testchimp-runner-core 0.0.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (74) hide show
  1. package/dist/auth-config.d.ts +33 -0
  2. package/dist/auth-config.d.ts.map +1 -0
  3. package/dist/auth-config.js +69 -0
  4. package/dist/auth-config.js.map +1 -0
  5. package/dist/env-loader.d.ts +20 -0
  6. package/dist/env-loader.d.ts.map +1 -0
  7. package/dist/env-loader.js +83 -0
  8. package/dist/env-loader.js.map +1 -0
  9. package/dist/execution-service.d.ts +61 -0
  10. package/dist/execution-service.d.ts.map +1 -0
  11. package/dist/execution-service.js +822 -0
  12. package/dist/execution-service.js.map +1 -0
  13. package/dist/file-handler.d.ts +59 -0
  14. package/dist/file-handler.d.ts.map +1 -0
  15. package/dist/file-handler.js +75 -0
  16. package/dist/file-handler.js.map +1 -0
  17. package/dist/index.d.ts +46 -0
  18. package/dist/index.d.ts.map +1 -0
  19. package/dist/index.js +196 -0
  20. package/dist/index.js.map +1 -0
  21. package/dist/llm-facade.d.ts +101 -0
  22. package/dist/llm-facade.d.ts.map +1 -0
  23. package/dist/llm-facade.js +289 -0
  24. package/dist/llm-facade.js.map +1 -0
  25. package/dist/playwright-mcp-service.d.ts +42 -0
  26. package/dist/playwright-mcp-service.d.ts.map +1 -0
  27. package/dist/playwright-mcp-service.js +167 -0
  28. package/dist/playwright-mcp-service.js.map +1 -0
  29. package/dist/prompts.d.ts +34 -0
  30. package/dist/prompts.d.ts.map +1 -0
  31. package/dist/prompts.js +237 -0
  32. package/dist/prompts.js.map +1 -0
  33. package/dist/scenario-service.d.ts +25 -0
  34. package/dist/scenario-service.d.ts.map +1 -0
  35. package/dist/scenario-service.js +119 -0
  36. package/dist/scenario-service.js.map +1 -0
  37. package/dist/scenario-worker-class.d.ts +30 -0
  38. package/dist/scenario-worker-class.d.ts.map +1 -0
  39. package/dist/scenario-worker-class.js +263 -0
  40. package/dist/scenario-worker-class.js.map +1 -0
  41. package/dist/script-utils.d.ts +44 -0
  42. package/dist/script-utils.d.ts.map +1 -0
  43. package/dist/script-utils.js +100 -0
  44. package/dist/script-utils.js.map +1 -0
  45. package/dist/types.d.ts +171 -0
  46. package/dist/types.d.ts.map +1 -0
  47. package/dist/types.js +28 -0
  48. package/dist/types.js.map +1 -0
  49. package/dist/utils/browser-utils.d.ts +13 -0
  50. package/dist/utils/browser-utils.d.ts.map +1 -0
  51. package/dist/utils/browser-utils.js +269 -0
  52. package/dist/utils/browser-utils.js.map +1 -0
  53. package/dist/utils/page-info-utils.d.ts +16 -0
  54. package/dist/utils/page-info-utils.d.ts.map +1 -0
  55. package/dist/utils/page-info-utils.js +77 -0
  56. package/dist/utils/page-info-utils.js.map +1 -0
  57. package/env.prod +1 -0
  58. package/env.staging +1 -0
  59. package/package.json +38 -0
  60. package/src/auth-config.ts +84 -0
  61. package/src/env-loader.ts +91 -0
  62. package/src/execution-service.ts +999 -0
  63. package/src/file-handler.ts +104 -0
  64. package/src/index.ts +205 -0
  65. package/src/llm-facade.ts +413 -0
  66. package/src/playwright-mcp-service.ts +203 -0
  67. package/src/prompts.ts +247 -0
  68. package/src/scenario-service.ts +138 -0
  69. package/src/scenario-worker-class.ts +330 -0
  70. package/src/script-utils.ts +109 -0
  71. package/src/types.ts +202 -0
  72. package/src/utils/browser-utils.ts +272 -0
  73. package/src/utils/page-info-utils.ts +93 -0
  74. package/tsconfig.json +19 -0
@@ -0,0 +1,999 @@
1
+ import { PlaywrightMCPService as PlaywrightService } from './playwright-mcp-service';
2
+ import {
3
+ PlaywrightExecutionRequest,
4
+ PlaywrightExecutionResponse,
5
+ ScriptResult,
6
+ ScriptExecutionRequest,
7
+ ScriptExecutionResponse,
8
+ ScriptStep,
9
+ ExecutionMode,
10
+ StepOperation,
11
+ StepRepairAction
12
+ } from './types';
13
+ import { RepairSuggestionResponse, RepairConfidenceResponse } from './llm-facade';
14
+ import { Browser, BrowserContext, Page } from 'playwright';
15
+ import { expect } from '@playwright/test';
16
+ import { getEnhancedPageInfo, PageInfo } from './utils/page-info-utils';
17
+ import { initializeBrowser } from './utils/browser-utils';
18
+ import { LLMFacade } from './llm-facade';
19
+ import { AuthConfig } from './auth-config';
20
+ import { addTestChimpComment } from './script-utils';
21
+
22
+ /**
23
+ * Service for orchestrating Playwright script execution
24
+ */
25
+ export class ExecutionService {
26
+ private playwrightService: PlaywrightService;
27
+ private llmFacade: LLMFacade;
28
+
29
+ constructor(authConfig?: AuthConfig) {
30
+ this.playwrightService = new PlaywrightService();
31
+ this.llmFacade = new LLMFacade(authConfig);
32
+ }
33
+
34
+ /**
35
+ * Initialize the execution service
36
+ */
37
+ async initialize(): Promise<void> {
38
+ await this.playwrightService.initialize();
39
+ }
40
+
41
+
42
+ /**
43
+ * Execute a script with optional AI repair capabilities
44
+ */
45
+ async executeScript(request: ScriptExecutionRequest): Promise<ScriptExecutionResponse> {
46
+ const startTime = Date.now();
47
+ const model = request.model || 'gpt-4.1-mini';
48
+
49
+ try {
50
+ if (request.mode === ExecutionMode.RUN_EXACTLY) {
51
+ return await this.runExactly(request, startTime);
52
+ } else {
53
+ return await this.runWithAIRepair(request, startTime, model);
54
+ }
55
+ } catch (error) {
56
+ return {
57
+ run_status: 'failed',
58
+ executionTime: Date.now() - startTime,
59
+ error: error instanceof Error ? error.message : 'Unknown error'
