universal-agent-memory 0.7.2 → 0.7.3

package/dist/benchmarks/model-integration.js

@@ -0,0 +1,923 @@
+/**
+ * Model Integration Benchmark
+ *
+ * Runs real API calls against multiple LLM providers via Factory.ai droid exec CLI
+ * to compare model performance on UAM memory-enhanced tasks.
+ *
+ * Assumptions:
+ * - FACTORY_API_KEY is set in environment for Factory.ai API access
+ * - Models: Claude Opus 4.5, GLM 4.7, GPT 5.2
+ * - droid CLI is installed and accessible
+ * - UAM CLI is available for memory initialization
+ *
+ * What this handles:
+ * - Full UAM setup (init, analyze, generate, memory start, prepopulate)
+ * - CLAUDE.md reading and context injection
+ * - Real API calls to multiple LLM providers via droid exec
+ * - Task execution comparison across models with/without UAM
+ * - Performance metrics collection (latency, success, tokens)
+ * - Result aggregation and reporting
+ *
+ * What this does NOT handle:
+ * - Rate limiting (caller responsibility)
+ * - Cost tracking (would require billing API)
+ * - Streaming responses (uses completion mode)
+ */
+import { readFileSync, writeFileSync, existsSync } from 'fs';
+import { join, dirname } from 'path';
+import { fileURLToPath } from 'url';
+import { execSync } from 'child_process';
+const __filename = fileURLToPath(import.meta.url);
+const __dirname = dirname(__filename);
+const PROJECT_ROOT = join(__dirname, '../..');
+/**
+ * Initialize UAM system for benchmark testing
+ * Runs: uam init, uam analyze, uam generate, uam memory start, uam memory prepopulate
+ */
+async function setupUAM(verbose = false) {
+    const result = {
+        initialized: false,
+        memoryStarted: false,
+        memoryPrepopulated: false,
+        claudeMdLoaded: false,
+        errors: [],
+    };
+    const log = (msg) => { if (verbose)
+        console.log(`  [UAM Setup] ${msg}`); };
+    try {
+        // Step 1: Check if UAM CLI is available
+        log('Checking UAM CLI availability...');
+        try {
+            execSync('uam --version', { encoding: 'utf-8', cwd: PROJECT_ROOT, stdio: 'pipe' });
+        }
+        catch {
+            // Try with npx
+            execSync('npx uam --version', { encoding: 'utf-8', cwd: PROJECT_ROOT, stdio: 'pipe' });
+        }
+        // Step 2: Initialize UAM (idempotent - safe to run multiple times)
+        log('Running uam init...');
+        try {
+            execSync('uam init --non-interactive 2>/dev/null || true', {
+                encoding: 'utf-8',
+                cwd: PROJECT_ROOT,
+                stdio: 'pipe',
+                timeout: 30000,
+            });
+            result.initialized = true;
+        }
+        catch (e) {
+            result.errors.push(`init failed: ${e instanceof Error ? e.message : String(e)}`);
+        }
+        // Step 3: Analyze project structure
+        log('Running uam analyze...');
+        try {
+            execSync('uam analyze 2>/dev/null || true', {
+                encoding: 'utf-8',
+                cwd: PROJECT_ROOT,
+                stdio: 'pipe',
+                timeout: 60000,
+            });
+        }
+        catch (e) {
+            result.errors.push(`analyze failed: ${e instanceof Error ? e.message : String(e)}`);
+        }
+        // Step 4: Generate/update CLAUDE.md
+        log('Running uam generate...');
+        try {
+            execSync('uam generate 2>/dev/null || true', {
+                encoding: 'utf-8',
+                cwd: PROJECT_ROOT,
+                stdio: 'pipe',
+                timeout: 30000,
+            });
+        }
+        catch (e) {
+            result.errors.push(`generate failed: ${e instanceof Error ? e.message : String(e)}`);
+        }
+        // Step 5: Start memory services
+        log('Starting memory services...');
+        try {
+            execSync('uam memory start 2>/dev/null || true', {
+                encoding: 'utf-8',
+                cwd: PROJECT_ROOT,
+                stdio: 'pipe',
+                timeout: 60000,
+            });
+            result.memoryStarted = true;
+        }
+        catch (e) {
+            result.errors.push(`memory start failed: ${e instanceof Error ? e.message : String(e)}`);
+        }
+        // Step 6: Prepopulate memory from docs and git history
+        log('Prepopulating memory from docs and git...');
+        try {
+            execSync('uam memory prepopulate --docs --git --limit 100 2>/dev/null || true', {
+                encoding: 'utf-8',
+                cwd: PROJECT_ROOT,
+                stdio: 'pipe',
+                timeout: 120000,
+            });
+            result.memoryPrepopulated = true;
+        }
