ? `## Current Execution Flows
+All distinct outcomes/behaviors this component can currently produce:
 \`\`\`json
-${JSON.stringify(
-  Object.keys(relevantDependentKeyAttributes).reduce((acc: any, path) => {
-    acc[path] = Object.keys(relevantDependentKeyAttributes[path]).reduce(
-      (entityAcc: any, entityName) => {
-        entityAcc[entityName] = relevantDependentKeyAttributes[path][
-          entityName
-        ].map((ka: any) => ({
-          dataStructurePath: ka.dataStructurePath,
-          localVariable: ka.path || ka.internalPath,
-          description: ka.description,
-          validValueOptions: ka.validValueOptions,
-          // Include new fields when present
-          ...(ka.valueType && { valueType: ka.valueType }),
-          ...(ka.dependencies && { dependencies: ka.dependencies }),
-        }));
-        return entityAcc;
-      },
-      {},
-    );
-    return acc;
-  }, {}),
-  null,
-  2,
-)}
-\`\`\``
+${JSON.stringify(executionFlows, null, 2)}
+\`\`\`
+`
       : '';
 
-  return `## Added Key Attributes
+  return `## Added Execution Flows
+New flows that need scenario coverage:
 \`\`\`json
 ${addedSection}
 \`\`\`
 
-## Removed Key Attributes
+## Removed Execution Flows
+Flows that no longer exist (remove from scenarios):
 \`\`\`json
 ${removedSection}
 \`\`\`
 ${changedFieldsSection}
-## Current Key Attributes
-\`\`\`json
-${JSON.stringify(formatKeyAttrs(keyAttributes), null, 2)}
-\`\`\`
-
+${executionFlowsSection}
 ## Existing Scenarios
 \`\`\`json
 ${existingScenariosJson}
@@ -172,5 +102,5 @@ ${diffSection}
 ## Current Code
 \`\`\`typescript
 ${entity.code}
-\`\`\`${dependencySection}`;
+\`\`\``;
 }

package/analyzer-template/packages/ai/src/lib/promptGenerators/generateChunkPrompt.ts

@@ -0,0 +1,82 @@
+import {
+  ExecutionFlow,
+  Scenario,
+  ScenariosDataStructure,
+} from '~codeyam/types';
+import simplifyKeysForLLM from './simplifyKeysForLLM';
+import noErrorAttributes from './noErrorAttributes';
+import { JsonTypeDefinition } from '~codeyam/types';
+
+type RequiredValue = NonNullable<ExecutionFlow['requiredValues']>[number];
+
+interface GenerateChunkPromptArgs {
+  scenario: Scenario;
+  chunk: ScenariosDataStructure['dataForMocks'];
+  chunkIndex: number;
+  totalChunks: number;
+  relevantRequiredValues: RequiredValue[];
+}
+
+/**
+ * Generate a prompt for a focused LLM call to generate mock data for a single chunk.
+ *
+ * Large data structures overwhelm LLMs, causing them to make mistakes on some keys.
+ * By processing smaller chunks, each key gets more focused attention.
+ */
+export default function generateChunkPrompt({
+  scenario,
+  chunk,
+  chunkIndex,
+  totalChunks,
+  relevantRequiredValues,
+}: GenerateChunkPromptArgs): string {
+  // Clean up the schema for LLM consumption
+  const cleanedSchema = simplifyKeysForLLM(
+    noErrorAttributes(chunk) as JsonTypeDefinition,
+  );
+
+  const keysInChunk = Object.keys(chunk || {});
+
+  // Build requirements section if there are relevant requiredValues
+  const requirementsSection =
+    relevantRequiredValues.length > 0
+      ? `## Execution Flow Requirements for These Keys
+These specific values MUST be set to produce the correct scenario behavior:
+\`\`\`json
+${JSON.stringify(relevantRequiredValues, null, 2)}
+\`\`\`
+
+`
+      : '';
+
+  return `Generate mock data for chunk ${chunkIndex + 1} of ${totalChunks}.
+
+## Scenario
+\`\`\`json
+${JSON.stringify({ name: scenario.name, description: scenario.description }, null, 2)}
+\`\`\`
+
+${requirementsSection}## Keys in This Chunk (${keysInChunk.length} keys)
+Generate data for ONLY these keys:
+${keysInChunk.map((k) => `- \`${k}\``).join('\n')}
+
+## Schema for This Chunk
+\`\`\`json
+${JSON.stringify(cleanedSchema, null, 2)}
+\`\`\`
+
+## Instructions
+1. Match the EXACT structure shown in the schema
+2. Keys with special characters like \`*\` are LITERAL keys, not patterns
+3. If requiredValues are shown above, set values that satisfy them
+4. For paths like \`functionCall().functionCallReturnValue.property\`, the mock data key is just \`functionCall()\`
+
+Return ONLY a JSON object with this structure:
+\`\`\`json
+{
+  "mockData": {
+    // fill in ONLY the keys from this chunk
+  }
+}
+\`\`\``;
+}

package/analyzer-template/packages/ai/src/lib/promptGenerators/generateCriticalKeysPrompt.ts