60
+ };
61
+ }
62
+ }
63
+
64
+ /**
65
+ * Execute a complete Playwright test suite as a single job
66
+ */
67
+ async executeTestSuite(request: PlaywrightExecutionRequest): Promise<PlaywrightExecutionResponse> {
68
+ try {
69
+ // Parse Playwright configuration
70
+ const config = this.parsePlaywrightConfig(request.playwrightConfig);
71
+
72
+ // Execute the entire job (prescript + script + postscript) as one unit
73
+ const jobResult = await this.playwrightService.executeJob(
74
+ request.prescript,
75
+ request.script,
76
+ request.postscript,
77
+ config
78
+ );
79
+
80
+ return {
81
+ success: jobResult.success,
82
+ results: jobResult.results,
83
+ executionTime: jobResult.executionTime,
84
+ error: jobResult.error
85
+ };
86
+
87
+ } catch (error) {
88
+ return {
89
+ success: false,
90
+ results: {
91
+ script: { success: false, output: '', error: '', executionTime: 0 }
92
+ },
93
+ executionTime: 0,
94
+ error: error instanceof Error ? error.message : 'Unknown error occurred'
95
+ };
96
+ }
97
+ }
98
+
99
+ /**
100
+ * Parse Playwright configuration from string
101
+ */
102
+ private parsePlaywrightConfig(configString: string): any {
103
+ try {
104
+ // Try to parse as JSON first
105
+ const config = JSON.parse(configString);
106
+ return {
107
+ browserType: config.browserType || 'chromium',
108
+ headless: config.headless === true,
109
+ viewport: config.viewport || { width: 1280, height: 720 },
110
+ options: config.options || {}
111
+ };
112
+ } catch {
113
+ // If not JSON, try to extract basic config from JavaScript
114
+ try {
115
+ // Simple regex-based extraction for common config patterns
116
+ const headlessMatch = configString.match(/headless:\s*(true|false)/);
117
+ const viewportMatch = configString.match(/viewport:\s*\{\s*width:\s*(\d+),\s*height:\s*(\d+)\s*\}/);
118
+ const browserMatch = configString.match(/browserType:\s*['"`](chromium|firefox|webkit)['"`]/);
119
+
120
+ return {
121
+ browserType: browserMatch ? browserMatch[1] : 'chromium',
122
+ headless: headlessMatch ? headlessMatch[1] === 'true' : true,
123
+ viewport: viewportMatch ?
124
+ { width: parseInt(viewportMatch[1]), height: parseInt(viewportMatch[2]) } :
125
+ { width: 1280, height: 720 },
126
+ options: {}
127
+ };
128
+ } catch {
129
+ // Return default config if parsing fails
130
+ return {
131
+ browserType: 'chromium',
132
+ headless: false,
133
+ viewport: { width: 1280, height: 720 },
134
+ options: {}
135
+ };
136
+ }
137
+ }
138
+ }
139
+
140
+ /**
141
+ * Close the execution service
142
+ */
143
+ async close(): Promise<void> {
144
+ await this.playwrightService.close();
145
+ }
146
+
147
+ /**
148
+ * Check if the service is ready
149
+ */
150
+ isReady(): boolean {
151
+ return this.playwrightService.isReady();
152
+ }
153
+
154
+ private async runExactly(request: ScriptExecutionRequest, startTime: number, model?: string): Promise<ScriptExecutionResponse> {
155
+ const deflakeRunCount = request.deflake_run_count !== undefined ? request.deflake_run_count : 1;
156
+ const totalAttempts = deflakeRunCount + 1; // Original run + deflake attempts
157
+ let lastError: Error | null = null;
158
+
159
+ console.log(`runExactly: deflake_run_count = ${request.deflake_run_count}, totalAttempts = ${totalAttempts}`);
160
+
161
+ // Script content should be provided by the caller (TestChimpService)
162
+ // The TestChimpService handles file reading through the appropriate FileHandler
163
+ if (!request.script) {
164
+ throw new Error('Script content is required for execution. The TestChimpService should read the file and provide script content.');
165
+ }
166
+
167
+ for (let attempt = 1; attempt <= totalAttempts; attempt++) {
168
+ console.log(`Attempting deflake run ${attempt}/${totalAttempts}`);
169
+ const { browser, context, page } = await this.initializeBrowser(request.playwrightConfig, request.headless, request.playwrightConfigFilePath);
170
+
171
+ try {
172
+ // Execute the script as-is
173
+ await this.executeScriptContent(request.script, page);
174
+
175
+ await browser.close();
176
+
177
+ // Success! Return immediately
178
+ return {
179
+ run_status: 'success',
180
+ num_deflake_runs: attempt - 1, // Count only deflaking runs (exclude original run)
181
+ executionTime: Date.now() - startTime
182
+ };
183
+ } catch (error) {
184
+ lastError = error instanceof Error ? error : new Error('Script execution failed');
185
+ console.log(`Initial run failed: ${lastError.message}`);
186
+
187
+ try {
188
+ await browser.close();
189
+ } catch (closeError) {
190
+ // Browser might already be closed
191
+ }
192
+
193
+ // If this is not the last attempt, continue to next attempt
194
+ if (attempt < totalAttempts) {
195
+ console.log(`Deflaking attempt ${attempt} failed, trying again... (${attempt + 1}/${totalAttempts})`);
196
+ continue;
197
+ }
198
+ }
199
+ }
200
+
201
+ // All attempts failed
202
+ return {
203
+ run_status: 'failed',
204
+ num_deflake_runs: deflakeRunCount, // Count only deflaking runs (exclude original run)
205
+ executionTime: Date.now() - startTime,
206
+ error: lastError?.message || 'All deflaking attempts failed'
207
+ };
208
+ }
209
+
210
+ private async runWithAIRepair(request: ScriptExecutionRequest, startTime: number, model: string): Promise<ScriptExecutionResponse> {
211
+ const repairFlexibility = request.repair_flexibility || 3;
212
+
213
+ // Script content should be provided by the caller (TestChimpService)
