@yasserkhanorg/e2e-agents 0.3.2

package/dist/e2e-test-gen/spec_parser.js

@@ -0,0 +1,786 @@
+"use strict";
+// Copyright (c) 2015-present Mattermost, Inc. All Rights Reserved.
+// See LICENSE.txt for license information.
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.SpecificationParser = void 0;
+const fs_1 = require("fs");
+const path_1 = require("path");
+const crypto_1 = require("crypto");
+/**
+ * Specification Parser
+ *
+ * Parses user-provided specification documents to guide autonomous testing.
+ * Supports multiple formats:
+ * - Markdown (.md) with embedded screenshots and Given-When-Then scenarios
+ * - JSON (.json) with structured feature definitions
+ * - PDF (.pdf) with text, diagrams, and screenshots extracted via LLM
+ * - Plain text focus strings (natural language)
+ *
+ * Extracts:
+ * - Feature name, description, priority
+ * - Target URLs to prioritize in crawling
+ * - Business scenarios (Given-When-Then)
+ * - Acceptance criteria
+ * - Reference screenshots for visual comparison
+ * - UI mockups and wireframes from PDFs
+ *
+ * The parsed specifications guide:
+ * - Crawler URL prioritization
+ * - Test scenario generation
+ * - Visual regression comparison
+ * - Coverage gap detection
+ */
+class SpecificationParser {
+    constructor(llmProvider, cache) {
+        this.llmProvider = llmProvider;
+        this.cache = cache;
+    }
+    /**
+     * Parse specification from file or string
+     */
+    async parse(source, sourceType = 'file') {
+        let content;
+        let sourcePath;
+        if (sourceType === 'file') {
+            if (!(0, fs_1.existsSync)(source)) {
+                throw new Error(`Specification file not found: ${source}`);
+            }
+            sourcePath = source;
+            // Determine format from extension
+            if (source.endsWith('.md')) {
+                content = (0, fs_1.readFileSync)(source, 'utf-8');
+                return this.parseMarkdown(content, sourcePath);
+            }
+            else if (source.endsWith('.json')) {
+                content = (0, fs_1.readFileSync)(source, 'utf-8');
+                return this.parseJSON(content, sourcePath);
+            }
+            else if (source.endsWith('.pdf')) {
+                return this.parsePDF(sourcePath);
+            }
+            else {
+                throw new Error(`Unsupported specification format: ${source}. Use .md, .json, or .pdf`);
+            }
+        }
+        else {
+            // Plain text focus string - use LLM to interpret
+            return this.parseFocusString(source);
+        }
+    }
+    /**
+     * Parse Markdown specification
+     *
+     * Expected format:
+     * # Feature: Feature Name
+     * **Priority**: High
+     * **Target URLs**: /path1, /path2
+     *
+     * ## Description
+     * Feature description...
+     *
+     * ## Business Scenarios
+     * ### Scenario 1: Name
+     * - **Given**: Precondition
+     * - **When**: Action
+     * - **Then**: Expected outcome
+     *
+     * ## Acceptance Criteria
+     * - Criterion 1
+     * - Criterion 2
+     *
+     * ## Screenshots
+     * ![Description](path/to/image.png)
+     */
+    async parseMarkdown(content, sourcePath) {
+        // Dynamic import to handle ESM marked module in CommonJS context
+        const { marked } = await import('marked');
+        const tokens = marked.lexer(content);
+        const specs = [];
+        let currentSpec = null;
+        let currentSection = null;
+        let currentScenario = null;
+        for (let i = 0; i < tokens.length; i++) {
+            const token = tokens[i];
+            if (token.type === 'heading') {
+                const headingToken = token;
+                const text = headingToken.text;
+                if (headingToken.depth === 1) {
+                    // New feature
+                    if (currentSpec) {
+                        specs.push(this.finalizeSpec(currentSpec, sourcePath));
+                    }
+                    currentSpec = {
+                        name: text.replace(/^Feature:\s*/i, '').trim(),
+                        targetUrls: [],
+                        scenarios: [],
+                        screenshots: [],
+                        acceptanceCriteria: [],
+                    };
+                    currentSection = null;
+                }
+                else if (headingToken.depth === 2) {
+                    // Section
+                    currentSection = text.toLowerCase();
+                }
+                else if (headingToken.depth === 3 && currentSection === 'business scenarios') {
+                    // New scenario
+                    if (currentScenario && currentSpec) {
+                        currentSpec.scenarios.push(this.finalizeScenario(currentScenario));
+                    }
+                    currentScenario = {
