claude-code-router-config 1.0.1 → 1.1.0

package/config/smart-intent-router.js

@@ -0,0 +1,543 @@
+/**
+ * Enhanced Smart Intent Router
+ * Intelligent routing with cost, performance, and quality optimization
+ *
+ * Features:
+ * - Intent-based routing
+ * - Cost-aware provider selection
+ * - Performance-based routing
+ * - Health monitoring integration
+ * - Auto-fallback mechanisms
+ * - Learning from past performance
+ *
+ * Configuration by Halil Ertekin
+ */
+
+const fs = require('fs');
+const path = require('path');
+const os = require('os');
+
+// Performance metrics storage
+const METRICS_PATH = path.join(os.homedir(), '.claude-code-router', 'routing-metrics.json');
+
+// Load historical performance data
+function loadMetrics() {
+  try {
+    if (fs.existsSync(METRICS_PATH)) {
+      return JSON.parse(fs.readFileSync(METRICS_PATH, 'utf8'));
+    }
+  } catch (error) {
+    console.error('Failed to load metrics:', error);
+  }
+  return {
+    providers: {},
+    intents: {},
+    requests: []
+  };
+}
+
+// Save performance metrics
+function saveMetrics(metrics) {
+  try {
+    // Keep only last 1000 requests
+    if (metrics.requests.length > 1000) {
+      metrics.requests = metrics.requests.slice(-1000);
+    }
+    fs.writeFileSync(METRICS_PATH, JSON.stringify(metrics, null, 2));
+  } catch (error) {
+    console.error('Failed to save metrics:', error);
+  }
+}
+
+// Enhanced intent definitions with routing strategies
+const INTENTS = {
+  // Coding tasks → OpenAI (GPT-4o, O1) with fallbacks
+  CODING: {
+    patterns: [
+      /\b(implement|refactor|debug|fix|write|code|function|class|method|bug|error|compile|syntax)\b/i,
+      /\b(typescript|javascript|python|rust|go|java|react|vue|angular|swift|kotlin)\b/i,
+      /\b(api|endpoint|database|query|migration|schema|test|unit test)\b/i,
+      /\b(codex|o1|reasoning)\b/i
+    ],
+    route: "openai,gpt-4o",
+    strategy: "quality",
+    fallbacks: ["anthropic,claude-sonnet-4-latest", "qwen,qwen3-coder-plus"],
+    priority: "high"
+  },
+
+  // Deep reasoning → Anthropic Claude with cost consideration
+  REASONING: {
+    patterns: [
+      /\b(architect|design|analyze|plan|strategy|structure|system|trade-?off)\b/i,
+      /\b(why|explain|reason|understand|compare|evaluate|consider|review)\b/i,
+      /\b(decision|approach|best practice|pattern|principle|philosophy)\b/i
+    ],
+    route: "anthropic,claude-sonnet-4-latest",
+    strategy: "quality",
+    fallbacks: ["openai,gpt-4o", "gemini,gemini-2.5-pro"],
+    priority: "high"
+  },
+
+  // Fast responses → Gemini Flash
+  FAST: {
+    patterns: [
+      /\b(fast|quick|brief|short|summary|tldr|overview|hızlı)\b/i,
+      /\b(scan|check|verify|validate)\b/i
+    ],
+    route: "gemini,gemini-2.5-flash",
+    strategy: "performance",
+    fallbacks: ["qwen,qwen-turbo", "glm,glm-4.5"],
+    priority: "medium"
+  },
+
+  // Simple/cheap tasks → Qwen
+  SIMPLE: {
+    patterns: [
+      /\b(list|show|what is|simple|basic|help|how to|format)\b/i,
+      /\b(rename|move|delete|create file|mkdir|copy)\b/i,
+      /\b(ucuz|basit|kolay)\b/i
+    ],
+    route: "qwen,qwen-plus",
+    strategy: "cost",
+    fallbacks: ["glm,glm-4.5", "gemini,gemini-2.5-flash"],
+    priority: "low"
+  },
+
+  // Multilingual → GLM
+  MULTILINGUAL: {
+    patterns: [
+      /\b(translate|çevir|tercüme|chinese|türkçe|multilingual)\b/i,
+      /[\u4e00-\u9fff]/, // Chinese characters