214
+ // The TestChimpService handles file reading through the appropriate FileHandler
215
+ if (!request.script) {
216
+ throw new Error('Script content is required for AI repair. The TestChimpService should read the file and provide script content.');
217
+ }
218
+
219
+ // First, try runExactly (which includes deflaking if configured)
220
+ console.log('Attempting runExactly first (with deflaking if configured)...');
221
+ const runExactlyResult = await this.runExactly(request, startTime, model);
222
+
223
+ // If runExactly succeeded, return that result
224
+ if (runExactlyResult.run_status === 'success') {
225
+ return runExactlyResult;
226
+ }
227
+
228
+ // runExactly failed, start AI repair
229
+ console.log('runExactly failed, starting AI repair process...');
230
+
231
+ try {
232
+
233
+ // Start browser initialization and script parsing in parallel for faster startup
234
+ console.log('Initializing repair browser and parsing script...');
235
+ const [steps, { browser: repairBrowser, context: repairContext, page: repairPage }] = await Promise.all([
236
+ this.parseScriptIntoSteps(request.script, model),
237
+ this.initializeBrowser(request.playwrightConfig, request.headless, request.playwrightConfigFilePath) // Use request.headless (defaults to false/headed)
238
+ ]);
239
+
240
+ console.log('Starting AI repair with parsed steps...');
241
+ const updatedSteps = await this.repairStepsWithAI(steps, repairPage, repairFlexibility, model);
242
+
243
+ // Always generate the updated script
244
+ const updatedScript = this.generateUpdatedScript(updatedSteps);
245
+
246
+ // Check if repair was successful by seeing if we completed all steps
247
+ const allStepsSuccessful = updatedSteps.length > 0 && updatedSteps.every(step => step.success);
248
+
249
+ // Check if we have any successful repairs (partial success)
250
+ const hasSuccessfulRepairs = updatedSteps.some(step => step.success);
251
+
252
+ // Debug: Log step success status
253
+ console.log('Step success status:', updatedSteps.map((step, index) => `Step ${index + 1}: ${step.success ? 'SUCCESS' : 'FAILED'}`));
254
+ console.log('All steps successful:', allStepsSuccessful);
255
+ console.log('Has successful repairs:', hasSuccessfulRepairs);
256
+
257
+ // Debug: Log individual step details
258
+ updatedSteps.forEach((step, index) => {
259
+ console.log(`Step ${index + 1} details: success=${step.success}, description="${step.description}"`);
260
+ });
261
+
262
+ // Update file if we have any successful repairs (partial or complete)
263
+ if (hasSuccessfulRepairs) {
264
+ const confidenceResponse = await this.llmFacade.assessRepairConfidence(request.script!, updatedScript, model);
265
+ const finalScript = await this.llmFacade.generateFinalScript(request.script!, updatedScript, confidenceResponse.advice, model);
266
+
267
+ // Ensure the final script has the correct TestChimp comment format with repair advice
268
+ const scriptWithRepairAdvice = addTestChimpComment(finalScript, confidenceResponse.advice);
269
+
270
+ await repairBrowser.close();
271
+
272
+ return {
273
+ run_status: 'failed', // Original script failed
274
+ repair_status: allStepsSuccessful ? 'success' : 'partial', // Complete or partial repair success
275
+ repair_confidence: confidenceResponse.confidence,
276
+ repair_advice: confidenceResponse.advice,
277
+ updated_script: scriptWithRepairAdvice, // Return the drop-in replacement script with proper TestChimp comment
278
+ num_deflake_runs: runExactlyResult.num_deflake_runs, // All deflaking attempts failed
279
+ executionTime: Date.now() - startTime
280
+ };
281
+ } else {
282
+ // No successful repairs at all
283
+ await repairBrowser.close();
284
+
285
+ return {
286
+ run_status: 'failed', // Original script failed
287
+ repair_status: 'failed',
288
+ repair_confidence: 0,
289
+ repair_advice: 'AI repair could not fix any steps',
290
+ updated_script: request.script!, // Return original script since no repairs were successful
291
+ num_deflake_runs: runExactlyResult.num_deflake_runs, // All deflaking attempts failed
292
+ executionTime: Date.now() - startTime,
293
+ error: 'AI repair could not fix any steps'
294
+ };
295
+ }
296
+ } catch (error) {
297
+ return {
298
+ run_status: 'failed',
299
+ repair_status: 'failed',
300
+ num_deflake_runs: runExactlyResult.num_deflake_runs,
301
+ executionTime: Date.now() - startTime,
302
+ error: error instanceof Error ? error.message : 'Script execution failed'
303
+ };
304
+ }
305
+ }
306
+
307
+ private async parseScriptIntoSteps(script: string, model: string): Promise<(ScriptStep & { success?: boolean; error?: string })[]> {
308
+ // First try LLM-based parsing
309
+ try {
310
+ console.log('Attempting LLM-based script parsing...');
311
+ const result = await this.llmFacade.parseScriptIntoSteps(script, model);
312
+ console.log('LLM parsing successful, got', result.length, 'steps');
313
+ return result;
314
+ } catch (error) {
315
+ console.log('LLM parsing failed, falling back to code parsing:', error);
316
+ const fallbackResult = this.parseScriptIntoStepsFallback(script);
317
+ console.log('Fallback parsing successful, got', fallbackResult.length, 'steps');
318
+ return fallbackResult;
319
+ }
320
+ }
321
+
322
+
323
+ private parseScriptIntoStepsFallback(script: string): (ScriptStep & { success?: boolean; error?: string })[] {
324
+ const lines = script.split('\n');