+                        name: text.replace(/^Scenario \d+:\s*/i, '').trim(),
+                        priority: 'should-have',
+                    };
+                }
+            }
+            else if (token.type === 'paragraph' && currentSpec) {
+                const paragraphToken = token;
+                const text = paragraphToken.text;
+                // Extract metadata from bold markers
+                const priorityMatch = text.match(/\*\*Priority\*\*:\s*(\w+)/i);
+                if (priorityMatch) {
+                    currentSpec.priority = priorityMatch[1].toLowerCase();
+                }
+                const urlsMatch = text.match(/\*\*Target URLs?\*\*:\s*(.+)/i);
+                if (urlsMatch) {
+                    currentSpec.targetUrls = urlsMatch[1]
+                        .split(',')
+                        .map((url) => url.trim())
+                        .filter(Boolean);
+                }
+                // Extract scenario details
+                if (currentScenario) {
+                    const givenMatch = text.match(/\*\*Given\*\*:\s*(.+)/i);
+                    const whenMatch = text.match(/\*\*When\*\*:\s*(.+)/i);
+                    const thenMatch = text.match(/\*\*Then\*\*:\s*(.+)/i);
+                    const priorityMatch = text.match(/\*\*Priority\*\*:\s*(\w+)/i);
+                    if (givenMatch)
+                        currentScenario.given = givenMatch[1].trim();
+                    if (whenMatch)
+                        currentScenario.when = whenMatch[1].trim();
+                    if (thenMatch)
+                        currentScenario.then = thenMatch[1].trim();
+                    if (priorityMatch) {
+                        const priority = priorityMatch[1].toLowerCase();
+                        if (priority.includes('must'))
+                            currentScenario.priority = 'must-have';
+                        else if (priority.includes('should'))
+                            currentScenario.priority = 'should-have';
+                        else
+                            currentScenario.priority = 'nice-to-have';
+                    }
+                }
+                // Description section
+                if (currentSection === 'description') {
+                    currentSpec.description = (currentSpec.description || '') + text + ' ';
+                }
+            }
+            else if (token.type === 'list' && currentSpec) {
+                const listToken = token;
+                if (currentSection === 'acceptance criteria') {
+                    // Extract acceptance criteria
+                    for (const item of listToken.items) {
+                        const itemToken = item;
+                        currentSpec.acceptanceCriteria.push(itemToken.text);
+                    }
+                }
+                else if (currentSection === 'business scenarios' && currentScenario) {
+                    // Extract Given-When-Then from list items
+                    for (const item of listToken.items) {
+                        const itemToken = item;
+                        const text = itemToken.text;
+                        const givenMatch = text.match(/\*\*Given\*\*:\s*(.+)/i);
+                        const whenMatch = text.match(/\*\*When\*\*:\s*(.+)/i);
+                        const thenMatch = text.match(/\*\*Then\*\*:\s*(.+)/i);
+                        if (givenMatch)
+                            currentScenario.given = givenMatch[1].trim();
+                        if (whenMatch)
+                            currentScenario.when = whenMatch[1].trim();
+                        if (thenMatch)
+                            currentScenario.then = thenMatch[1].trim();
+                    }
+                }
+            }
+            else if (token.type === 'image' && currentSpec && currentSection === 'screenshots') {
+                const imageToken = token;
+                // Resolve image path relative to spec file
+                const imagePath = this.resolveImagePath(imageToken.href, sourcePath);
+                currentSpec.screenshots.push({
+                    path: imagePath,
+                    description: imageToken.text || 'Screenshot',
+                });
+            }
+        }
+        // Finalize last scenario and spec
+        if (currentScenario && currentSpec) {
+            currentSpec.scenarios.push(this.finalizeScenario(currentScenario));
+        }
+        if (currentSpec) {
+            specs.push(this.finalizeSpec(currentSpec, sourcePath));
+        }
+        // Load screenshot data
+        await this.loadScreenshots(specs);
+        return specs;
+    }
+    /**
+     * Parse JSON specification
+     */
+    parseJSON(content, sourcePath) {
+        try {
+            let data;
+            try {
