opc-agent 4.1.0 → 4.1.2

package/src/memory/user-profiler.ts

@@ -1,215 +1,215 @@
-import type { Message } from '../core/types';
-
-export interface UserProfile {
-  preferences: Record<string, string>;
-  communication_style: string;
-  expertise_areas: string[];
-  common_requests: string[];
-  last_updated: number;
-}
-
-const EMPTY_PROFILE: UserProfile = {
-  preferences: {},
-  communication_style: 'unknown',
-  expertise_areas: [],
-  common_requests: [],
-  last_updated: 0,
-};
-
-/**
- * DeepBrain-powered user profiling from conversation signals.
- */
-export class UserProfiler {
-  private observationCount = 0;
-  private signals = {
-    languages: new Map<string, number>(),
-    techKeywords: new Set<string>(),
-    styleSignals: { brief: 0, detailed: 0, formal: 0, casual: 0 },
-    requestTypes: new Map<string, number>(),
-  };
-
-  private readonly LEARN_INTERVAL = 20;
-
-  /**
-   * Detect language mix of a text.
-   */
-  private detectLanguage(text: string): string {
-    let cn = 0, en = 0;
-    for (const char of text) {
-      const code = char.codePointAt(0) ?? 0;
-      if (code >= 0x4e00 && code <= 0x9fff) cn++;
-      else if ((code >= 0x41 && code <= 0x5a) || (code >= 0x61 && code <= 0x7a)) en++;
-    }
-    if (cn > 0 && en > 0) return 'mixed';
-    if (cn > en) return 'chinese';
-    return 'english';
-  }
-
-  /**
-   * Detect technical level from content.
-   */
-  private detectTechLevel(text: string): string {
-    const expertTerms = /\b(kubernetes|k8s|microservice|architecture|distributed|consensus|raft|paxos|sharding|vector db|embedding|fine-?tun|transformer|CUDA|inference|quantiz|LoRA|RAG)\b/i;
-    const intermediateTerms = /\b(API|REST|GraphQL|Docker|CI\/CD|database|deploy|cloud|React|TypeScript|Python|async|cache|redis|nginx)\b/i;
-
-    if (expertTerms.test(text)) return 'expert';
-    if (intermediateTerms.test(text)) return 'intermediate';
-    return 'beginner';
-  }
-
-  /**
-   * Detect communication style.
-   */
-  private detectStyle(text: string): void {
-    if (text.length < 50) this.signals.styleSignals.brief++;
-    else if (text.length > 300) this.signals.styleSignals.detailed++;
-
-    if (/\b(please|kindly|would you|could you|请问|烦请|麻烦)\b/i.test(text)) {
-      this.signals.styleSignals.formal++;
-    }
-    if (/[!]{2,}|lol|haha|😂|🤣|哈哈|牛|666|👍/i.test(text)) {
-      this.signals.styleSignals.casual++;
-    }
-  }
-
-  /**
-   * Extract domain keywords.
-   */
-  private extractDomainKeywords(text: string): void {
-    const techWords = text.match(/\b[A-Z][a-zA-Z]{2,}(?:\.js|\.ts|\.py)?\b/g) ?? [];
-    techWords.forEach(w => this.signals.techKeywords.add(w));
-    // Chinese tech terms
-    const cnTerms = text.match(/(?:人工智能|机器学习|深度学习|大模型|微服务|架构|部署|运维|前端|后端|数据库|缓存|分布式)/g) ?? [];
-    cnTerms.forEach(w => this.signals.techKeywords.add(w));
-  }
-
-  /**
-   * Classify request type.
-   */
-  private classifyRequest(text: string): string {
-    if (/\b(how to|怎么|如何|how do)\b/i.test(text)) return 'how-to';
-    if (/\b(why|为什么|原因)\b/i.test(text)) return 'explanation';
-    if (/\b(fix|error|bug|报错|出错|failed)\b/i.test(text)) return 'debugging';
-    if (/\b(review|评审|看看|check)\b/i.test(text)) return 'review';
-    if (/\b(create|build|write|写|创建|生成)\b/i.test(text)) return 'creation';
-    return 'general';
-  }
-
-  /**
-   * Observe a user message and accumulate profile signals.