325
+ const steps: (ScriptStep & { success?: boolean; error?: string })[] = [];
326
+ let currentStep: ScriptStep | null = null;
327
+ let currentCode: string[] = [];
328
+
329
+ for (const line of lines) {
330
+ const trimmedLine = line.trim();
331
+
332
+ // Check for step comment
333
+ if (trimmedLine.startsWith('// Step ')) {
334
+ // Save previous step if exists and has code
335
+ if (currentStep) {
336
+ const code = currentCode.join('\n').trim();
337
+ const cleanedCode = this.cleanStepCode(code);
338
+ if (cleanedCode) {
339
+ currentStep.code = cleanedCode;
340
+ steps.push(currentStep);
341
+ }
342
+ }
343
+
344
+ // Start new step
345
+ const description = trimmedLine.replace(/^\/\/\s*Step\s*\d+:\s*/, '').replace(/\s*\[FAILED\]\s*$/, '').trim();
346
+ currentStep = { description, code: '' };
347
+ currentCode = [];
348
+ } else if (trimmedLine && !trimmedLine.startsWith('import') && !trimmedLine.startsWith('test(') && !trimmedLine.startsWith('});')) {
349
+ // Add code line to current step
350
+ if (currentStep) {
351
+ currentCode.push(line);
352
+ }
353
+ }
354
+ }
355
+
356
+ // Add the last step if it has code
357
+ if (currentStep) {
358
+ const code = currentCode.join('\n').trim();
359
+ const cleanedCode = this.cleanStepCode(code);
360
+ if (cleanedCode) {
361
+ currentStep.code = cleanedCode;
362
+ steps.push(currentStep);
363
+ }
364
+ }
365
+
366
+ return steps;
367
+ }
368
+
369
+ private async repairStepsWithAI(
370
+ steps: (ScriptStep & { success?: boolean; error?: string })[],
371
+ page: Page,
372
+ repairFlexibility: number,
373
+ model: string
374
+ ): Promise<(ScriptStep & { success?: boolean; error?: string })[]> {
375
+ let updatedSteps = [...steps];
376
+ const maxTries = 3;
377
+ const recentRepairs: Array<{
378
+ stepNumber: number;
379
+ operation: string;
380
+ originalDescription?: string;
381
+ newDescription?: string;
382
+ originalCode?: string;
383
+ newCode?: string;
384
+ }> = [];
385
+
386
+ // Create a shared execution context that accumulates all executed code for variable tracking
387
+ let executionContext = '';
388
+ const contextVariables = new Map<string, any>();
389
+
390
+ let i = 0;
391
+ while (i < updatedSteps.length) {
392
+ const step = updatedSteps[i];
393
+ console.log(`Loop iteration: i=${i}, step description="${step.description}", total steps=${updatedSteps.length}`);
394
+
395
+ try {
396
+ // Try to execute the step directly without context replay
397
+ console.log(`Attempting Step ${i + 1}: ${step.description}`);
398
+ console.log(` Code: ${step.code}`);
399
+ await this.executeStepCode(step.code, page);
400
+ step.success = true;
401
+ console.log(`Step ${i + 1} executed successfully: ${step.description}`);
402
+ console.log(`Step ${i + 1} success status set to: ${step.success}`);
403
+
404
+ // Add this step's code to the execution context for future steps (for variable tracking)
405
+ executionContext += step.code + '\n';
406
+ i++; // Move to next step
407
+ } catch (error) {
408
+ console.log(`Step ${i + 1} failed: ${step.description}`);
409
+ console.log(` Failed code: ${step.code}`);
410
+ console.log(` Error: ${error instanceof Error ? error.message : 'Unknown error'}`);
411
+ if (error instanceof Error && error.stack) {
412
+ console.log(` Stack trace: ${error.stack}`);
413
+ }
414
+ step.success = false;
415
+ step.error = this.safeSerializeError(error);
416
+
417
+ // Try multiple repair attempts
418
+ const repairHistory: Array<{
419
+ attempt: number;
420
+ action: StepRepairAction;
421
+ error: string;
422
+ pageInfo: PageInfo;
423
+ }> = [];
424
+
425
+ let repairSuccess = false;
426
+ const originalDescription = step.description;
427
+ const originalCode = step.code;
428
+
429
+ for (let attempt = 1; attempt <= maxTries; attempt++) {
430
+ console.log(`Step ${i + 1} repair attempt ${attempt}/${maxTries}`);
431
+
432
+ // Get current page state for AI repair
433
+ const pageInfo = await this.getEnhancedPageInfo(page);
434
+
435
+ // Build failure history for LLM context
436
+ const failureHistory = this.buildFailureHistory(repairHistory, step, error);
437
+
438
+ // Build recent repairs context for LLM
439
+ const recentRepairsContext = this.buildRecentRepairsContext(recentRepairs);
440
+
441
+ // Ask AI for repair suggestion with failure history and recent repairs
442
+ const repairSuggestion = await this.llmFacade.getRepairSuggestion(
443
+ step.description,
444
+ step.code,
445
+ step.error || 'Unknown error',
446
+ pageInfo,
447
+ failureHistory,
448
+ recentRepairsContext,
449
+ model
450
+ );
451
+
452
+ if (!repairSuggestion.shouldContinue) {
453
+ console.log(`AI decided to stop repair at attempt ${attempt}: ${repairSuggestion.reason}`);
454
+ break;
455
+ }
456
+
457
+ // Apply the repair action
458
+ try {
459
+ // Set the step index and insertAfterIndex on the client side based on current step being processed
460
+ const repairAction = {
461
+ ...repairSuggestion.action,
462
+ stepIndex: i, // Client-side step index management
463
+ insertAfterIndex: repairSuggestion.action.operation === StepOperation.INSERT ? i - 1 : undefined // For INSERT, insert before current step
464
+ };
465
+
466
+ console.log(`🔧 Applying repair action:`, {
467
+ operation: repairAction.operation,
468
+ stepIndex: repairAction.stepIndex,
469
+ insertAfterIndex: repairAction.insertAfterIndex,
470
+ newStepDescription: repairAction.newStep?.description,