+                data = JSON.parse(content);
+            }
+            catch (parseError) {
+                throw new Error(`Invalid JSON syntax: ${parseError.message}`);
+            }
+            if (!data) {
+                throw new Error('JSON is empty or null');
+            }
+            // Support both single spec and array of specs
+            const specsData = Array.isArray(data) ? data : [data];
+            if (specsData.length === 0) {
+                throw new Error('No specifications found in JSON');
+            }
+            const specs = specsData.map((specData, index) => {
+                // Validate required fields
+                if (!specData.name && !specData.feature) {
+                    throw new Error(`Specification ${index + 1} missing required "name" or "feature" field`);
+                }
+                if (!specData.scenarios || !Array.isArray(specData.scenarios)) {
+                    // eslint-disable-next-line no-console
+                    console.warn(`Specification "${specData.name || specData.feature}" has no scenarios`);
+                }
+                const scenarios = (specData.scenarios || []).map((s, sIndex) => {
+                    if (!s.name) {
+                        throw new Error(`Scenario ${sIndex + 1} in "${specData.name || specData.feature}" missing "name" field`);
+                    }
+                    return {
+                        name: s.name,
+                        given: s.given || '',
+                        when: s.when || '',
+                        then: s.then || '',
+                        priority: s.priority || 'should-have',
+                    };
+                });
+                const screenshots = (specData.screenshots || []).map((s) => {
+                    if (typeof s === 'string') {
+                        return { path: this.resolveImagePath(s, sourcePath), description: 'Screenshot' };
+                    }
+                    return {
+                        path: this.resolveImagePath(s.path, sourcePath),
+                        description: s.description || 'Screenshot',
+                    };
+                });
+                return {
+                    id: this.generateSpecId(specData.feature || specData.name),
+                    name: specData.feature || specData.name,
+                    description: specData.description || '',
+                    priority: specData.priority || 'medium',
+                    targetUrls: specData.targetUrls || specData.urls || [],
+                    scenarios,
+                    screenshots,
+                    acceptanceCriteria: specData.acceptanceCriteria || [],
+                    sourcePath,
+                };
+            });
+            // Load screenshot data
+            this.loadScreenshots(specs);
+            return specs;
+        }
+        catch (error) {
+            throw new Error(`Failed to parse JSON specification: ${error instanceof Error ? error.message : String(error)}`);
+        }
+    }
+    /**
+     * Parse PDF specification using LLM vision
+     *
+     * PDFs can contain:
+     * - Product requirement documents (PRDs)
+     * - Feature specifications with screenshots
+     * - UI mockups and wireframes
+     * - User flow diagrams
+     * - Acceptance criteria and test plans
+     *
+     * The LLM will extract structured information including:
+     * - Feature name and description
+     * - Business scenarios
+     * - Target URLs (inferred from screenshots/mockups)
+     * - Acceptance criteria
+     * - Screenshots/mockups for visual comparison
+     */
+    async parsePDF(pdfPath) {
+        /* eslint-disable no-console */
+        // Console output is expected for PDF parsing progress
+        console.log(`📄 Parsing PDF specification: ${pdfPath}`);
+        // Read PDF file first to calculate hash
+        const pdfBuffer = (0, fs_1.readFileSync)(pdfPath);
+        const pdfHash = (0, crypto_1.createHash)('sha256').update(pdfBuffer).digest('hex');
+        // Check if already cached
+        if (this.cache && this.cache.isSpecificationCached(pdfPath, pdfHash)) {
+            console.log(`   ✓ Using cached PDF specification (hash: ${pdfHash.substring(0, 8)}...)`);
+            const cachedSpecs = this.cache.getCachedSpecifications(pdfPath, pdfHash);
+            console.log(`✅ Loaded ${cachedSpecs.length} feature(s) from cache`);
+            // Print scenario count for each feature
+            for (const spec of cachedSpecs) {
+                const priority = spec.priority === 'critical' ? 'critical' : spec.priority;
+                console.log(`   ✓ Loaded 1 specifications`);
+                console.log(`     - ${spec.name} (${priority}): ${spec.scenarios.length} scenarios`);
+            }
+            return cachedSpecs;
+        }