-   */
-  async observe(message: Message, brain?: any): Promise<void> {
-    if (message.role !== 'user') return;
-
-    const text = message.content;
-
-    // Language
-    const lang = this.detectLanguage(text);
-    this.signals.languages.set(lang, (this.signals.languages.get(lang) ?? 0) + 1);
-
-    // Style
-    this.detectStyle(text);
-
-    // Domain keywords
-    this.extractDomainKeywords(text);
-
-    // Request type
-    const reqType = this.classifyRequest(text);
-    this.signals.requestTypes.set(reqType, (this.signals.requestTypes.get(reqType) ?? 0) + 1);
-
-    this.observationCount++;
-
-    // Periodically persist to brain
-    if (brain?.learn && this.observationCount % this.LEARN_INTERVAL === 0) {
-      const profile = this.buildProfileFromSignals();
-      try {
-        await brain.learn(JSON.stringify(profile), { insight_type: 'user_profile' });
-      } catch { /* non-critical */ }
-    }
-  }
-
-  private buildProfileFromSignals(): UserProfile {
-    // Language preference
-    let topLang = 'english';
-    let maxCount = 0;
-    for (const [lang, count] of this.signals.languages) {
-      if (count > maxCount) { topLang = lang; maxCount = count; }
-    }
-
-    // Communication style
-    const ss = this.signals.styleSignals;
-    const styles = [
-      ['brief', ss.brief], ['detailed', ss.detailed],
-      ['formal', ss.formal], ['casual', ss.casual],
-    ] as const;
-    const topStyle = styles.reduce((a, b) => (b[1] > a[1] ? b : a))[0];
-
-    // Top request types
-    const topRequests = [...this.signals.requestTypes.entries()]
-      .sort((a, b) => b[1] - a[1])
-      .slice(0, 5)
-      .map(([t]) => t);
-
-    return {
-      preferences: { language: topLang },
-      communication_style: topStyle,
-      expertise_areas: [...this.signals.techKeywords].slice(0, 20),
-      common_requests: topRequests,
-      last_updated: Date.now(),
-    };
-  }
-
-  /**
-   * Get user profile, optionally from brain recall.
-   */
-  async getProfile(brain?: any): Promise<UserProfile> {
-    // First try local signals
-    if (this.observationCount > 0) {
-      return this.buildProfileFromSignals();
-    }
-
-    // Fallback to brain recall
-    if (brain?.recall) {
-      try {
-        const results = await brain.recall('user profile preferences style');
-        if (Array.isArray(results) && results.length > 0) {
-          const raw = typeof results[0] === 'string' ? results[0] : results[0].content;
-          return { ...EMPTY_PROFILE, ...JSON.parse(raw) };
-        }
-      } catch { /* ignore parse errors */ }
-    }
-
-    return { ...EMPTY_PROFILE };
-  }
-
-  /**
-   * Enhance a system prompt with user profile context.
-   */
-  enhance(systemPrompt: string, profile: UserProfile): string {
-    const hints: string[] = [];
-
-    if (profile.preferences.language) {
-      const langMap: Record<string, string> = {
-        chinese: 'User prefers Chinese responses.',
-        english: 'User prefers English responses.',
-        mixed: 'User uses mixed Chinese/English. Match their style.',
-      };
-      if (langMap[profile.preferences.language]) hints.push(langMap[profile.preferences.language]);
-    }
-
-    if (profile.communication_style && profile.communication_style !== 'unknown') {
-      hints.push(`User communication style: ${profile.communication_style}.`);
-    }
-
-    if (profile.expertise_areas.length > 0) {
-      hints.push(`User expertise: ${profile.expertise_areas.slice(0, 10).join(', ')}.`);
-    }
-
-    if (profile.common_requests.length > 0) {
-      hints.push(`Common request types: ${profile.common_requests.join(', ')}.`);
-    }
-
-    if (hints.length === 0) return systemPrompt;
-    return `${systemPrompt}\n\n[User Profile] ${hints.join(' ')}`;
-  }
-}
+import type { Message } from '../core/types';
+
+export interface UserProfile {
+  preferences: Record<string, string>;
+  communication_style: string;
+  expertise_areas: string[];
+  common_requests: string[];
+  last_updated: number;
+}
+
+const EMPTY_PROFILE: UserProfile = {
+  preferences: {},
+  communication_style: 'unknown',
+  expertise_areas: [],
+  common_requests: [],
+  last_updated: 0,
+};
+
+/**
+ * DeepBrain-powered user profiling from conversation signals.