+      /[\u0600-\u06FF]/, // Arabic
+    ],
+    route: "glm,glm-4.6",
+    strategy: "quality",
+    fallbacks: ["qwen,qwen-plus", "gemini,gemini-2.5-flash"],
+    priority: "medium"
+  },
+
+  // Heavy reasoning → O1
+  HEAVY_REASONING: {
+    patterns: [
+      /\b(complex algorithm|optimization|performance critical|system design)\b/i,
+      /\b(prove|mathematical|theorem|formal verification)\b/i
+    ],
+    route: "openai,o1",
+    strategy: "quality",
+    fallbacks: ["anthropic,claude-sonnet-4-latest", "openai,gpt-4o"],
+    priority: "highest"
+  },
+
+  // AgentSkills commands → Anthropic with specialized routing
+  AGENT_SKILLS: {
+    patterns: [
+      /\b\/sc:[\w-]+\b/i,  // All /sc: commands
+      /\b(business panel|expert analysis|strategic review)\b/i,
+      /\b(mcp:|context7|magic|morphllm|playwright|serena)\b/i,
+      /\b(skill:|capability:|expertise:)\b/i
+    ],
+    route: "anthropic,claude-sonnet-4-latest",
+    strategy: "quality",
+    fallbacks: ["openai,gpt-4o", "gemini,gemini-2.5-pro"],
+    priority: "highest"
+  }
+};
+
+// Provider capabilities and costs
+const PROVIDER_PROFILES = {
+  openai: {
+    costTier: "medium",
+    performanceTier: "high",
+    qualityTier: "high",
+    speedTier: "medium",
+    specialties: ["coding", "reasoning", "math"],
+    models: {
+      "gpt-4o": { cost: 1, speed: 1, quality: 1, capability: "general" },
+      "gpt-4o-mini": { cost: 0.1, speed: 1.5, quality: 0.8, capability: "general" },
+      "gpt-4-turbo": { cost: 2, speed: 1, quality: 1.1, capability: "general" },
+      "o1": { cost: 3, speed: 0.3, quality: 1.3, capability: "reasoning" },
+      "o1-mini": { cost: 0.6, speed: 0.5, quality: 1.1, capability: "reasoning" }
+    }
+  },
+  anthropic: {
+    costTier: "high",
+    performanceTier: "high",
+    qualityTier: "highest",
+    speedTier: "medium",
+    specialties: ["reasoning", "writing", "analysis", "safety"],
+    models: {
+      "claude-sonnet-4-latest": { cost: 3, speed: 1, quality: 1.3, capability: "general" },
+      "claude-3-5-sonnet-latest": { cost: 0.6, speed: 1.2, quality: 1, capability: "general" },
+      "claude-3-5-haiku-latest": { cost: 0.2, speed: 2, quality: 0.7, capability: "general" }
+    }
+  },
+  gemini: {
+    costTier: "low",
+    performanceTier: "high",
+    qualityTier: "medium",
+    speedTier: "highest",
+    specialties: ["speed", "context", "multilingual"],
+    models: {
+      "gemini-2.5-flash": { cost: 0.01, speed: 3, quality: 0.8, capability: "general" },
+      "gemini-2.5-pro": { cost: 0.25, speed: 1.5, quality: 1, capability: "general" },
+      "gemini-2.0-flash": { cost: 0.01, speed: 3, quality: 0.7, capability: "general" }
+    }
+  },
+  qwen: {
+    costTier: "low",
+    performanceTier: "medium",
+    qualityTier: "medium",
+    speedTier: "high",
+    specialties: ["cost-effective", "multilingual", "coding"],
+    models: {
+      "qwen-plus": { cost: 0.1, speed: 2, quality: 0.8, capability: "general" },
+      "qwen-max": { cost: 0.4, speed: 1.5, quality: 1, capability: "general" },
+      "qwen-turbo": { cost: 0.03, speed: 2.5, quality: 0.6, capability: "general" },
+      "qwen3-coder-plus": { cost: 0.4, speed: 1.8, quality: 1.1, capability: "coding" }
+    }
+  },
+  glm: {
+    costTier: "low",
+    performanceTier: "medium",
+    qualityTier: "medium",
+    speedTier: "high",
+    specialties: ["chinese", "multilingual", "translation"],