+        // Check if provider supports vision (required for PDF parsing)
+        if (!this.llmProvider.capabilities.vision) {
+            throw new Error('PDF parsing requires a vision-capable LLM provider (e.g., Anthropic Claude). ' +
+                'Current provider does not support vision. ' +
+                'Use --llm-provider anthropic or --llm-provider hybrid');
+        }
+        // SECURITY WARNING: PDF will be sent to external LLM
+        console.warn('');
+        console.warn('⚠️  SECURITY WARNING: PDF Document Transmission ⚠️');
+        console.warn('━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━');
+        console.warn('The PDF specification will be sent to an external LLM provider.');
+        console.warn('');
+        console.warn('Ensure the PDF does NOT contain:');
+        console.warn('  • Internal API keys or credentials');
+        console.warn('  • Sensitive architecture details');
+        console.warn('  • Production URLs or IP addresses');
+        console.warn('  • Personal Identifying Information (PII)');
+        console.warn('  • Proprietary business information');
+        console.warn('');
+        console.warn('The LLM provider may retain this data per their policy.');
+        console.warn('For Anthropic: https://www.anthropic.com/legal/privacy');
+        console.warn('━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━');
+        console.warn('');
+        // Check if user has consented (via environment variable)
+        if (!process.env.AUTONOMOUS_ALLOW_PDF_UPLOAD) {
+            console.error('❌ PDF upload blocked for security.');
+            console.error('❌ To proceed, set environment variable:');
+            console.error('❌   AUTONOMOUS_ALLOW_PDF_UPLOAD=true');
+            console.error('❌ Only set this if you have reviewed the PDF for sensitive data.');
+            throw new Error('PDF upload requires explicit consent via AUTONOMOUS_ALLOW_PDF_UPLOAD=true');
+        }
+        try {
+            console.log(`📄 Extracting text from PDF... (${(pdfBuffer.length / 1024).toFixed(2)} KB)`);
+            // Parse PDF to extract text content (dynamic import with default export)
+            const pdfParse = (await import('pdf-parse')).default;
+            const pdfData = await pdfParse(pdfBuffer);
+            const pdfText = pdfData.text;
+            console.log(`   ✓ Extracted ${pdfText.length} characters, ${pdfData.numpages} pages`);
+            // Use LLM for semantic parsing (text-only, no vision)
+            const prompt = `
+You are a test specification parser. Parse this UX/product specification document and extract structured, machine-operable testing information.
+
+The document describes features, user flows, permissions, edge cases, and acceptance criteria.
+
+Extract the following in JSON format (this schema is optimized for test automation):
+
+{
+  "features": [
+    {
+      "name": "Feature Name (e.g., Auto-translation)",
+      "scope": "MVP|v1|v2",
+      "description": "What this feature does",
+      "priority": "critical|high|medium|low",
+
+      "roles": {
+        "role_name": {
+          "permissions": ["permission1", "permission2"]
+        }
+      },
+
+      "enablement": {
+        "location_type": {
+          "enabled_by": "who can enable",
+          "default_state": "on|off",
+          "side_effects": ["system_message", "label_visible"]
+        }
+      },
+
+      "scenarios": [
+        {
+          "name": "Scenario name",
+          "given": "Precondition",
+          "when": "Action",
+          "then": "Expected outcome",
+          "priority": "must-have|should-have|nice-to-have"
+        }
+      ],
+
+      "acceptanceCriteria": [
+        "Testable criterion"
+      ],
+
+      "stateMachines": {
+        "message_lifecycle": {
+          "states": ["translating", "translated", "failed"],
+          "transitions": ["new_message -> translating", "translating -> translated"]
+        }
+      },
+
+      "uiIndicators": {
+        "location": "visual_indicator"
+      },
+
+      "edgeCases": [
+        "What happens when X"
+      ],
+
+      "platformDifferences": {
+        "desktop": "behavior",
+        "mobile": "behavior"
+      },
+
+      "nonGoals": [
+        "Out of scope items"
+      ]
+    }
+  ]
+}
+
+IMPORTANT:
+- Extract ALL features, flows, roles, and permissions
+- Map state machines if described
+- Identify edge cases and failure modes
+- List platform differences
+- Capture non-goals/out-of-scope items
+- DO NOT infer or generate target URLs - feature specs describe behavior, not routes
+- Be precise with permissions and access control
+- For MVP features or P0 features, set priority to "critical"
+- For must-have scenarios in MVP features, set scenario priority to "must-have"
+
+Respond with ONLY valid JSON, no markdown formatting.