@@ -0,0 +1,103 @@
+import {
+  ExecutionFlow,
+  Scenario,
+  ScenariosDataStructure,
+} from '~codeyam/types';
+import simplifyKeysForLLM from './simplifyKeysForLLM';
+import noErrorAttributes from './noErrorAttributes';
+import { JsonTypeDefinition } from '~codeyam/types';
+
+interface GenerateCriticalKeysPromptArgs {
+  scenario: Scenario;
+  executionFlows: ExecutionFlow[] | undefined;
+  criticalKeys: string[];
+  fullDataForMocks: ScenariosDataStructure['dataForMocks'];
+}
+
+/**
+ * Generate a prompt for a focused LLM call to generate only the critical mock data keys.
+ *
+ * When a scenario has covered flows with requiredValues, only a subset of the full
+ * dataForMocks schema is actually needed to satisfy those requirements. This prompt
+ * focuses the LLM on JUST those critical keys, avoiding confusion from large schemas.
+ *
+ * This is called BEFORE the main data generation to ensure critical keys are correctly
+ * generated with full attention from the LLM.
+ */
+export default function generateCriticalKeysPrompt({
+  scenario,
+  executionFlows,
+  criticalKeys,
+  fullDataForMocks,
+}: GenerateCriticalKeysPromptArgs): string {
+  // Build the subset schema with only critical keys
+  const criticalSchema: Record<string, any> = {};
+  if (fullDataForMocks && typeof fullDataForMocks === 'object') {
+    for (const key of criticalKeys) {
+      if (key in fullDataForMocks) {
+        criticalSchema[key] = (fullDataForMocks as Record<string, any>)[key];
+      }
+    }
+  }
+
+  // Clean up the schema for LLM consumption
+  const cleanedSchema = simplifyKeysForLLM(
+    noErrorAttributes(criticalSchema) as JsonTypeDefinition,
+  );
+
+  // Get the covered flows and their requirements
+  const coveredFlowIds = scenario.metadata?.coveredFlows || [];
+  const coveredFlowRequirements = coveredFlowIds
+    .map((flowId) => {
+      const flow = executionFlows?.find((f) => f.id === flowId);
+      if (!flow) return null;
+      return {
+        flowId: flow.id,
+        flowName: flow.name,
+        flowDescription: flow.description,
+        requiredValues: flow.requiredValues,
+      };
+    })
+    .filter(Boolean);
+
+  return `Generate mock data for these CRITICAL keys. Pay close attention to the requiredValues.
+
+## IMPORTANT: These keys are critical for satisfying the execution flow requirements
+The keys below directly control the scenario's behavior. Generate values that EXACTLY satisfy the requiredValues.
+
+## Scenario
+\`\`\`json
+${JSON.stringify({ name: scenario.name, description: scenario.description }, null, 2)}
+\`\`\`
+
+## Execution Flow Requirements
+These are the specific values that MUST be set:
+\`\`\`json
+${JSON.stringify(coveredFlowRequirements, null, 2)}
+\`\`\`
+
+## Critical Keys Schema (generate data for ONLY these keys)
+\`\`\`json
+${JSON.stringify(cleanedSchema, null, 2)}
+\`\`\`
+
+## Instructions
+1. Match the EXACT structure shown in the schema
+2. Keys with special characters like \`*\` are LITERAL keys, not patterns
+3. Set values that satisfy ALL the requiredValues above
+4. For paths like \`functionCall().functionCallReturnValue.property\`, the mock data key is just \`functionCall()\`
+
+Return ONLY a JSON object with this structure:
+\`\`\`json
+{
+  "scenarioData": {
+    "scenarioName": "${scenario.name}",
+    "data": {
+      "mockData": {
+        // fill in ONLY the critical keys above
+      }
+    }
+  }
+}
+\`\`\``;
+}