+        catch (e) {
+            result.errors.push(`memory prepopulate failed: ${e instanceof Error ? e.message : String(e)}`);
+        }
+        // Step 7: Verify CLAUDE.md exists
+        const claudeMdPath = join(PROJECT_ROOT, 'CLAUDE.md');
+        if (existsSync(claudeMdPath)) {
+            result.claudeMdLoaded = true;
+            log('CLAUDE.md found and ready');
+        }
+        else {
+            result.errors.push('CLAUDE.md not found after setup');
+        }
+    }
+    catch (error) {
+        result.errors.push(`UAM setup error: ${error instanceof Error ? error.message : String(error)}`);
+    }
+    return result;
+}
+/**
+ * Load UAM memory context from CLAUDE.md and short-term memory
+ */
+function loadUAMMemoryContext() {
+    const sections = [];
+    // Read CLAUDE.md
+    const claudeMdPath = join(PROJECT_ROOT, 'CLAUDE.md');
+    if (existsSync(claudeMdPath)) {
+        const claudeMd = readFileSync(claudeMdPath, 'utf-8');
+        // Extract key sections from CLAUDE.md
+        sections.push('## UAM Memory Context (from CLAUDE.md)\n');
+        // Extract Code Field section
+        const codeFieldMatch = claudeMd.match(/## .*CODE FIELD.*?(?=\n## |\n---\n|$)/s);
+        if (codeFieldMatch) {
+            sections.push('### Code Field Guidelines\n');
+            sections.push(codeFieldMatch[0].slice(0, 1500) + '\n');
+        }
+        // Extract Testing Requirements
+        const testingMatch = claudeMd.match(/## .*Testing Requirements.*?(?=\n## |\n---\n|$)/s);
+        if (testingMatch) {
+            sections.push('### Testing Requirements\n');
+            sections.push(testingMatch[0].slice(0, 500) + '\n');
+        }
+        // Extract Repository Structure
+        const structureMatch = claudeMd.match(/## Repository Structure.*?```[\s\S]*?```/);
+        if (structureMatch) {
+            sections.push('### Repository Structure\n');
+            sections.push(structureMatch[0].slice(0, 1000) + '\n');
+        }
+    }
+    // Query short-term memory from SQLite
+    const dbPath = join(PROJECT_ROOT, 'agents/data/memory/short_term.db');
+    if (existsSync(dbPath)) {
+        try {
+            const recentMemories = execSync(`sqlite3 "${dbPath}" "SELECT type, content FROM memories ORDER BY id DESC LIMIT 10;" 2>/dev/null || true`, { encoding: 'utf-8', cwd: PROJECT_ROOT }).trim();
+            if (recentMemories) {
+                sections.push('### Recent Session Memory\n');
+                sections.push('```\n' + recentMemories.slice(0, 1000) + '\n```\n');
+            }
+            // Get lessons learned
+            const lessons = execSync(`sqlite3 "${dbPath}" "SELECT content FROM memories WHERE type='lesson' ORDER BY id DESC LIMIT 5;" 2>/dev/null || true`, { encoding: 'utf-8', cwd: PROJECT_ROOT }).trim();
+            if (lessons) {
+                sections.push('### Lessons Learned\n');
+                sections.push(lessons.slice(0, 500) + '\n');
+            }
+        }
+        catch {
+            // Memory DB not available
+        }
+    }
+    // Add static context as fallback/supplement
+    sections.push(`
+### Project Coding Standards
+- Use TypeScript strict mode
+- All functions must have JSDoc comments with @param and @returns
+- Error handling uses custom AppError class that extends Error
+- Prefer async/await over callbacks and Promises
+- Use zod for runtime input validation
+- Export types and interfaces alongside implementations
+- Use Map for key-value storage, Set for unique collections
+
+### Common Patterns
+- Singleton pattern: private constructor + static getInstance()
+- Strategy pattern: interface + multiple implementations
+- Factory pattern: static create() methods
+- Error handling: try/catch with specific error types
+- Exponential backoff: delay = baseMs * Math.pow(2, attempt)
+
+### Known Gotchas (from memory)
+- Always check array bounds: use i < length, not i <= length
+- Handle empty arrays explicitly before operations
+- Include cleanup logic for resources (timers, connections)
+- JSON.parse throws on invalid input - always wrap in try/catch
+- Array methods like reduce need initial value for empty arrays
+- Map.get() returns undefined for missing keys
+
+---
+
+`);
+    return sections.join('\n');
+}
+// Cached memory context (loaded once per benchmark run)
+let cachedMemoryContext = null;
+function getUAMMemoryContext() {
+    if (!cachedMemoryContext) {
+        cachedMemoryContext = loadUAMMemoryContext();
+    }
+    return cachedMemoryContext;
+}
+// ============================================================================
+// Model Configurations (per Factory.ai droid CLI available models)