+
+DOCUMENT TEXT:
+${pdfText}
+            `.trim();
+            // Call LLM with text (no vision)
+            // Use higher maxTokens for complex documents (16K for large specs)
+            const response = await this.llmProvider.generateText(prompt, {
+                maxTokens: 16000,
+                temperature: 0.1, // Very low temperature for structured extraction
+            });
+            // Parse LLM response
+            let jsonText = response.text.trim();
+            // Remove markdown code blocks if present (handle various formats)
+            // Try multiple patterns to be more robust
+            const jsonMatch = jsonText.match(/```(?:json)?\s*([\s\S]+?)\s*```/);
+            if (jsonMatch) {
+                jsonText = jsonMatch[1].trim();
+            }
+            else if (jsonText.startsWith('```')) {
+                // Fallback: remove all triple backticks
+                jsonText = jsonText
+                    .replace(/```[^\n]*\n?/g, '')
+                    .replace(/```\s*$/g, '')
+                    .trim();
+            }
+            let data;
+            try {
+                data = JSON.parse(jsonText);
+            }
+            catch (parseError) {
+                // Log the problematic JSON for debugging
+                console.error('Failed to parse LLM response as JSON');
+                console.error('First 500 chars:', jsonText.substring(0, 500));
+                console.error('Last 500 chars:', jsonText.substring(Math.max(0, jsonText.length - 500)));
+                // Try to repair common JSON issues
+                console.log('Attempting to repair JSON...');
+                let repairedJson = jsonText;
+                // Common fixes:
+                // 1. Fix unquoted property names (e.g., {name: "value"} -> {"name": "value"})
+                repairedJson = repairedJson.replace(/([{,]\s*)([a-zA-Z_][a-zA-Z0-9_]*)\s*:/g, '$1"$2":');
+                // 2. Fix single quotes to double quotes
+                repairedJson = repairedJson.replace(/'/g, '"');
+                // 3. Remove trailing commas
+                repairedJson = repairedJson.replace(/,(\s*[}\]])/g, '$1');
+                // 4. Fix missing quotes around string values (this is tricky, skip for now)
+                try {
+                    data = JSON.parse(repairedJson);
+                    console.log('✓ JSON repair successful!');
+                }
+                catch {
+                    // If repair failed, ask LLM to fix it
+                    console.log('JSON repair failed, requesting LLM to fix the JSON...');
+                    const fixPrompt = `
+The following JSON is malformed. Please fix it and return ONLY valid JSON with no markdown formatting:
+
+${jsonText}
+
+Error: ${parseError instanceof Error ? parseError.message : String(parseError)}
+
+Return the corrected JSON:`.trim();
+                    const fixResponse = await this.llmProvider.generateText(fixPrompt, {
+                        maxTokens: 16000,
+                        temperature: 0,
+                    });
+                    let fixedJson = fixResponse.text.trim();
+                    // Remove markdown code blocks again
+                    const fixedMatch = fixedJson.match(/```(?:json)?\s*([\s\S]+?)\s*```/);
+                    if (fixedMatch) {
+                        fixedJson = fixedMatch[1].trim();
+                    }
+                    else if (fixedJson.startsWith('```')) {
+                        fixedJson = fixedJson
+                            .replace(/```[^\n]*\n?/g, '')
+                            .replace(/```\s*$/g, '')
+                            .trim();
+                    }
+                    try {
+                        data = JSON.parse(fixedJson);
+                        console.log('✓ LLM successfully fixed the JSON!');
+                    }
+                    catch (finalError) {
+                        console.error('Failed to parse even after LLM repair');
+                        console.error('Original error:', parseError);
+                        console.error('Repair error:', finalError);
+                        throw parseError; // Throw original error
+                    }
+                }
+            }
+            if (!data.features || !Array.isArray(data.features)) {
+                throw new Error('Invalid PDF parse response: missing features array');
+            }
+            console.log(`✅ Extracted ${data.features.length} feature(s) from PDF`);
+            // Convert to FeatureSpecification format
+            const specs = data.features.map((feature) => {