+ */
+export class UserProfiler {
+  private observationCount = 0;
+  private signals = {
+    languages: new Map<string, number>(),
+    techKeywords: new Set<string>(),
+    styleSignals: { brief: 0, detailed: 0, formal: 0, casual: 0 },
+    requestTypes: new Map<string, number>(),
+  };
+
+  private readonly LEARN_INTERVAL = 20;
+
+  /**
+   * Detect language mix of a text.
+   */
+  private detectLanguage(text: string): string {
+    let cn = 0, en = 0;
+    for (const char of text) {
+      const code = char.codePointAt(0) ?? 0;
+      if (code >= 0x4e00 && code <= 0x9fff) cn++;
+      else if ((code >= 0x41 && code <= 0x5a) || (code >= 0x61 && code <= 0x7a)) en++;
+    }
+    if (cn > 0 && en > 0) return 'mixed';
+    if (cn > en) return 'chinese';
+    return 'english';
+  }
+
+  /**
+   * Detect technical level from content.
+   */
+  private detectTechLevel(text: string): string {
+    const expertTerms = /\b(kubernetes|k8s|microservice|architecture|distributed|consensus|raft|paxos|sharding|vector db|embedding|fine-?tun|transformer|CUDA|inference|quantiz|LoRA|RAG)\b/i;
+    const intermediateTerms = /\b(API|REST|GraphQL|Docker|CI\/CD|database|deploy|cloud|React|TypeScript|Python|async|cache|redis|nginx)\b/i;
+
+    if (expertTerms.test(text)) return 'expert';
+    if (intermediateTerms.test(text)) return 'intermediate';
+    return 'beginner';
+  }
+
+  /**
+   * Detect communication style.
+   */
+  private detectStyle(text: string): void {
+    if (text.length < 50) this.signals.styleSignals.brief++;
+    else if (text.length > 300) this.signals.styleSignals.detailed++;
+
+    if (/\b(please|kindly|would you|could you|请问|烦请|麻烦)\b/i.test(text)) {
+      this.signals.styleSignals.formal++;
+    }
+    if (/[!]{2,}|lol|haha|😂|🤣|哈哈|牛|666|👍/i.test(text)) {
+      this.signals.styleSignals.casual++;
+    }
+  }
+
+  /**
+   * Extract domain keywords.
+   */
+  private extractDomainKeywords(text: string): void {
+    const techWords = text.match(/\b[A-Z][a-zA-Z]{2,}(?:\.js|\.ts|\.py)?\b/g) ?? [];
+    techWords.forEach(w => this.signals.techKeywords.add(w));
+    // Chinese tech terms
+    const cnTerms = text.match(/(?:人工智能|机器学习|深度学习|大模型|微服务|架构|部署|运维|前端|后端|数据库|缓存|分布式)/g) ?? [];
+    cnTerms.forEach(w => this.signals.techKeywords.add(w));
+  }
+
+  /**
+   * Classify request type.
+   */
+  private classifyRequest(text: string): string {
+    if (/\b(how to|怎么|如何|how do)\b/i.test(text)) return 'how-to';
+    if (/\b(why|为什么|原因)\b/i.test(text)) return 'explanation';
+    if (/\b(fix|error|bug|报错|出错|failed)\b/i.test(text)) return 'debugging';
+    if (/\b(review|评审|看看|check)\b/i.test(text)) return 'review';
+    if (/\b(create|build|write|写|创建|生成)\b/i.test(text)) return 'creation';
+    return 'general';
+  }
+
+  /**
+   * Observe a user message and accumulate profile signals.