+// ============================================================================
+const MODELS = [
+    {
+        id: 'opus-4.5',
+        name: 'Claude Opus 4.5',
+        provider: 'anthropic',
+        apiModel: 'claude-opus-4-5-20251101',
+    },
+    {
+        id: 'glm-4.7',
+        name: 'GLM 4.7 (Droid Core)',
+        provider: 'zhipu',
+        apiModel: 'glm-4.7',
+    },
+    {
+        id: 'gpt-5.2-codex',
+        name: 'GPT 5.2 Codex',
+        provider: 'openai',
+        apiModel: 'gpt-5.2-codex',
+    },
+    {
+        id: 'gpt-5.2',
+        name: 'GPT 5.2',
+        provider: 'openai',
+        apiModel: 'gpt-5.2',
+    },
+];
+// ============================================================================
+// Benchmark Tasks
+// ============================================================================
+const BENCHMARK_TASKS = [
+    {
+        id: 'task-001-code-generation',
+        name: 'TypeScript Function Generation',
+        description: 'Generate a well-typed TypeScript function',
+        prompt: `Write a TypeScript function called 'calculateAverage' that:
+1. Takes an array of numbers as input
+2. Returns the arithmetic mean
+3. Handles empty arrays (return 0)
+4. Has proper type annotations
+
+Return ONLY the function code, no explanations.`,
+        difficulty: 'easy',
+        category: 'code-generation',
+        expectedPatterns: [
+            'function calculateAverage',
+            'number[]',
+            ': number',
+            'length',
+            'return',
+        ],
+        maxTokens: 500,
+    },
+    {
+        id: 'task-002-bug-fix',
+        name: 'Bug Detection and Fix',
+        description: 'Identify and fix a bug in code',
+        prompt: `Find and fix the bug in this TypeScript code:
+
+function sumPositive(nums: number[]): number {
+  let sum = 0;
+  for (let i = 0; i <= nums.length; i++) {
+    if (nums[i] > 0) {
+      sum += nums[i];
+    }
+  }
+  return sum;
+}
+
+Return ONLY the corrected function code.`,
+        difficulty: 'easy',
+        category: 'bug-fix',
+        expectedPatterns: [
+            'i < nums.length',
+            'function sumPositive',
+            'return sum',
+        ],
+        maxTokens: 500,
+    },
+    {
+        id: 'task-003-pattern-application',
+        name: 'Design Pattern Implementation',
+        description: 'Implement a singleton pattern',
+        prompt: `Implement a TypeScript singleton class called 'ConfigManager' that:
+1. Has a private constructor
+2. Has a static getInstance() method
+3. Has get(key: string) and set(key: string, value: any) methods
+4. Stores configuration in a private Map
+
+Return ONLY the class code.`,
+        difficulty: 'medium',
+        category: 'patterns',
+        expectedPatterns: [
+            'class ConfigManager',
+            'private constructor',
+            'static getInstance',
+            'private static instance',
+            'Map',
+        ],
+        maxTokens: 800,
+    },
+    {
+        id: 'task-004-refactoring',
+        name: 'Code Refactoring',
+        description: 'Refactor code for better maintainability',
+        prompt: `Refactor this code to follow SOLID principles and improve readability:
+
+function processOrder(order: any) {
+  if (order.type === 'digital') {
+    console.log('Sending email with download link');
+    order.status = 'delivered';
+  } else if (order.type === 'physical') {
+    console.log('Creating shipping label');
+    order.status = 'shipped';
+  } else if (order.type === 'subscription') {
+    console.log('Activating subscription');
+    order.status = 'active';
+  }
+  console.log('Order processed: ' + order.id);
+  return order;
+}
+
+Provide the refactored TypeScript code using proper interfaces and a strategy pattern.`,
+        difficulty: 'medium',
+        category: 'refactoring',
+        expectedPatterns: [
+            'interface',
+            'class',
+            'implements',
+            'process',
+        ],
+        maxTokens: 1200,
+    },
+    {
+        id: 'task-005-memory-context',
+        name: 'Context-Aware Code Generation',
+        description: 'Generate code using provided context',
+        prompt: `Given the following project context from memory:
+
+MEMORY CONTEXT:
+- Project uses src/utils/ for utility functions
+- All functions must have JSDoc comments
+- Error handling uses custom AppError class
+- Prefer async/await over callbacks
+- Use zod for input validation
+
+Write a utility function 'validateAndParseJSON' that:
+1. Takes a string input
+2. Validates it's valid JSON using zod
+3. Returns the parsed object or throws AppError