+                const scenarios = (feature.scenarios || []).map((s) => ({
+                    name: s.name,
+                    given: s.given || '',
+                    when: s.when || '',
+                    then: s.then || '',
+                    priority: s.priority || 'should-have',
+                }));
+                // Store screenshot metadata (page numbers for extraction)
+                const screenshots = (feature.screenshots || []).map((s) => ({
+                    path: `${pdfPath}#page=${s.pageNumber}`,
+                    description: s.description || `Page ${s.pageNumber}`,
+                    pageNumber: s.pageNumber,
+                }));
+                const spec = {
+                    id: this.generateSpecId(feature.name),
+                    name: feature.name,
+                    description: feature.description || '',
+                    priority: feature.priority || 'medium',
+                    targetUrls: feature.targetUrls || [],
+                    scenarios,
+                    screenshots,
+                    acceptanceCriteria: feature.acceptanceCriteria || [],
+                    sourcePath: pdfPath,
+                    sourceHash: pdfHash,
+                    metadata: {
+                        uiElements: feature.uiElements || [],
+                        userFlows: feature.userFlows || [],
+                        extractedFrom: 'pdf',
+                    },
+                };
+                return spec;
+            });
+            // Save to cache if knowledge base is available
+            if (this.cache) {
+                for (const spec of specs) {
+                    this.cache.saveSpecification(spec);
+                }
+            }
+            /* eslint-enable no-console */
+            return specs;
+        }
+        catch (error) {
+            if (error instanceof Error && error.message.includes('vision')) {
+                throw error; // Re-throw vision capability errors
+            }
+            throw new Error(`Failed to parse PDF specification: ${error instanceof Error ? error.message : String(error)}. ` +
+                'Ensure the PDF contains readable text and images. ' +
+                'For scanned PDFs, ensure they have been OCR processed.');
+        }
+    }
+    /**
+     * Parse natural language focus string using LLM
+     */
+    async parseFocusString(focusString) {
+        const prompt = `
+You are a test specification generator. Given a natural language focus string, extract:
+1. Feature name (infer from the focus string)
+2. Target URLs (guess likely URLs based on feature names)
+3. Business scenarios (generate 2-3 test scenarios)
+4. Priority (infer from words like "critical", "thoroughly", etc.)
+
+Focus string: "${focusString}"
+
+Respond with valid JSON in this format:
+{
+  "feature": "Feature Name",
+  "priority": "high|medium|low",
+  "targetUrls": ["/url1", "/url2"],
+  "scenarios": [
+    {
+      "name": "Scenario name",
+      "given": "Precondition",
+      "when": "Action",
+      "then": "Expected result",
+      "priority": "must-have|should-have|nice-to-have"
+    }
+  ],
+  "acceptanceCriteria": ["criterion 1", "criterion 2"]
+}
+        `.trim();
+        try {
+            const response = await this.llmProvider.generateText(prompt, {
+                maxTokens: 1000,
+                temperature: 0.3,
+            });
+            // Extract JSON from response (might have markdown code blocks)
+            let jsonText = response.text.trim();
+            // Try multiple patterns to be more robust
+            const jsonMatch = jsonText.match(/```(?:json)?\s*([\s\S]+?)\s*```/);
+            if (jsonMatch) {
+                jsonText = jsonMatch[1].trim();
+            }
+            else if (jsonText.startsWith('```')) {
+                // Fallback: remove all triple backticks
+                jsonText = jsonText
+                    .replace(/```[^\n]*\n?/g, '')
+                    .replace(/```\s*$/g, '')
+                    .trim();
+            }
+            const data = JSON.parse(jsonText);
+            const scenarios = (data.scenarios || []).map((s) => ({
+                name: s.name,
+                given: s.given,
+                when: s.when,
+                then: s.then,
+                priority: s.priority || 'should-have',
+            }));
+            return [
+                {
+                    id: this.generateSpecId(data.feature),
+                    name: data.feature,
+                    description: `Generated from focus string: ${focusString}`,
+                    priority: data.priority || 'medium',