471
+ newStepCode: repairAction.newStep?.code
472
+ });
473
+ console.log(`🔧 Steps array before repair:`, updatedSteps.map((s, idx) => `${idx}: "${s.description}" (success: ${s.success})`));
474
+
475
+ const result = await this.applyRepairActionInContext(repairAction, updatedSteps, i, page, executionContext, contextVariables);
476
+
477
+ if (result.success) {
478
+ repairSuccess = true;
479
+ console.log(`🔧 Steps array after repair:`, updatedSteps.map((s, idx) => `${idx}: "${s.description}" (success: ${s.success})`));
480
+
481
+ // Mark the appropriate step(s) as successful based on operation type
482
+ if (repairAction.operation === StepOperation.MODIFY) {
483
+ // For MODIFY: mark the modified step as successful
484
+ step.success = true;
485
+ step.error = undefined;
486
+ updatedSteps[i].success = true;
487
+ updatedSteps[i].error = undefined;
488
+ console.log(`Step ${i + 1} marked as successful after MODIFY repair`);
489
+ } else if (repairAction.operation === StepOperation.INSERT) {
490
+ // For INSERT: mark the newly inserted step as successful
491
+ const insertIndex = repairAction.insertAfterIndex !== undefined ? repairAction.insertAfterIndex + 1 : i + 1;
492
+ if (updatedSteps[insertIndex]) {
493
+ updatedSteps[insertIndex].success = true;
494
+ updatedSteps[insertIndex].error = undefined;
495
+ }
496
+ } else if (repairAction.operation === StepOperation.REMOVE) {
497
+ // For REMOVE: no step to mark as successful since we removed it
498
+ // The step is already removed from the array
499
+ }
500
+
501
+ const commandInfo = repairAction.operation === StepOperation.MODIFY ?
502
+ `MODIFY: "${repairAction.newStep?.code || 'N/A'}"` :
503
+ repairAction.operation === StepOperation.INSERT ?
504
+ `INSERT: "${repairAction.newStep?.code || 'N/A'}"` :
505
+ repairAction.operation === StepOperation.REMOVE ?
506
+ `REMOVE: step at index ${repairAction.stepIndex}` :
507
+ repairAction.operation;
508
+ console.log(`Step ${i + 1} repair action ${commandInfo} executed successfully on attempt ${attempt}`);
509
+
510
+ // Update execution context based on the repair action
511
+ if (repairAction.operation === StepOperation.MODIFY && repairAction.newStep) {
512
+ // Update the step in the execution context for variable tracking
513
+ executionContext = executionContext.replace(originalCode, repairAction.newStep.code);
514
+ } else if (repairAction.operation === StepOperation.INSERT && repairAction.newStep) {
515
+ // Insert the new step code into execution context for variable tracking
516
+ executionContext += repairAction.newStep.code + '\n';
517
+ } else if (repairAction.operation === StepOperation.REMOVE) {
518
+ // Remove the step code from execution context for variable tracking
519
+ executionContext = executionContext.replace(originalCode, '');
520
+ }
521
+
522
+ // Record this successful repair
523
+ recentRepairs.push({
524
+ stepNumber: i + 1,
525
+ operation: repairAction.operation,
526
+ originalDescription: repairAction.operation === StepOperation.REMOVE ? originalDescription : undefined,
527
+ newDescription: repairAction.newStep?.description,
528
+ originalCode: repairAction.operation === StepOperation.REMOVE ? originalCode : undefined,
529
+ newCode: repairAction.newStep?.code
530
+ });
531
+
532
+ // Keep only the last 3 repairs for context
533
+ if (recentRepairs.length > 3) {
534
+ recentRepairs.shift();
535
+ }
536
+
537
+ // Update step index based on operation
538
+ if (repairAction.operation === StepOperation.INSERT) {
539
+ // For INSERT: inserted step is already executed
540
+ console.log(`INSERT operation: current i=${i}, insertAfterIndex=${repairAction.insertAfterIndex}`);
541
+ console.log(`INSERT: Steps array length before: ${updatedSteps.length}`);
542
+ console.log(`INSERT: Steps before operation:`, updatedSteps.map((s, idx) => `${idx}: "${s.description}" (success: ${s.success})`));
543
+
544
+ if (repairAction.insertAfterIndex !== undefined && repairAction.insertAfterIndex < i) {
545
+ // If inserting before current position, current step moved down by 1
546
+ console.log(`INSERT before current position: incrementing i from ${i} to ${i + 1}`);
547
+ i++; // Move to the original step that was pushed to the next position
548
+ } else {
549
+ // If inserting at or after current position, stay at current step
550
+ console.log(`INSERT at/after current position: keeping i at ${i}`);
551
+ }
552
+
553
+ console.log(`INSERT: Steps array length after: ${updatedSteps.length}`);
554
+ console.log(`INSERT: Steps after operation:`, updatedSteps.map((s, idx) => `${idx}: "${s.description}" (success: ${s.success})`));
555
+ } else if (repairAction.operation === StepOperation.REMOVE) {
556
+ // For REMOVE: stay at same index since the next step moved to current position
557
+ // Don't increment i because the array shifted left
558
+ } else {
559
+ // For MODIFY: move to next step since modified step was executed
560
+ i++; // Move to next step for MODIFY
561
+ }
562
+
563
+ // Add the repaired step's code to execution context for variable tracking
564
+ executionContext += step.code + '\n';
565
+
566
+ break;
567
+ } else {
568
+ throw new Error(result.error || 'Repair action failed');
569
+ }
570
+ } catch (repairError) {
571
+ const repairErrorMessage = repairError instanceof Error ? repairError.message : 'Repair failed';
572
+ const commandInfo = repairSuggestion.action.operation === StepOperation.MODIFY ?