+4. Has proper JSDoc documentation
+
+Return ONLY the function code with JSDoc.`,
+        difficulty: 'medium',
+        category: 'memory',
+        expectedPatterns: [
+            'async',
+            'zod',
+            'AppError',
+            '@param',
+            '@returns',
+            'validateAndParseJSON',
+        ],
+        maxTokens: 800,
+    },
+    {
+        id: 'task-006-complex-algorithm',
+        name: 'Algorithm Implementation',
+        description: 'Implement a complex algorithm with proper typing',
+        prompt: `Implement a TypeScript function 'findShortestPath' using Dijkstra's algorithm:
+
+1. Input: weighted graph as adjacency list Map<string, Map<string, number>>
+2. Input: start node (string), end node (string)
+3. Output: { path: string[], distance: number } or null if no path
+4. Handle disconnected nodes properly
+5. Use proper TypeScript types
+
+Return ONLY the function code with type definitions.`,
+        difficulty: 'hard',
+        category: 'algorithms',
+        expectedPatterns: [
+            'function findShortestPath',
+            'Map<string',
+            'distance',
+            'path',
+            'while',
+            'return',
+        ],
+        maxTokens: 1500,
+    },
+    {
+        id: 'task-007-multi-step-task',
+        name: 'Multi-Step Code Generation',
+        description: 'Complete a multi-step implementation task',
+        prompt: `Create a complete TypeScript module for a rate limiter with these requirements:
+
+1. Interface RateLimiterConfig { maxRequests: number; windowMs: number; }
+2. Class RateLimiter with:
+   - constructor(config: RateLimiterConfig)
+   - isAllowed(clientId: string): boolean
+   - getRemainingRequests(clientId: string): number
+   - reset(clientId?: string): void
+3. Use Map for tracking requests per client
+4. Include proper cleanup of expired entries
+5. Export both the class and interface
+
+Return the complete module code.`,
+        difficulty: 'hard',
+        category: 'multi-step',
+        expectedPatterns: [
+            'interface RateLimiterConfig',
+            'class RateLimiter',
+            'isAllowed',
+            'getRemainingRequests',
+            'reset',
+            'Map',
+            'export',
+        ],
+        maxTokens: 2000,
+    },
+    {
+        id: 'task-008-error-handling',
+        name: 'Comprehensive Error Handling',
+        description: 'Implement robust error handling',
+        prompt: `Create a TypeScript async function 'fetchWithRetry' that:
+
+1. Takes url: string, options?: RequestInit, retryConfig?: { maxRetries: number; backoffMs: number; }
+2. Implements exponential backoff retry logic
+3. Handles network errors, timeout, and HTTP errors (4xx, 5xx)
+4. Returns Promise<Response> or throws a detailed custom error
+5. Logs each retry attempt
+6. Has proper TypeScript types for all parameters and return values
+
+Return ONLY the function code with any necessary type definitions.`,
+        difficulty: 'hard',
+        category: 'error-handling',
+        expectedPatterns: [
+            'async function fetchWithRetry',
+            'retry',
+            'backoff',
+            'catch',
+            'throw',
+            'Promise<Response>',
+        ],
+        maxTokens: 1200,
+    },
+];
+// ============================================================================
+// Droid Exec Client
+// ============================================================================
+class DroidExecClient {
+    apiKey;
+    tmpDir;
+    autoLevel;
+    constructor(apiKey, autoLevel = 'low') {
+        this.apiKey = apiKey;
+        this.autoLevel = autoLevel;
+        this.tmpDir = '/tmp/uam-benchmark';
+        try {
+            execSync(`mkdir -p ${this.tmpDir}`, { encoding: 'utf-8' });
+        }
+        catch {
+            // ignore
+        }
+    }
+    async complete(model, prompt) {
+        const startTime = Date.now();
+        // Write prompt to temp file to avoid shell escaping issues
+        const promptFile = `${this.tmpDir}/prompt-${Date.now()}.txt`;
+        writeFileSync(promptFile, prompt, 'utf-8');
+        try {
+            // Use --auto low to allow file operations without system modifications
+            const result = execSync(`FACTORY_API_KEY="${this.apiKey}" droid exec --model "${model}" --auto ${this.autoLevel} -f "${promptFile}"`, {
+                encoding: 'utf-8',
+                timeout: 300000, // 5 minutes for complex tasks
+                maxBuffer: 10 * 1024 * 1024,
+                env: { ...process.env, FACTORY_API_KEY: this.apiKey },
+            });
+            const latencyMs = Date.now() - startTime;