+    models: {
+      "glm-4.6": { cost: 0.1, speed: 2, quality: 0.8, capability: "multilingual" },
+      "glm-4.5": { cost: 0.1, speed: 2, quality: 0.8, capability: "multilingual" },
+      "glm-4-plus": { cost: 0.2, speed: 1.5, quality: 0.9, capability: "general" }
+    }
+  },
+  openrouter: {
+    costTier: "variable",
+    performanceTier: "variable",
+    qualityTier: "variable",
+    speedTier: "variable",
+    specialties: ["variety", "fallback", "specialized"],
+    models: {
+      "deepseek/deepseek-chat": { cost: 0.14, speed: 1.5, quality: 0.9, capability: "coding" },
+      "meta-llama/llama-3.2-3b-instruct": { cost: 0.1, speed: 2.5, quality: 0.7, capability: "general" }
+    }
+  }
+};
+
+// Smart routing strategies
+const ROUTING_STRATEGIES = {
+  cost: {
+    name: "Cost-Optimized",
+    select: (candidates) => {
+      // Sort by cost (ascending)
+      return candidates.sort((a, b) => {
+        const costA = PROVIDER_PROFILES[a.provider]?.models[a.model]?.cost || 999;
+        const costB = PROVIDER_PROFILES[b.provider]?.models[b.model]?.cost || 999;
+        return costA - costB;
+      });
+    }
+  },
+
+  performance: {
+    name: "Performance-Optimized",
+    select: (candidates) => {
+      // Sort by speed (descending)
+      return candidates.sort((a, b) => {
+        const speedA = PROVIDER_PROFILES[a.provider]?.models[a.model]?.speed || 0;
+        const speedB = PROVIDER_PROFILES[b.provider]?.models[b.model]?.speed || 0;
+        return speedB - speedA;
+      });
+    }
+  },
+
+  quality: {
+    name: "Quality-Optimized",
+    select: (candidates) => {
+      // Sort by quality (descending)
+      return candidates.sort((a, b) => {
+        const qualityA = PROVIDER_PROFILES[a.provider]?.models[a.model]?.quality || 0;
+        const qualityB = PROVIDER_PROFILES[b.provider]?.models[b.model]?.quality || 0;
+        return qualityB - qualityA;
+      });
+    }
+  },
+
+  adaptive: {
+    name: "Adaptive",
+    select: (candidates, metrics, intent) => {
+      // Consider historical performance
+      const scored = candidates.map(candidate => {
+        let score = 0;
+        const profile = PROVIDER_PROFILES[candidate.provider]?.models[candidate.model];
+
+        if (!profile) return { ...candidate, score: 0 };
+
+        // Base score from profile
+        score += profile.quality * 2;
+        score += profile.speed;
+        score -= profile.cost * 0.5;
+
+        // Historical performance adjustment
+        const historical = getHistoricalPerformance(candidate.provider, candidate.model, intent, metrics);
+        score += historical.performanceModifier;
+
+        return { ...candidate, score };
+      });
+
+      return scored.sort((a, b) => b.score - a.score);
+    }
+  }
+};
+
+// Get historical performance for provider/model combination
+function getHistoricalPerformance(provider, model, intent, metrics) {
+  const key = `${provider}/${model}`;
+  const intentMetrics = metrics.intents[intent] || {};
+  const providerMetrics = metrics.providers[key] || {};
+
+  // Default performance
+  const performance = {
+    avgLatency: 3000,
+    successRate: 0.95,
+    costPerRequest: 0.01,
+    performanceModifier: 0
+  };
+
+  // Apply historical data
+  if (providerMetrics.avgLatency) {
+    performance.avgLatency = providerMetrics.avgLatency;
+    performance.performanceModifier -= (performance.avgLatency - 3000) / 10000; // Penalty for slow
