@codeyam/codeyam-cli 0.1.0-staging.da5baf5 → 0.1.0-staging.e090cb3

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
+      }
+    }
+  }
+}
+\`\`\``;
+}

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

@@ -1,21 +1,85 @@
-import { ScenariosDataStructure, Scenario, ScenarioData } from '~codeyam/types';
+import {
+  ScenariosDataStructure,
+  Scenario,
+  ScenarioData,
+  JsonTypeDefinition,
+  ExecutionFlow,
+} from '~codeyam/types';
 import noErrorAttributes from './noErrorAttributes';
+import simplifyKeysForLLM from './simplifyKeysForLLM';
+import collapseNullableObjects from './collapseNullableObjects';
+
+/**
+ * Sort object keys by the size of their JSON-stringified values (smallest first).
+ * This ensures small keys appear before large keys in the LLM prompt, increasing
+ * the chance they get generated before the LLM loses focus on large schemas.
+ */
+function sortKeysBySize(obj: Record<string, any>): Record<string, any> {
+  const entries = Object.entries(obj);
+  entries.sort(([, a], [, b]) => {
+    const sizeA = JSON.stringify(a).length;
+    const sizeB = JSON.stringify(b).length;
+    return sizeA - sizeB;
+  });
+  return Object.fromEntries(entries);
+}
+
+interface GeneratePromptOptions {
+  /** When true, mockData has already been generated via chunks - skip it in main call */
+  mockDataAlreadyGenerated?: boolean;
+}
 
 export default function generateEntityScenarioDataGenerator(
   structure: ScenariosDataStructure,
   scenario: Scenario,
+  executionFlows: ExecutionFlow[] | undefined,
   defaultScenarioData?: ScenarioData,
   incompleteResponse?: string,
+  options?: GeneratePromptOptions,
 ) {
   if (incompleteResponse) {
     return incompleteResponse;
   }
 
+  // Get the covered flows for this scenario and look up their requiredValues
+  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);
+
+  // Get blocking flows that are NOT covered by this scenario
+  // These should be explicitly avoided in the mock data
+  const blockingFlowsToAvoid = (executionFlows || [])
+    .filter(
+      (flow) =>
+        flow.blocksOtherFlows === true && !coveredFlowIds.includes(flow.id),
+    )
+    .map((flow) => ({
+      flowId: flow.id,
+      flowName: flow.name,
+      flowDescription: flow.description,
+      requiredValues: flow.requiredValues,
+    }));
+
   const scenarioJson = JSON.stringify(
     {
       name: scenario.name,
       description: scenario.description,
-      keyAttributeInstructions: scenario.metadata.keyAttributeInstructions,
+      // Include the flow requirements to guide data generation
+      coveredFlows: coveredFlowRequirements,
+      // Include blocking flows to avoid
+      ...(blockingFlowsToAvoid.length > 0 && {
+        blockingFlowsToAvoid,
+      }),
     },
     null,
     2,
@@ -30,12 +94,52 @@ ${JSON.stringify(defaultScenarioData.data, null, 2)}
 `
     : '';
 