573
+ `MODIFY: "${repairSuggestion.action.newStep?.code || 'N/A'}"` :
574
+ repairSuggestion.action.operation === StepOperation.INSERT ?
575
+ `INSERT: "${repairSuggestion.action.newStep?.code || 'N/A'}"` :
576
+ repairSuggestion.action.operation === StepOperation.REMOVE ?
577
+ `REMOVE: step at index ${repairSuggestion.action.stepIndex}` :
578
+ repairSuggestion.action.operation;
579
+ console.log(`Step ${i + 1} repair attempt ${attempt} failed (${commandInfo}): ${repairErrorMessage}`);
580
+ if (repairError instanceof Error && repairError.stack) {
581
+ console.log(` Repair stack trace: ${repairError.stack}`);
582
+ }
583
+
584
+ // Record this attempt in history
585
+ repairHistory.push({
586
+ attempt,
587
+ action: repairSuggestion.action,
588
+ error: repairErrorMessage,
589
+ pageInfo
590
+ });
591
+
592
+ step.error = repairErrorMessage;
593
+ }
594
+ }
595
+
596
+ if (!repairSuccess) {
597
+ console.log(`Step ${i + 1} failed after ${maxTries} repair attempts`);
598
+ break;
599
+ }
600
+ }
601
+ }
602
+
603
+ return updatedSteps;
604
+ }
605
+
606
+ private async executeStepCode(code: string, page: Page): Promise<void> {
607
+ // Set timeout for individual step execution
608
+ page.setDefaultTimeout(5000); // 5 seconds for individual commands
609
+
610
+ try {
611
+ // Clean and validate the code before execution
612
+ const cleanedCode = this.cleanStepCode(code);
613
+
614
+ if (!cleanedCode || cleanedCode.trim().length === 0) {
615
+ throw new Error('Step code is empty or contains only comments');
616
+ }
617
+
618
+ // Create an async function that has access to page, expect, and other Playwright globals
619
+ const executeCode = new Function('page', 'expect', `return (async () => { ${cleanedCode} })()`);
620
+ const result = executeCode(page, expect);
621
+ await result;
622
+ } finally {
623
+ // Restore to reasonable default timeout
624
+ page.setDefaultTimeout(10000);
625
+ }
626
+ }
627
+
628
+ /**
629
+ * Validate step code has executable content (preserves comments)
630
+ */
631
+ private cleanStepCode(code: string): string {
632
+ if (!code || code.trim().length === 0) {
633
+ return '';
634
+ }
635
+
636
+ // Check if there are any executable statements (including those with comments)
637
+ const hasExecutableCode = /[a-zA-Z_$][a-zA-Z0-9_$]*\s*\(|await\s+|return\s+|if\s*\(|for\s*\(|while\s*\(|switch\s*\(|try\s*\{|catch\s*\(/.test(code);
638
+
639
+ if (!hasExecutableCode) {
640
+ return '';
641
+ }
642
+
643
+ return code; // Return the original code without removing comments
644
+ }
645
+
646
+ private async executeStepInContext(
647
+ code: string,
648
+ page: Page,
649
+ executionContext: string,
650
+ contextVariables: Map<string, any>
651
+ ): Promise<void> {
652
+ // Set timeout for individual step execution
653
+ page.setDefaultTimeout(5000); // 5 seconds for individual commands
654
+
655
+ try {
656
+ // Execute only the current step code, but make context variables available
657
+ const fullCode = code;
658
+
659
+ // Create a function that has access to page, expect, and the context variables
660
+ const executeCode = new Function(
661
+ 'page',
662
+ 'expect',
663
+ 'contextVariables',
664
+ `return (async () => {
665
+ // Make context variables available in the execution scope
666
+ for (const [key, value] of contextVariables) {
667
+ globalThis[key] = value;
668
+ }
669
+
670
+ ${fullCode}
671
+
672
+ // Capture any new variables that might have been created
673
+ const newVars = {};
674
+ for (const key in globalThis) {
675
+ if (!contextVariables.has(key) && typeof globalThis[key] !== 'function' && key !== 'page' && key !== 'expect') {
676
+ newVars[key] = globalThis[key];
677
+ }
678
+ }
679
+ return newVars;
680
+ })()`
681
+ );
682
+
683
+ const newVars = await executeCode(page, expect, contextVariables);
684
+
685
+ // Update the context variables with any new variables created
686
+ for (const [key, value] of Object.entries(newVars)) {
687
+ contextVariables.set(key, value);
688
+ }
689
+ } finally {
690
+ // Restore to reasonable default timeout
691
+ page.setDefaultTimeout(5000);
692
+ }
693
+ }
694
+
695
+ private async executeScriptContent(script: string, page: Page): Promise<void> {
696
+ // Extract the test function content
697
+ const testMatch = script.match(/test\([^,]+,\s*async\s*\(\s*\{\s*page[^}]*\}\s*\)\s*=>\s*\{([\s\S]*)\}\s*\);/);
698
+ if (!testMatch) {
699
+ throw new Error('Could not extract test function from script');
700
+ }
701
+
702
+ const testBody = testMatch[1];
703
+ // Execute the entire test body as one async function
704
+ const executeTest = new Function('page', 'expect', `return (async () => { ${testBody} })()`);
705
+ await executeTest(page, expect);
706
+ }
707
+
708
+ private async getEnhancedPageInfo(page: Page): Promise<PageInfo> {
709
+ try {
710
+ return await getEnhancedPageInfo(page);
711
+ } catch (error) {
712
+ return {
713
+ url: page.url(),
714
+ title: 'Unknown',
715
+ elements: 'Unable to extract',
716
+ formFields: 'Unable to extract',
717