+            // Clean up temp file
+            try {
+                execSync(`rm "${promptFile}"`, { encoding: 'utf-8' });
+            }
+            catch {
+                // ignore cleanup failures
+            }
+            return {
+                content: result.trim(),
+                tokensUsed: 0,
+                latencyMs,
+            };
+        }
+        catch (error) {
+            // Clean up temp file
+            try {
+                execSync(`rm "${promptFile}"`, { encoding: 'utf-8' });
+            }
+            catch {
+                // ignore cleanup failures
+            }
+            const errMsg = error instanceof Error ? error.message : String(error);
+            throw new Error(`droid exec failed: ${errMsg}`);
+        }
+    }
+}
+// ============================================================================
+// Benchmark Runner
+// ============================================================================
+function evaluateResponse(response, expectedPatterns) {
+    const normalizedResponse = response.toLowerCase();
+    return expectedPatterns.filter(pattern => normalizedResponse.includes(pattern.toLowerCase()));
+}
+async function runTaskForModel(client, model, task, withMemory = false) {
+    const result = {
+        taskId: task.id,
+        modelId: model.id,
+        success: false,
+        latencyMs: 0,
+        tokensUsed: 0,
+        response: '',
+        matchedPatterns: [],
+    };
+    try {
+        // Inject UAM memory context if enabled (loaded from CLAUDE.md + memory DB)
+        const prompt = withMemory
+            ? getUAMMemoryContext() + task.prompt
+            : task.prompt;
+        const completion = await client.complete(model.apiModel, prompt);
+        result.response = completion.content;
+        result.latencyMs = completion.latencyMs;
+        result.tokensUsed = completion.tokensUsed;
+        result.matchedPatterns = evaluateResponse(completion.content, task.expectedPatterns);
+        const matchRatio = result.matchedPatterns.length / task.expectedPatterns.length;
+        result.success = matchRatio >= 0.6;
+    }
+    catch (error) {
+        result.error = error instanceof Error ? error.message : String(error);
+    }
+    return result;
+}
+async function runBenchmarkForModel(client, model, tasks, withMemory = false) {
+    const memoryLabel = withMemory ? ' (with UAM Memory)' : ' (without Memory)';
+    console.log(`\n${'='.repeat(60)}`);
+    console.log(`Running benchmark for: ${model.name}${memoryLabel}`);
+    console.log(`${'='.repeat(60)}`);
+    const results = [];
+    for (const task of tasks) {
+        console.log(`  [${task.difficulty.toUpperCase()}] ${task.name}...`);
+        const result = await runTaskForModel(client, model, task, withMemory);
+        results.push(result);
+        if (result.success) {
+            console.log(`    ✓ Success (${result.latencyMs}ms)`);
+        }
+        else {
+            console.log(`    ✗ Failed: ${result.error || 'Pattern mismatch'}`);
+        }
+        // Small delay between tasks
+        await new Promise(r => setTimeout(r, 1000));
+    }
+    const succeeded = results.filter(r => r.success).length;
+    const successfulResults = results.filter(r => r.latencyMs > 0);
+    const avgLatency = successfulResults.length > 0
+        ? successfulResults.reduce((sum, r) => sum + r.latencyMs, 0) / successfulResults.length
+        : 0;
+    const totalTokens = results.reduce((sum, r) => sum + r.tokensUsed, 0);
+    return {
+        modelId: model.id,
+        modelName: model.name,
+        tasksRun: tasks.length,
+        tasksSucceeded: succeeded,
+        successRate: (succeeded / tasks.length) * 100,
+        avgLatencyMs: Math.round(avgLatency),
+        totalTokens,
+        results,
+    };
+}
+function generateComparison(modelResults) {
+    const sorted = [...modelResults].sort((a, b) => b.successRate - a.successRate);
+    const fastest = [...modelResults].sort((a, b) => a.avgLatencyMs - b.avgLatencyMs);
+    const byDifficulty = {};
+    for (const diff of ['easy', 'medium', 'hard']) {
+        let bestModel = '';
+        let bestRate = 0;
+        for (const modelResult of modelResults) {
+            const diffTasks = modelResult.results.filter(r => {
+                const task = BENCHMARK_TASKS.find(t => t.id === r.taskId);
+                return task?.difficulty === diff;
+            });
+            if (diffTasks.length > 0) {
+                const rate = (diffTasks.filter(t => t.success).length / diffTasks.length) * 100;
+                if (rate > bestRate) {
+                    bestRate = rate;