+                    targetUrls: data.targetUrls || [],
+                    scenarios,
+                    screenshots: [],
+                    acceptanceCriteria: data.acceptanceCriteria || [],
+                    sourcePath: 'focus-string',
+                },
+            ];
+        }
+        catch (error) {
+            throw new Error(`Failed to parse focus string with LLM: ${error instanceof Error ? error.message : String(error)}`);
+        }
+    }
+    /**
+     * Finalize partial spec
+     */
+    finalizeSpec(partial, sourcePath) {
+        return {
+            id: this.generateSpecId(partial.name || 'unknown'),
+            name: partial.name || 'Unnamed Feature',
+            description: (partial.description || '').trim(),
+            priority: partial.priority || 'medium',
+            targetUrls: partial.targetUrls || [],
+            scenarios: partial.scenarios || [],
+            screenshots: partial.screenshots || [],
+            acceptanceCriteria: partial.acceptanceCriteria || [],
+            sourcePath,
+        };
+    }
+    /**
+     * Finalize partial scenario
+     */
+    finalizeScenario(partial) {
+        return {
+            name: partial.name || 'Unnamed Scenario',
+            given: partial.given || '',
+            when: partial.when || '',
+            then: partial.then || '',
+            priority: partial.priority || 'should-have',
+        };
+    }
+    /**
+     * Generate spec ID from name
+     */
+    generateSpecId(name) {
+        return name
+            .toLowerCase()
+            .replace(/[^a-z0-9]+/g, '-')
+            .replace(/^-|-$/g, '');
+    }
+    /**
+     * Resolve image path relative to spec file
+     */
+    resolveImagePath(imagePath, specPath) {
+        if (imagePath.startsWith('http://') || imagePath.startsWith('https://')) {
+            return imagePath; // Absolute URL
+        }
+        if (imagePath.startsWith('/')) {
+            return imagePath; // Absolute path
+        }
+        // Relative path - resolve relative to spec file
+        const specDir = (0, path_1.dirname)(specPath);
+        return (0, path_1.join)(specDir, imagePath);
+    }
+    /**
+     * Load screenshot data from files
+     */
+    async loadScreenshots(specs) {
+        for (const spec of specs) {
+            for (const screenshot of spec.screenshots) {
+                try {
+                    if ((0, fs_1.existsSync)(screenshot.path)) {
+                        const imageData = (0, fs_1.readFileSync)(screenshot.path);
+                        screenshot.data = imageData.toString('base64');
+                    }
+                    else {
+                        // eslint-disable-next-line no-console
+                        console.warn(`Screenshot not found: ${screenshot.path}`);
+                    }
+                }
+                catch (error) {
+                    // eslint-disable-next-line no-console
+                    console.warn(`Failed to load screenshot ${screenshot.path}:`, error);
+                }
+            }
+        }
+    }
+    /**
+     * Validate parsed specification
+     */
+    validateSpec(spec) {
+        const errors = [];
+        if (!spec.name || spec.name.trim().length === 0) {
+            errors.push('Feature name is required');
+        }
+        if (spec.scenarios.length === 0) {
+            errors.push('At least one scenario is required');
+        }
+        for (let i = 0; i < spec.scenarios.length; i++) {
+            const scenario = spec.scenarios[i];
+            if (!scenario.given || !scenario.when || !scenario.then) {
+                errors.push(`Scenario ${i + 1} is incomplete (missing given/when/then)`);
+            }
+        }
+        return {
+            valid: errors.length === 0,
+            errors,
+        };
+    }
+    /**
+     * Get spec coverage summary
+     */
+    getSpecSummary(spec) {
+        return {
+            id: spec.id,
+            name: spec.name,
+            priority: spec.priority,
+            targetUrlCount: spec.targetUrls.length,
+            scenarioCount: spec.scenarios.length,
+            mustHaveScenarios: spec.scenarios.filter((s) => s.priority === 'must-have').length,
+            shouldHaveScenarios: spec.scenarios.filter((s) => s.priority === 'should-have').length,
+            niceToHaveScenarios: spec.scenarios.filter((s) => s.priority === 'nice-to-have').length,
+            acceptanceCriteriaCount: spec.acceptanceCriteria.length,
+            hasScreenshots: spec.screenshots.length > 0,
+        };
+    }
+}
+exports.SpecificationParser = SpecificationParser;