+  }
+
+  if (providerMetrics.successRate) {
+    performance.successRate = providerMetrics.successRate;
+    performance.performanceModifier += (performance.successRate - 0.95) * 10; // Bonus for reliability
+  }
+
+  if (intentMetrics.avgLatency) {
+    performance.performanceModifier -= (intentMetrics.avgLatency - 3000) / 8000;
+  }
+
+  return performance;
+}
+
+// Extract content from request
+function extractContent(req) {
+  const messages = req.body?.messages || [];
+  return messages
+    .filter(m => m.role === "user" || m.role === "system")
+    .map(m => typeof m.content === "string" ? m.content : JSON.stringify(m.content))
+    .join(" ")
+    .slice(0, 3000); // Limit for performance
+}
+
+// Detect intent with scoring
+function detectIntent(content) {
+  const scores = {};
+
+  // Calculate scores for each intent
+  for (const [intent, config] of Object.entries(INTENTS)) {
+    scores[intent] = config.patterns.reduce((score, pattern) => {
+      const matches = (content.match(pattern) || []).length;
+      return score + matches;
+    }, 0);
+
+    // Apply priority weighting
+    const priorityWeight = {
+      highest: 1.5,
+      high: 1.3,
+      medium: 1.0,
+      low: 0.8
+    };
+    scores[intent] *= priorityWeight[config.priority] || 1.0;
+  }
+
+  // Sort by score and filter out zero scores
+  const sorted = Object.entries(scores)
+    .filter(([_, score]) => score > 0)
+    .sort((a, b) => b[1] - a[1]);
+
+  return sorted.length > 0 ? {
+    intent: sorted[0][0],
+    confidence: sorted[0][1],
+    alternatives: sorted.slice(1, 3)
+  } : null;
+}
+
+// Get available providers from config
+function getAvailableProviders(config) {
+  const providers = [];
+
+  if (config.Providers) {
+    config.Providers.forEach(provider => {
+      provider.models.forEach(model => {
+        providers.push({
+          provider: provider.name,
+          model,
+          route: `${provider.name},${model}`
+        });
+      });
+    });
+  }
+
+  return providers;
+}
+
+// Generate candidates for routing
+function generateCandidates(detectedIntent, availableProviders) {
+  const candidates = [];
+
+  if (!detectedIntent || !availableProviders.length) {
+    return candidates;
+  }
+
+  // Add primary route
+  const primary = availableProviders.find(p => p.route === detectedIntent.route);
+  if (primary) {
+    candidates.push({
+      ...primary,
+      source: 'primary',
+      reason: 'Intent match'
+    });
+  }
+
+  // Add fallbacks
+  if (INTENTS[detectedIntent.intent]?.fallbacks) {
+    INTENTS[detectedIntent.intent].fallbacks.forEach((fallbackRoute, index) => {
+      const fallback = availableProviders.find(p => p.route === fallbackRoute);
+      if (fallback) {
+        candidates.push({
+          ...fallback,
+          source: 'fallback',
+          priority: index + 1,
+          reason: `Fallback ${index + 1}`
+        });
+      }
+    });
+  }
+
+  return candidates;
+}
+
+// Apply routing strategy
+function applyRoutingStrategy(candidates, strategy, metrics, intent) {
+  if (!candidates.length) return null;
+
+  const router = ROUTING_STRATEGIES[strategy] || ROUTING_STRATEGIES.adaptive;
+  const sorted = router.select(candidates, metrics, intent?.intent);
+
+  return sorted[0];
+}
+
+// Main router function
+module.exports = async function smartRouter(req, config) {
+  const metrics = loadMetrics();