+                    bestModel = modelResult.modelName;
+                }
+            }
+        }
+        byDifficulty[diff] = { model: bestModel, successRate: bestRate };
+    }
+    return {
+        bestOverall: sorted[0]?.modelName || 'N/A',
+        fastestModel: fastest[0]?.modelName || 'N/A',
+        mostAccurate: sorted[0]?.modelName || 'N/A',
+        byDifficulty,
+    };
+}
+function generateMarkdownReport(report) {
+    const lines = [
+        '# Model Integration Benchmark Results',
+        '',
+        `**Generated:** ${report.timestamp}`,
+        `**Models Tested:** ${report.models.map(m => m.modelName).join(', ')}`,
+        `**Tasks Run:** ${BENCHMARK_TASKS.length}`,
+        '',
+        '---',
+        '',
+        '## Executive Summary',
+        '',
+        '| Model | Success Rate | Avg Latency | Total Tokens |',
+        '|-------|--------------|-------------|--------------|',
+    ];
+    for (const model of report.models) {
+        lines.push(`| ${model.modelName} | ${model.successRate.toFixed(1)}% | ${model.avgLatencyMs}ms | ${model.totalTokens} |`);
+    }
+    lines.push('', '---', '', '## Comparison', '');
+    lines.push(`- **Best Overall:** ${report.comparison.bestOverall}`);
+    lines.push(`- **Fastest Model:** ${report.comparison.fastestModel}`);
+    lines.push(`- **Most Accurate:** ${report.comparison.mostAccurate}`);
+    lines.push('', '### By Difficulty', '');
+    lines.push('| Difficulty | Best Model | Success Rate |');
+    lines.push('|------------|------------|--------------|');
+    for (const [diff, data] of Object.entries(report.comparison.byDifficulty)) {
+        lines.push(`| ${diff} | ${data.model} | ${data.successRate.toFixed(1)}% |`);
+    }
+    lines.push('', '---', '', '## Detailed Results', '');
+    for (const model of report.models) {
+        lines.push(`### ${model.modelName}`, '');
+        lines.push('| Task | Difficulty | Success | Latency | Patterns Matched |');
+        lines.push('|------|------------|---------|---------|------------------|');
+        for (const result of model.results) {
+            const task = BENCHMARK_TASKS.find(t => t.id === result.taskId);
+            const status = result.success ? '✓' : '✗';
+            const patterns = `${result.matchedPatterns.length}/${task?.expectedPatterns.length || 0}`;
+            lines.push(`| ${task?.name || result.taskId} | ${task?.difficulty || 'N/A'} | ${status} | ${result.latencyMs}ms | ${patterns} |`);
+        }
+        lines.push('');
+    }
+    // Add memory comparison section if available
+    if (report.memoryComparison) {
+        lines.push('---', '', '## UAM Memory Impact Analysis', '');
+        lines.push('### Success Rate Comparison', '');
+        lines.push('| Model | Without Memory | With Memory | Improvement |');
+        lines.push('|-------|----------------|-------------|-------------|');
+        for (const withMem of report.memoryComparison.withMemory) {
+            const without = report.memoryComparison.withoutMemory.find(r => r.modelId === withMem.modelId);
+            const imp = report.memoryComparison.improvement[withMem.modelId];
+            if (without && imp) {
+                const sign = imp.successDelta >= 0 ? '+' : '';
+                lines.push(`| ${withMem.modelName} | ${without.successRate.toFixed(1)}% | ${withMem.successRate.toFixed(1)}% | ${sign}${imp.successDelta.toFixed(1)}% |`);
+            }
+        }
+        lines.push('', '### Latency Comparison', '');
+        lines.push('| Model | Without Memory | With Memory | Speed Ratio |');
+        lines.push('|-------|----------------|-------------|-------------|');
+        for (const withMem of report.memoryComparison.withMemory) {
+            const without = report.memoryComparison.withoutMemory.find(r => r.modelId === withMem.modelId);
+            const imp = report.memoryComparison.improvement[withMem.modelId];
+            if (without && imp) {
+                const speedLabel = imp.speedupRatio > 1 ? `${imp.speedupRatio.toFixed(2)}x faster` :
+                    imp.speedupRatio < 1 ? `${(1 / imp.speedupRatio).toFixed(2)}x slower` : 'same';
+                lines.push(`| ${withMem.modelName} | ${without.avgLatencyMs}ms | ${withMem.avgLatencyMs}ms | ${speedLabel} |`);
+            }
+        }
+        lines.push('', '### Key Findings', '');
+        // Find best improvement
+        const improvements = Object.entries(report.memoryComparison.improvement);
+        if (improvements.length > 0) {
+            const bestImprovement = improvements.reduce((a, b) => a[1].successDelta > b[1].successDelta ? a : b);
+            const bestModel = BENCHMARK_TASKS.length > 0 ?