+   */
+  async observe(message: Message, brain?: any): Promise<void> {
+    if (message.role !== 'user') return;
+
+    const text = message.content;
+
+    // Language
+    const lang = this.detectLanguage(text);
+    this.signals.languages.set(lang, (this.signals.languages.get(lang) ?? 0) + 1);
+
+    // Style
+    this.detectStyle(text);
+
+    // Domain keywords
+    this.extractDomainKeywords(text);
+
+    // Request type
+    const reqType = this.classifyRequest(text);
+    this.signals.requestTypes.set(reqType, (this.signals.requestTypes.get(reqType) ?? 0) + 1);
+
+    this.observationCount++;
+
+    // Periodically persist to brain
+    if (brain?.learn && this.observationCount % this.LEARN_INTERVAL === 0) {
+      const profile = this.buildProfileFromSignals();
+      try {
+        await brain.learn(JSON.stringify(profile), { insight_type: 'user_profile' });
+      } catch { /* non-critical */ }
+    }
+  }
+
+  private buildProfileFromSignals(): UserProfile {
+    // Language preference
+    let topLang = 'english';
+    let maxCount = 0;
+    for (const [lang, count] of this.signals.languages) {
+      if (count > maxCount) { topLang = lang; maxCount = count; }
+    }
+
+    // Communication style
+    const ss = this.signals.styleSignals;
+    const styles = [
+      ['brief', ss.brief], ['detailed', ss.detailed],
+      ['formal', ss.formal], ['casual', ss.casual],
+    ] as const;
+    const topStyle = styles.reduce((a, b) => (b[1] > a[1] ? b : a))[0];
+
+    // Top request types
+    const topRequests = [...this.signals.requestTypes.entries()]
+      .sort((a, b) => b[1] - a[1])
+      .slice(0, 5)
+      .map(([t]) => t);
+
+    return {
+      preferences: { language: topLang },
+      communication_style: topStyle,
+      expertise_areas: [...this.signals.techKeywords].slice(0, 20),
+      common_requests: topRequests,
+      last_updated: Date.now(),
+    };
+  }
+
+  /**
+   * Get user profile, optionally from brain recall.
+   */
+  async getProfile(brain?: any): Promise<UserProfile> {
+    // First try local signals
+    if (this.observationCount > 0) {
+      return this.buildProfileFromSignals();
+    }
+
+    // Fallback to brain recall
+    if (brain?.recall) {
+      try {
+        const results = await brain.recall('user profile preferences style');
+        if (Array.isArray(results) && results.length > 0) {
+          const raw = typeof results[0] === 'string' ? results[0] : results[0].content;
+          return { ...EMPTY_PROFILE, ...JSON.parse(raw) };
+        }
+      } catch { /* ignore parse errors */ }
+    }
+
+    return { ...EMPTY_PROFILE };
+  }
+
+  /**
+   * Enhance a system prompt with user profile context.
+   */
+  enhance(systemPrompt: string, profile: UserProfile): string {
+    const hints: string[] = [];
+
+    if (profile.preferences.language) {
+      const langMap: Record<string, string> = {
+        chinese: 'User prefers Chinese responses.',
+        english: 'User prefers English responses.',
+        mixed: 'User uses mixed Chinese/English. Match their style.',
+      };
+      if (langMap[profile.preferences.language]) hints.push(langMap[profile.preferences.language]);
+    }
+
+    if (profile.communication_style && profile.communication_style !== 'unknown') {
+      hints.push(`User communication style: ${profile.communication_style}.`);
+    }
+
+    if (profile.expertise_areas.length > 0) {
+      hints.push(`User expertise: ${profile.expertise_areas.slice(0, 10).join(', ')}.`);
+    }
+
+    if (profile.common_requests.length > 0) {
+      hints.push(`Common request types: ${profile.common_requests.join(', ')}.`);
+    }
+
+    if (hints.length === 0) return systemPrompt;
+    return `${systemPrompt}\n\n[User Profile] ${hints.join(' ')}`;
+  }
+}

package/src/plugins/content-filter.ts

@@ -1,23 +1,23 @@
-import type { Plugin } from './index';
-
-/**
- * Content filter plugin — filters messages containing blocked keywords.
- */
-export function createContentFilterPlugin(blocklist: string[] = []): Plugin {
-  return {
-    name: 'content-filter',
-    version: '1.0.0',
-    description: 'Filter inappropriate content',
-    onMessage: async (msg: any, next: (m: any) => Promise<any>) => {
-      const content = (msg.content || '').toLowerCase();
-      for (const word of blocklist) {
-        if (content.includes(word.toLowerCase())) {
-          throw new Error(`Content blocked: message contains prohibited content`);
-        }
-      }
-      return next(msg);
-    },
-  };
-}
-
-export const contentFilterPlugin: Plugin = createContentFilterPlugin([]);
+import type { Plugin } from './index';
+
+/**
+ * Content filter plugin — filters messages containing blocked keywords.
+ */
+export function createContentFilterPlugin(blocklist: string[] = []): Plugin {
+  return {
+    name: 'content-filter',
+    version: '1.0.0',
+    description: 'Filter inappropriate content',
+    onMessage: async (msg: any, next: (m: any) => Promise<any>) => {
+      const content = (msg.content || '').toLowerCase();
+      for (const word of blocklist) {
+        if (content.includes(word.toLowerCase())) {
+          throw new Error(`Content blocked: message contains prohibited content`);
+        }
+      }
+      return next(msg);
+    },
+  };
+}
+
+export const contentFilterPlugin: Plugin = createContentFilterPlugin([]);

package/src/plugins/logger.ts

@@ -1,18 +1,18 @@
-import type { Plugin } from './index';
-
-/**
- * Logger plugin — logs all messages and responses via middleware chain.