+  const startTime = Date.now();
+
+  try {
+    // Extract content and detect intent
+    const content = extractContent(req);
+    const detectedIntent = detectIntent(content);
+
+    // Get available providers
+    const availableProviders = getAvailableProviders(config);
+
+    // Generate routing candidates
+    const candidates = generateCandidates(detectedIntent, availableProviders);
+
+    // Determine routing strategy
+    let strategy = 'adaptive'; // Default
+
+    // Override strategy based on intent or config
+    if (detectedIntent && INTENTS[detectedIntent.intent]) {
+      strategy = INTENTS[detectedIntent.intent].strategy;
+    }
+
+    // Check for cost/quality optimization settings in config
+    if (config.CostOptimization?.enabled) {
+      strategy = 'cost';
+    } else if (config.PerformanceOptimization?.enabled) {
+      strategy = 'performance';
+    } else if (config.QualityOptimization?.enabled) {
+      strategy = 'quality';
+    }
+
+    // Apply routing strategy
+    const selected = applyRoutingStrategy(candidates, strategy, metrics, detectedIntent);
+
+    if (selected) {
+      const latency = Date.now() - startTime;
+
+      // Log routing decision
+      console.log(`[SmartRouter] ${detectedIntent?.intent || 'unknown'} → ${selected.route} (${strategy}, ${selected.reason})`);
+
+      // Update metrics
+      updateRoutingMetrics(metrics, detectedIntent, selected, latency);
+
+      return selected.route;
+    }
+
+    // Ultimate fallback - use config default or first available
+    const fallback = config.Router?.default || availableProviders[0]?.route;
+    console.log(`[SmartRouter] No match → ${fallback}`);
+
+    return fallback;
+
+  } catch (error) {
+    console.error('[SmartRouter] Error:', error);
+    return config.Router?.default || null;
+  } finally {
+    saveMetrics(metrics);
+  }
+};
+
+// Update routing metrics
+function updateRoutingMetrics(metrics, intent, selected, latency) {
+  const timestamp = new Date().toISOString();
+  const key = `${selected.provider}/${selected.model}`;
+
+  // Track request
+  metrics.requests.push({
+    timestamp,
+    intent: intent?.intent || 'unknown',
+    provider: selected.provider,
+    model: selected.model,
+    latency,
+    source: selected.source
+  });
+
+  // Update provider metrics
+  if (!metrics.providers[key]) {
+    metrics.providers[key] = {
+      requests: 0,
+      totalLatency: 0,
+      minLatency: Infinity,
+      maxLatency: 0,
+      errors: 0
+    };
+  }
+
+  const providerMetrics = metrics.providers[key];
+  providerMetrics.requests++;
+  providerMetrics.totalLatency += latency;
+  providerMetrics.minLatency = Math.min(providerMetrics.minLatency, latency);
+  providerMetrics.maxLatency = Math.max(providerMetrics.maxLatency, latency);
+
+  // Update intent metrics
+  if (intent?.intent) {
+    if (!metrics.intents[intent.intent]) {
+      metrics.intents[intent.intent] = {
+        requests: 0,
+        totalLatency: 0,
+        avgLatency: 0
+      };
+    }
+
+    const intentMetrics = metrics.intents[intent.intent];
+    intentMetrics.requests++;
+    intentMetrics.totalLatency += latency;
+    intentMetrics.avgLatency = intentMetrics.totalLatency / intentMetrics.requests;
+  }
+}
+
+// Export utilities for testing
+module.exports.INTENTS = INTENTS;
+module.exports.PROVIDER_PROFILES = PROVIDER_PROFILES;
+module.exports.ROUTING_STRATEGIES = ROUTING_STRATEGIES;