+                report.memoryComparison.withMemory.find(m => m.modelId === bestImprovement[0])?.modelName : 'N/A';
+            lines.push(`- **Best Memory Benefit:** ${bestModel} (+${bestImprovement[1].successDelta.toFixed(1)}% success rate)`);
+            const avgImprovement = improvements.reduce((sum, [_, imp]) => sum + imp.successDelta, 0) / improvements.length;
+            lines.push(`- **Average Improvement:** +${avgImprovement.toFixed(1)}% success rate across all models`);
+            lines.push('', '### Interpretation', '');
+            lines.push('UAM memory context injection provides models with:');
+            lines.push('- Project structure knowledge (file locations, patterns)');
+            lines.push('- Coding standards (JSDoc, error handling, async patterns)');
+            lines.push('- Common gotchas and lessons learned from previous sessions');
+            lines.push('- Design pattern templates (singleton, strategy, factory)');
+        }
+    }
+    lines.push('', '---', '', '**Report Generated by UAM Model Integration Benchmark**');
+    return lines.join('\n');
+}
+// ============================================================================
+// Parallel Execution Utilities
+// ============================================================================
+/**
+ * Run multiple model benchmarks in parallel with configurable concurrency
+ */
+async function runModelsInParallel(client, models, tasks, withMemory, concurrency) {
+    const results = [];
+    const queue = [...models];
+    const inProgress = [];
+    const runNext = async () => {
+        const model = queue.shift();
+        if (!model)
+            return;
+        const result = await runBenchmarkForModel(client, model, tasks, withMemory);
+        results.push(result);
+        if (queue.length > 0) {
+            await runNext();
+        }
+    };
+    // Start initial batch up to concurrency limit
+    const initialBatch = Math.min(concurrency, models.length);
+    for (let i = 0; i < initialBatch; i++) {
+        inProgress.push(runNext());
+    }
+    await Promise.all(inProgress);
+    // Sort results to match original model order
+    return models.map(m => results.find(r => r.modelId === m.id)).filter(Boolean);
+}
+export async function runModelBenchmark(apiKeyOrOptions, modelIds, compareMemory = true, parallelModels = 1) {
+    // Handle both old signature and new options object
+    let key;
+    let models;
+    let compare;
+    let parallel;
+    if (typeof apiKeyOrOptions === 'object' && apiKeyOrOptions !== null) {
+        key = apiKeyOrOptions.apiKey;
+        models = apiKeyOrOptions.modelIds;
+        compare = apiKeyOrOptions.compareMemory ?? true;
+        parallel = apiKeyOrOptions.parallelModels ?? 1;
+    }
+    else {
+        key = apiKeyOrOptions;
+        models = modelIds;
+        compare = compareMemory;
+        parallel = parallelModels;
+    }
+    key = key || process.env.FACTORY_API_KEY || process.env.DROID_API_KEY;
+    if (!key) {
+        throw new Error('FACTORY_API_KEY or DROID_API_KEY not provided and not found in environment');
+    }
+    const client = new DroidExecClient(key, 'medium');
+    const modelsToTest = models
+        ? MODELS.filter(m => models.includes(m.id))
+        : MODELS;
+    if (modelsToTest.length === 0) {
+        throw new Error('No valid models specified');
+    }
+    // Determine effective parallelism
+    const effectiveParallel = Math.min(parallel, modelsToTest.length);
+    const isParallel = effectiveParallel > 1;
+    console.log('\n' + '='.repeat(60));
+    console.log('   UAM MODEL INTEGRATION BENCHMARK');
+    console.log('='.repeat(60));
+    console.log(`\nModels: ${modelsToTest.map(m => m.name).join(', ')}`);
+    console.log(`Tasks: ${BENCHMARK_TASKS.length}`);
+    console.log(`Memory Comparison: ${compare ? 'ENABLED' : 'DISABLED'}`);
+    console.log(`Parallel Models: ${effectiveParallel}${isParallel ? ' (ENABLED)' : ' (sequential)'}`);
+    let withoutMemoryResults = [];
+    let withMemoryResults = [];
+    // Run without memory first
+    console.log('\n' + '█'.repeat(60));
+    console.log(`   PHASE 1: WITHOUT UAM MEMORY${isParallel ? ' (PARALLEL)' : ''}`);
+    console.log('█'.repeat(60));
+    if (isParallel) {
+        console.log(`\n  Running ${modelsToTest.length} models with concurrency=${effectiveParallel}...\n`);
+        withoutMemoryResults = await runModelsInParallel(client, modelsToTest, BENCHMARK_TASKS, false, effectiveParallel);
+    }
+    else {
+        for (const model of modelsToTest) {
+            const result = await runBenchmarkForModel(client, model, BENCHMARK_TASKS, false);
+            withoutMemoryResults.push(result);
+        }
+    }
+    // Run with memory if comparison enabled
+    if (compare) {
+        console.log('\n' + '█'.repeat(60));
+        console.log(`   PHASE 2: WITH UAM MEMORY${isParallel ? ' (PARALLEL)' : ''}`);
+        console.log('█'.repeat(60));
+        // Setup UAM before running with-memory tests