+ interactiveElements: 'Unable to extract',
718
+ pageStructure: 'Unable to extract'
719
+ };
720
+ }
721
+ }
722
+
723
+ private buildFailureHistory(
724
+ repairHistory: Array<{ attempt: number; action: StepRepairAction; error: string; pageInfo: PageInfo }>,
725
+ originalStep: ScriptStep,
726
+ originalError: any
727
+ ): string {
728
+ if (repairHistory.length === 0) {
729
+ return `Original failure: ${this.safeSerializeError(originalError)}`;
730
+ }
731
+
732
+ let history = `Original failure: ${this.safeSerializeError(originalError)}\n\n`;
733
+ history += `Previous repair attempts:\n`;
734
+
735
+ repairHistory.forEach((attempt, index) => {
736
+ history += `Attempt ${attempt.attempt}:\n`;
737
+ history += ` Operation: ${attempt.action.operation}\n`;
738
+ if (attempt.action.newStep) {
739
+ history += ` Description: ${attempt.action.newStep.description}\n`;
740
+ history += ` Code: ${attempt.action.newStep.code}\n`;
741
+ }
742
+ history += ` Error: ${attempt.error}\n`;
743
+ if (index < repairHistory.length - 1) {
744
+ history += `\n`;
745
+ }
746
+ });
747
+
748
+ return history;
749
+ }
750
+
751
+ private buildRecentRepairsContext(
752
+ recentRepairs: Array<{
753
+ stepNumber: number;
754
+ operation: string;
755
+ originalDescription?: string;
756
+ newDescription?: string;
757
+ originalCode?: string;
758
+ newCode?: string;
759
+ }>
760
+ ): string {
761
+ if (recentRepairs.length === 0) {
762
+ return 'No recent repairs to consider.';
763
+ }
764
+
765
+ let context = 'Recent successful repairs that may affect this step:\n\n';
766
+
767
+ recentRepairs.forEach((repair, index) => {
768
+ context += `Step ${repair.stepNumber} was successfully repaired:\n`;
769
+ context += ` Operation: ${repair.operation}\n`;
770
+
771
+ if (repair.operation === 'REMOVE') {
772
+ context += ` Removed: "${repair.originalDescription}"\n`;
773
+ context += ` Code removed:\n ${repair.originalCode?.replace(/\n/g, '\n ')}\n`;
774
+ } else if (repair.operation === 'INSERT') {
775
+ context += ` Inserted: "${repair.newDescription}"\n`;
776
+ context += ` Code inserted:\n ${repair.newCode?.replace(/\n/g, '\n ')}\n`;
777
+ } else {
778
+ context += ` Original: "${repair.originalDescription}"\n`;
779
+ context += ` Repaired: "${repair.newDescription}"\n`;
780
+ context += ` Code changed from:\n ${repair.originalCode?.replace(/\n/g, '\n ')}\n`;
781
+ context += ` To:\n ${repair.newCode?.replace(/\n/g, '\n ')}\n`;
782
+ }
783
+
784
+ if (index < recentRepairs.length - 1) {
785
+ context += `\n`;
786
+ }
787
+ });
788
+
789
+ context += '\nConsider how these changes might affect the current step and adjust accordingly.';
790
+ return context;
791
+ }
792
+
793
+ private async applyRepairActionInContext(
794
+ action: StepRepairAction,
795
+ steps: (ScriptStep & { success?: boolean; error?: string })[],
796
+ currentIndex: number,
797
+ page: Page,
798
+ executionContext: string,
799
+ contextVariables: Map<string, any>
800
+ ): Promise<{ success: boolean; error?: string; updatedContext?: string }> {
801
+ try {
802
+ switch (action.operation) {
803
+ case StepOperation.MODIFY:
804
+ if (action.newStep && action.stepIndex !== undefined) {
805
+ // Modify existing step
806
+ steps[action.stepIndex] = {
807
+ ...action.newStep,
808
+ success: false,
809
+ error: undefined
810
+ };
811
+ // Test the modified step with current page state and variables
812
+ await this.executeStepCode(action.newStep.code, page);
813
+ return { success: true, updatedContext: executionContext + action.newStep.code };
814
+ }
815
+ break;
816
+
817
+ case StepOperation.INSERT:
818
+ if (action.newStep && action.insertAfterIndex !== undefined) {
819
+ // Insert new step after specified index
820
+ const insertIndex = action.insertAfterIndex + 1;
821
+ const newStep = {
822
+ ...action.newStep,
823
+ success: false,
824
+ error: undefined
825
+ };
826
+ console.log(`INSERT: Inserting step at index ${insertIndex} with description "${newStep.description}"`);
827
+ console.log(`INSERT: Steps before insertion:`, steps.map((s, i) => `${i}: "${s.description}" (success: ${s.success})`));
828
+
829
+ // Preserve success status of existing steps before insertion
830
+ const successStatusMap = new Map(steps.map((step, index) => [index, { success: step.success, error: step.error }]));
831
+
832
+ steps.splice(insertIndex, 0, newStep);
833
+
834
+ // Restore success status for steps that were shifted by the insertion
835
+ // Steps at insertIndex and before keep their original status
836
+ // Steps after insertIndex need to be shifted to their new positions
837
+ for (let i = insertIndex + 1; i < steps.length; i++) {
838
+ const originalIndex = i - 1; // The step that was originally at this position
839
+ if (successStatusMap.has(originalIndex)) {
840
+ const status = successStatusMap.get(originalIndex)!;
841
+ steps[i].success = status.success;
842
+ steps[i].error = status.error;
843
+ }
844
+ }
845
+
846
+ // CRITICAL FIX: Ensure the inserted step doesn't overwrite existing step data
847