- */
-export const loggerPlugin: Plugin = {
-  name: 'logger',
-  version: '1.0.0',
-  description: 'Log all messages and responses',
-  onMessage: async (msg: any, next: (m: any) => Promise<any>) => {
-    console.log(`[${new Date().toISOString()}] IN: ${msg.content?.slice(0, 100)}`);
-    return next(msg);
-  },
-  onResponse: async (res: any, next: (r: any) => Promise<any>) => {
-    console.log(`[${new Date().toISOString()}] OUT: ${res.content?.slice(0, 100)}`);
-    return next(res);
-  },
-};
+import type { Plugin } from './index';
+
+/**
+ * Logger plugin — logs all messages and responses via middleware chain.
+ */
+export const loggerPlugin: Plugin = {
+  name: 'logger',
+  version: '1.0.0',
+  description: 'Log all messages and responses',
+  onMessage: async (msg: any, next: (m: any) => Promise<any>) => {
+    console.log(`[${new Date().toISOString()}] IN: ${msg.content?.slice(0, 100)}`);
+    return next(msg);
+  },
+  onResponse: async (res: any, next: (r: any) => Promise<any>) => {
+    console.log(`[${new Date().toISOString()}] OUT: ${res.content?.slice(0, 100)}`);
+    return next(res);
+  },
+};

package/src/plugins/rate-limiter.ts

@@ -1,38 +1,38 @@
-import type { Plugin } from './index';
-
-/**
- * Rate limiter plugin — limits messages per session using in-memory sliding window.
- */
-export function createRateLimiterPlugin(maxPerMinute = 60): Plugin {
-  const sessionTimestamps: Map<string, number[]> = new Map();
-
-  return {
-    name: 'rate-limiter',
-    version: '1.0.0',
-    description: `Rate limit messages per session (${maxPerMinute}/min)`,
-    onMessage: async (msg: any, next: (m: any) => Promise<any>) => {
-      const sessionId = msg.metadata?.sessionId || msg.id || 'default';
-      const now = Date.now();
-      const windowStart = now - 60_000;
-
-      if (!sessionTimestamps.has(sessionId)) {
-        sessionTimestamps.set(sessionId, []);
-      }
-      const timestamps = sessionTimestamps.get(sessionId)!;
-
-      // Remove expired timestamps
-      while (timestamps.length > 0 && timestamps[0] < windowStart) {
-        timestamps.shift();
-      }
-
-      if (timestamps.length >= maxPerMinute) {
-        throw new Error(`Rate limit exceeded: ${maxPerMinute} messages per minute`);
-      }
-
-      timestamps.push(now);
-      return next(msg);
-    },
-  };
-}
-
-export const rateLimiterPlugin: Plugin = createRateLimiterPlugin(60);
+import type { Plugin } from './index';
+
+/**
+ * Rate limiter plugin — limits messages per session using in-memory sliding window.
+ */
+export function createRateLimiterPlugin(maxPerMinute = 60): Plugin {
+  const sessionTimestamps: Map<string, number[]> = new Map();
+
+  return {
+    name: 'rate-limiter',
+    version: '1.0.0',
+    description: `Rate limit messages per session (${maxPerMinute}/min)`,
+    onMessage: async (msg: any, next: (m: any) => Promise<any>) => {
+      const sessionId = msg.metadata?.sessionId || msg.id || 'default';
+      const now = Date.now();
+      const windowStart = now - 60_000;
+
+      if (!sessionTimestamps.has(sessionId)) {
+        sessionTimestamps.set(sessionId, []);
+      }
+      const timestamps = sessionTimestamps.get(sessionId)!;
+
+      // Remove expired timestamps
+      while (timestamps.length > 0 && timestamps[0] < windowStart) {
+        timestamps.shift();
+      }
+
+      if (timestamps.length >= maxPerMinute) {
+        throw new Error(`Rate limit exceeded: ${maxPerMinute} messages per minute`);
+      }
+
+      timestamps.push(now);
+      return next(msg);
+    },
+  };
+}
+
+export const rateLimiterPlugin: Plugin = createRateLimiterPlugin(60);