+        console.log('\n--- Setting up UAM (init, analyze, generate, memory start, prepopulate) ---');
+        const uamSetup = await setupUAM(true);
+        if (uamSetup.errors.length > 0) {
+            console.log('\nUAM Setup warnings:');
+            uamSetup.errors.forEach(e => console.log(`  - ${e}`));
+        }
+        console.log(`\nUAM Status:`);
+        console.log(`  Initialized: ${uamSetup.initialized ? '✓' : '✗'}`);
+        console.log(`  Memory Started: ${uamSetup.memoryStarted ? '✓' : '✗'}`);
+        console.log(`  Memory Prepopulated: ${uamSetup.memoryPrepopulated ? '✓' : '✗'}`);
+        console.log(`  CLAUDE.md Loaded: ${uamSetup.claudeMdLoaded ? '✓' : '✗'}`);
+        // Clear cached context to force reload with fresh memory
+        cachedMemoryContext = null;
+        // Log memory context size
+        const memoryContext = getUAMMemoryContext();
+        console.log(`  Memory Context Size: ${memoryContext.length} chars\n`);
+        if (isParallel) {
+            console.log(`  Running ${modelsToTest.length} models with concurrency=${effectiveParallel}...\n`);
+            withMemoryResults = await runModelsInParallel(client, modelsToTest, BENCHMARK_TASKS, true, effectiveParallel);
+        }
+        else {
+            for (const model of modelsToTest) {
+                const result = await runBenchmarkForModel(client, model, BENCHMARK_TASKS, true);
+                withMemoryResults.push(result);
+            }
+        }
+    }
+    // Calculate memory improvement for each model
+    const improvement = {};
+    if (compare) {
+        for (const model of modelsToTest) {
+            const without = withoutMemoryResults.find(r => r.modelId === model.id);
+            const withMem = withMemoryResults.find(r => r.modelId === model.id);
+            if (without && withMem) {
+                improvement[model.id] = {
+                    successDelta: withMem.successRate - without.successRate,
+                    speedupRatio: without.avgLatencyMs > 0 ? without.avgLatencyMs / withMem.avgLatencyMs : 1,
+                };
+            }
+        }
+    }
+    // Use with-memory results as primary if available, otherwise without
+    const primaryResults = compare && withMemoryResults.length > 0
+        ? withMemoryResults
+        : withoutMemoryResults;
+    const report = {
+        timestamp: new Date().toISOString(),
+        models: primaryResults,
+        comparison: generateComparison(primaryResults),
+        memoryComparison: compare ? {
+            withMemory: withMemoryResults,
+            withoutMemory: withoutMemoryResults,
+            improvement,
+        } : undefined,
+    };
+    // Generate and save markdown report
+    const markdown = generateMarkdownReport(report);
+    const reportPath = join(__dirname, '../../MODEL_BENCHMARK_RESULTS.md');
+    writeFileSync(reportPath, markdown);
+    console.log(`\nReport saved to: ${reportPath}`);
+    // Print summary
+    console.log('\n' + '='.repeat(60));
+    console.log('   BENCHMARK COMPLETE');
+    console.log('='.repeat(60));
+    if (compare) {
+        console.log('\n--- Without Memory ---');
+        for (const model of withoutMemoryResults) {
+            console.log(`  ${model.modelName}: ${model.successRate.toFixed(1)}% success, ${model.avgLatencyMs}ms avg`);
+        }
+        console.log('\n--- With UAM Memory ---');
+        for (const model of withMemoryResults) {
+            console.log(`  ${model.modelName}: ${model.successRate.toFixed(1)}% success, ${model.avgLatencyMs}ms avg`);
+        }
+        console.log('\n--- Memory Improvement ---');
+        for (const [modelId, imp] of Object.entries(improvement)) {
+            const model = modelsToTest.find(m => m.id === modelId);
+            const sign = imp.successDelta >= 0 ? '+' : '';
+            console.log(`  ${model?.name}: ${sign}${imp.successDelta.toFixed(1)}% success, ${imp.speedupRatio.toFixed(2)}x speed`);
+        }
+    }
+    else {
+        console.log('\nSummary:');
+        for (const model of primaryResults) {
+            console.log(`  ${model.modelName}: ${model.successRate.toFixed(1)}% success, ${model.avgLatencyMs}ms avg`);
+        }
+    }
+    console.log(`\nBest Overall: ${report.comparison.bestOverall}`);
+    return report;
+}
+// CLI entry point
+if (process.argv[1]?.includes('model-integration')) {
+    const envPath = join(__dirname, '../../.env');
+    if (existsSync(envPath)) {
+        const envContent = readFileSync(envPath, 'utf-8');
+        for (const line of envContent.split('\n')) {
+            const [key, ...valueParts] = line.split('=');
+            if (key && valueParts.length > 0) {
+                process.env[key.trim()] = valueParts.join('=').trim();
+            }
+        }
+    }
+    runModelBenchmark()
+        .then(() => process.exit(0))
+        .catch(err => {
+        console.error('Benchmark failed:', err);
+        process.exit(1);
+    });
+}
+export { MODELS, BENCHMARK_TASKS, setupUAM, loadUAMMemoryContext };
+//# sourceMappingURL=model-integration.js.map