+ // The new step should only have its own description, not inherit from existing steps
848
+ console.log(`INSERT: Final step array after restoration:`, steps.map((s, i) => `${i}: "${s.description}" (success: ${s.success})`));
849
+
850
+ console.log(`INSERT: Steps after insertion:`, steps.map((s, i) => `${i}: "${s.description}" (success: ${s.success})`));
851
+ // Test the new step with current page state
852
+ await this.executeStepCode(action.newStep.code, page);
853
+ return { success: true, updatedContext: executionContext + action.newStep.code };
854
+ }
855
+ break;
856
+
857
+ case StepOperation.REMOVE:
858
+ if (action.stepIndex !== undefined) {
859
+ // Remove step
860
+ steps.splice(action.stepIndex, 1);
861
+ return { success: true, updatedContext: executionContext };
862
+ }
863
+ break;
864
+ }
865
+
866
+ return { success: false, error: 'Invalid repair action' };
867
+ } catch (error) {
868
+ return {
869
+ success: false,
870
+ error: error instanceof Error ? error.message : 'Unknown error during repair action'
871
+ };
872
+ }
873
+ }
874
+
875
+ private async applyRepairAction(
876
+ action: StepRepairAction,
877
+ steps: (ScriptStep & { success?: boolean; error?: string })[],
878
+ currentIndex: number,
879
+ page: Page
880
+ ): Promise<{ success: boolean; error?: string }> {
881
+ try {
882
+ switch (action.operation) {
883
+ case StepOperation.MODIFY:
884
+ if (action.newStep && action.stepIndex !== undefined) {
885
+ // Modify existing step
886
+ steps[action.stepIndex] = {
887
+ ...action.newStep,
888
+ success: false,
889
+ error: undefined
890
+ };
891
+ // Test the modified step
892
+ await this.executeStepCode(action.newStep.code, page);
893
+ return { success: true };
894
+ }
895
+ break;
896
+
897
+ case StepOperation.INSERT:
898
+ if (action.newStep && action.insertAfterIndex !== undefined) {
899
+ // Insert new step after specified index
900
+ const insertIndex = action.insertAfterIndex + 1;
901
+ const newStep = {
902
+ ...action.newStep,
903
+ success: false,
904
+ error: undefined
905
+ };
906
+ steps.splice(insertIndex, 0, newStep);
907
+ // Test the inserted step
908
+ await this.executeStepCode(action.newStep.code, page);
909
+ return { success: true };
910
+ }
911
+ break;
912
+
913
+ case StepOperation.REMOVE:
914
+ if (action.stepIndex !== undefined) {
915
+ // Remove the step
916
+ steps.splice(action.stepIndex, 1);
917
+ return { success: true };
918
+ }
919
+ break;
920
+ }
921
+
922
+ return { success: false, error: 'Invalid repair action' };
923
+ } catch (error) {
924
+ return {
925
+ success: false,
926
+ error: error instanceof Error ? error.message : 'Repair action execution failed'
927
+ };
928
+ }
929
+ }
930
+
931
+
932
+
933
+ private generateUpdatedScript(steps: (ScriptStep & { success?: boolean; error?: string })[], repairAdvice?: string): string {
934
+ const scriptLines = [
935
+ "import { test, expect } from '@playwright/test';",
936
+ `test('repairedTest', async ({ page, browser, context }) => {`
937
+ ];
938
+
939
+ steps.forEach((step, index) => {
940
+ scriptLines.push(` // Step ${index + 1}: ${step.description}`);
941
+ const codeLines = step.code.split('\n');
942
+ codeLines.forEach(line => {
943
+ scriptLines.push(` ${line}`);
944
+ });
945
+ });
946
+
947
+ scriptLines.push('});');
948
+ const script = scriptLines.join('\n');
949
+
950
+ // Add TestChimp comment to the repaired script with repair advice
951
+ return addTestChimpComment(script, repairAdvice);
952
+ }
953
+
954
+
955
+ /**
956
+ * Initialize browser with configuration (delegates to utility function)
957
+ */
958
+ private async initializeBrowser(playwrightConfig?: string, headless?: boolean, playwrightConfigFilePath?: string): Promise<{ browser: Browser; context: BrowserContext; page: Page }> {
959
+ return initializeBrowser(playwrightConfig, headless, playwrightConfigFilePath);
960
+ }
961
+
962
+ /**
963
+ * Safely serialize error information, filtering out non-serializable values
964
+ */
965
+ private safeSerializeError(error: any): string {
966
+ try {
967
+ if (error instanceof Error) {
968
+ return error.message;
969
+ }
970
+
971
+ if (typeof error === 'string') {
972
+ return error;
973
+ }
974
+
975
+ if (typeof error === 'object' && error !== null) {
976
+ // Try to extract meaningful information without serializing the entire object
977
+ const safeError: any = {};
978
+
979
+ // Copy safe properties
980
+ if (error.message) safeError.message = error.message;
981
+ if (error.name) safeError.name = error.name;
982
+ if (error.code) safeError.code = error.code;
983
+ if (error.status) safeError.status = error.status;
984
+
985
+ // Try to get stack trace safely
986
+ if (error.stack && typeof error.stack === 'string') {
987
+ safeError.stack = error.stack;
988
+ }
989
+
990
+ return JSON.stringify(safeError);
991
+ }
992
+
993
+ return String(error);
994
+ } catch (serializationError) {
995
+ // If even safe serialization fails, return a basic string representation
996
+ return `Error: ${String(error)}`;
997
+ }
998
+ }
999
+ }