-  const mockDataSection = structure.dataForMocks
-    ? `## mockData Structure
+  // Simplify verbose function call keys for LLM consumption, then sort by size.
+  // The static analysis generates keys like "useQuery({ id: Number(id), onSuccess: async (quote) => { ...30 lines... } })"
+  // which confuse the LLM. simplifyKeysForLLM reduces them to "useQuery({ id, onSuccess: () => <truncated> })"
+  // while the reconcileMockDataKeys function will still match the simplified response back to original keys.
+  // Then sort keys by size (smallest first) so small keys appear before large keys,
+  // increasing the chance the LLM generates all keys for large schemas.
+
+  const afterNoError = structure.dataForMocks
+    ? (noErrorAttributes(structure.dataForMocks) as JsonTypeDefinition)
+    : null;
+
+  const afterSimplify = afterNoError ? simplifyKeysForLLM(afterNoError) : null;
+
+  const cleanedDataForMocks = afterSimplify
+    ? collapseNullableObjects(afterSimplify)
+    : null;
+
+  const sortedDataForMocks =
+    cleanedDataForMocks &&
+    typeof cleanedDataForMocks === 'object' &&
+    !Array.isArray(cleanedDataForMocks)
+      ? sortKeysBySize(cleanedDataForMocks as Record<string, any>)
+      : cleanedDataForMocks;
+
+  // Log nullable-related content in final prompt JSON
+  if (sortedDataForMocks) {
+    const finalJson = JSON.stringify(sortedDataForMocks, null, 2);
+    const hasNullableMarker = finalJson.includes('_nullable');
+    const nullCollapsed = (finalJson.match(/"object \| null"/g) || []).length;
+    const arrayCollapsed = (finalJson.match(/"array \| null"/g) || []).length;
+    if (hasNullableMarker || nullCollapsed > 0 || arrayCollapsed > 0) {
+      console.log(
+        `[generateEntityScenarioDataGenerator] final prompt: ${nullCollapsed} "object | null", ${arrayCollapsed} "array | null"${hasNullableMarker ? ', WARNING: _nullable still present!' : ''}`,
+      );
+    }
+  }
+
+  // When mockData has already been generated via chunks, skip it in the main call
+  const mockDataSection = options?.mockDataAlreadyGenerated
+    ? '## mockData Structure\nmockData has been pre-generated—return `{}` for mockData.'
+    : sortedDataForMocks
+      ? `## mockData Structure
 \`\`\`json
-${JSON.stringify(noErrorAttributes(structure.dataForMocks), null, 2)}
+${JSON.stringify(sortedDataForMocks, null, 2)}
 \`\`\``
-    : '## mockData Structure\nNone—return `{}` for mockData.';
+      : '## mockData Structure\nNone—return `{}` for mockData.';
 
   const argumentsSection = structure.arguments
     ? `## argumentsData Structure

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

@@ -1,63 +1,30 @@
-import { Analysis, Entity } from '~codeyam/types';
+import { Entity, ExecutionFlow } from '~codeyam/types';
 
 interface GenerateEntityScenariosGeneratorArgs {
   entity: Pick<Entity, 'code' | 'metadata' | 'filePath' | 'name'>;
-  keyAttributes: Analysis['metadata']['keyAttributes'];
+  executionFlows?: ExecutionFlow[];
   error: boolean;
 }
 
 export default function generateEntityScenariosGenerator({
   entity,
-  keyAttributes,
+  executionFlows,
   error,
 }: GenerateEntityScenariosGeneratorArgs) {
-  const thisEntityKeyAttributes = keyAttributes?.filter(
-    (ka) => !ka.dependentEntityName,
-  );
-
-  const relevantDependentKeyAttributes = keyAttributes?.filter(
-    (ka) => ka.dependentEntityName,
-  );
-
-  const keyAttrsJson = JSON.stringify(
-    thisEntityKeyAttributes.map((ka) => ({
-      dataStructurePath: ka.dataStructurePath,
-      localVariable: ka.internalPath,
-      description: ka.description,
-      validValueOptions: ka.validValueOptions,
-    })),
-    null,
-    2,
-  );
-
-  const dependencySection =
-    !error &&
-    relevantDependentKeyAttributes &&
-    relevantDependentKeyAttributes.length > 0
-      ? `
-## Dependency Key Attributes
-\`\`\`json
-${JSON.stringify(
-  relevantDependentKeyAttributes.map((ka) => ({
-    dataStructurePath: ka.dataStructurePath,
-    localVariable: ka.internalPath,
-    forDependency: ka.dependentEntityName,
-    description: ka.description,
-    validValueOptions: ka.validValueOptions,
-  })),
-  null,
-  2,
-)}
-\`\`\``
-      : '';
-
-  return `## Key Attributes
+  // Build execution flows section
+  const executionFlowsSection =
+    !error && executionFlows && executionFlows.length > 0
+      ? `## Execution Flows
+These are all the distinct outcomes/behaviors this component can produce. Create scenarios that maximize coverage of these flows.
 \`\`\`json
-${keyAttrsJson}
+${JSON.stringify(executionFlows, null, 2)}
 \`\`\`
 
-## Code
+`
+      : '';
+
+  return `${executionFlowsSection}## Code
 \`\`\`typescript
 ${entity.code}
-\`\`\`${dependencySection}`;
+\`\`\``;
 }

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

@@ -0,0 +1,58 @@
+import { ExecutionFlow, Scenario } from '~codeyam/types';
+
+interface GenerateMissingKeysPromptArgs {
+  scenario: Scenario;
+  executionFlows: ExecutionFlow[] | undefined;
+  generatedMockData: Record<string, any>;
+  missingSchema: Record<string, any>;
+}
+
+/**
+ * Generate a prompt for the follow-up LLM call to fill in missing mockData keys.
+ * This is used when the initial Default Scenario response is missing some keys.
+ */
+export default function generateMissingKeysPrompt({
+  scenario,
+  executionFlows,
+  generatedMockData,
+  missingSchema,
+}: GenerateMissingKeysPromptArgs): string {
+  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,
+        requiredValues: flow.requiredValues,
+      };
+    })
+    .filter(Boolean);
+
+  return `The initial response was missing some keys. Generate mock data for ONLY the missing keys listed below.
+
+## Scenario Context
+\`\`\`json
+${JSON.stringify({ name: scenario.name, description: scenario.description, coveredFlows: coveredFlowRequirements }, null, 2)}
+\`\`\`
+
+## Already Generated Data
+\`\`\`json
+${JSON.stringify(generatedMockData, null, 2)}
+\`\`\`
+
+## Missing Keys to Generate
+\`\`\`json
+${JSON.stringify(missingSchema, null, 2)}
+\`\`\`
+
+Return ONLY a JSON object with this structure:
+\`\`\`json
+{
+  "mockData": {
+    // fill in ONLY the missing keys above
+  }
+}
+\`\`\``;
+}

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

@@ -1,24 +1,50 @@
 import { ScenariosDataStructure, Scenario } from '~codeyam/types';
 
+export interface ExecutionFlowSelection {
+  path: string;
+  value: string;
+  valueOptionIndex?: number;
+  isCustom: boolean;
+}
+
 interface GenerateEntityScenariosGeneratorArgs {
   description: string;
   existingScenarios: Scenario[];
   scenariosDataStructure: ScenariosDataStructure;
+  flowSelections?: ExecutionFlowSelection[];
 }
 
 export default function guessNewScenarioDataFromDescriptionGenerator({
   description,
   existingScenarios,
   scenariosDataStructure,
+  flowSelections,
 }: GenerateEntityScenariosGeneratorArgs) {
+  let flowSelectionsSection = '';
+  if (flowSelections && flowSelections.length > 0) {
+    flowSelectionsSection = `
+  User-selected Execution Flow Values:
+  The user has specifically requested these values be used in the scenario:
+  ${flowSelections
+    .map(
+      (sel) =>
+        `  - ${sel.path}: ${sel.value}${sel.isCustom ? ' (custom value)' : ''}`,
+    )
+    .join('\n')}
+
+  IMPORTANT: The mockData MUST include these specific values for the specified paths. Generate a scenario name and description that reflects these choices.
+  `;
+  }
+
   return `Mock Scenario Data Structure:
   \`\`\`
   ${JSON.stringify(scenariosDataStructure, null, 2)}
-  \`\`\
+  \`\`\`
   Existing Mock Scenario Data:
   \`\`\`
   ${JSON.stringify(existingScenarios, null, 2)}
   \`\`\`
-  New Scenario user-created prompt: "${description}"
+${flowSelectionsSection}
+  New Scenario user-created prompt: "${description || '(No additional description - generate based on selected execution flow values)'}"
   `;
 }