opc-agent 1.1.1 → 1.1.2

package/src/tools/text-analysis.ts

@@ -1,116 +1,116 @@
-import type { MCPTool, MCPToolResult } from './mcp';
-
-/**
- * Text Analysis Tool — v0.8.0
- * Summarize, translate (stub), sentiment analysis as an LLM function tool.
- * Note: For production, connect to actual LLM/NLP APIs. This provides basic built-in analysis.
- */
-export const TextAnalysisTool: MCPTool = {
-  name: 'text_analysis',
-  description: 'Analyze text: word count, character count, reading time, keyword extraction, basic sentiment, and language detection.',
-  inputSchema: {
-    type: 'object',
-    properties: {
-      operation: {
-        type: 'string',
-        enum: ['stats', 'keywords', 'sentiment', 'detect_language', 'truncate', 'split_sentences'],
-        description: 'Analysis operation to perform',
-      },
-      text: {
-        type: 'string',
-        description: 'Text to analyze',
-      },
-      maxLength: {
-        type: 'number',
-        description: 'Max length for truncate operation',
-      },
-      topN: {
-        type: 'number',
-        description: 'Number of top keywords to extract (default: 10)',
-      },
-    },
-    required: ['operation', 'text'],
-  },
-
-  async execute(input: Record<string, unknown>): Promise<MCPToolResult> {
-    try {
-      const op = String(input.operation);
-      const text = String(input.text ?? '');
-
-      switch (op) {
-        case 'stats': {
-          const words = text.split(/\s+/).filter(Boolean);
-          const sentences = text.split(/[.!?。！？]+/).filter(Boolean);
-          const readingTimeMin = Math.ceil(words.length / 200);
-          return {
-            content: JSON.stringify({
-              characters: text.length,
-              words: words.length,
-              sentences: sentences.length,
-              paragraphs: text.split(/\n\s*\n/).filter(Boolean).length,
-              readingTimeMinutes: readingTimeMin,
-            }, null, 2),
-          };
-        }
-
-        case 'keywords': {
-          const topN = Number(input.topN ?? 10);
-          const words = text.toLowerCase().replace(/[^\w\s\u4e00-\u9fff]/g, '').split(/\s+/).filter(Boolean);
-          const stopWords = new Set(['the', 'a', 'an', 'is', 'are', 'was', 'were', 'in', 'on', 'at', 'to', 'for', 'of', 'and', 'or', 'but', 'not', 'with', 'this', 'that', 'it', 'be', 'has', 'have', 'had', 'do', 'does', 'did', 'will', 'would', 'can', 'could', 'should', 'may', 'might', 'from', 'by', 'as', 'if', 'so', 'than']);
-          const freq: Record<string, number> = {};
-          for (const w of words) {
-            if (w.length < 2 || stopWords.has(w)) continue;
-            freq[w] = (freq[w] ?? 0) + 1;
-          }
-          const sorted = Object.entries(freq).sort((a, b) => b[1] - a[1]).slice(0, topN);
-          return { content: JSON.stringify(sorted.map(([word, count]) => ({ word, count })), null, 2) };
-        }
-
-        case 'sentiment': {
-          // Basic lexicon-based sentiment
-          const positiveWords = new Set(['good', 'great', 'excellent', 'amazing', 'wonderful', 'fantastic', 'love', 'happy', 'best', 'awesome', 'perfect', 'brilliant', 'outstanding', 'superb', 'beautiful', 'nice', 'enjoy', 'like', 'positive', 'success']);
-          const negativeWords = new Set(['bad', 'terrible', 'awful', 'horrible', 'hate', 'worst', 'poor', 'ugly', 'fail', 'wrong', 'sad', 'angry', 'frustrating', 'disappointing', 'negative', 'annoying', 'broken', 'useless', 'boring', 'painful']);
-
-          const words = text.toLowerCase().split(/\s+/);
-          let pos = 0, neg = 0;
-          for (const w of words) {
-            if (positiveWords.has(w)) pos++;
-            if (negativeWords.has(w)) neg++;
-          }
-          const total = pos + neg || 1;
-          const score = (pos - neg) / total; // -1 to 1
-          const label = score > 0.1 ? 'positive' : score < -0.1 ? 'negative' : 'neutral';
-          return {
-            content: JSON.stringify({ score: Math.round(score * 100) / 100, label, positive: pos, negative: neg }, null, 2),
-          };
-        }
-
-        case 'detect_language': {
-          // Simple heuristic
-          const hasChinese = /[\u4e00-\u9fff]/.test(text);
-          const hasJapanese = /[\u3040-\u309f\u30a0-\u30ff]/.test(text);
-          const hasKorean = /[\uac00-\ud7af]/.test(text);
-          const hasArabic = /[\u0600-\u06ff]/.test(text);
-          const lang = hasChinese ? 'zh' : hasJapanese ? 'ja' : hasKorean ? 'ko' : hasArabic ? 'ar' : 'en';
-          return { content: JSON.stringify({ detected: lang, confidence: 'heuristic' }) };
-        }
-
-        case 'truncate': {
-          const maxLen = Number(input.maxLength ?? 100);
-          const truncated = text.length > maxLen ? text.slice(0, maxLen) + '...' : text;
-          return { content: truncated };
-        }
-
-        case 'split_sentences': {
-          const sentences = text.match(/[^.!?。！？]+[.!?。！？]+/g) ?? [text];
-          return { content: JSON.stringify(sentences.map((s) => s.trim()), null, 2) };
-        }
-
-        default:
-          return { content: `Unknown operation: ${op}`, isError: true };
-      }
-    } catch (err) {
-      return { content: `Error: ${(err as Error).message}`, isError: true };
-    }
-  },
-};
+import type { MCPTool, MCPToolResult } from './mcp';
+
+/**
+ * Text Analysis Tool — v0.8.0
+ * Summarize, translate (stub), sentiment analysis as an LLM function tool.
+ * Note: For production, connect to actual LLM/NLP APIs. This provides basic built-in analysis.
+ */
+export const TextAnalysisTool: MCPTool = {
+  name: 'text_analysis',
+  description: 'Analyze text: word count, character count, reading time, keyword extraction, basic sentiment, and language detection.',
+  inputSchema: {
+    type: 'object',
+    properties: {
+      operation: {
+        type: 'string',
+        enum: ['stats', 'keywords', 'sentiment', 'detect_language', 'truncate', 'split_sentences'],
+        description: 'Analysis operation to perform',
+      },
+      text: {
+        type: 'string',
+        description: 'Text to analyze',
+      },
+      maxLength: {
+        type: 'number',
+        description: 'Max length for truncate operation',
+      },
+      topN: {
+        type: 'number',
+        description: 'Number of top keywords to extract (default: 10)',
+      },
+    },
+    required: ['operation', 'text'],
+  },
+
+  async execute(input: Record<string, unknown>): Promise<MCPToolResult> {
+    try {
+      const op = String(input.operation);
+      const text = String(input.text ?? '');
+
+      switch (op) {
+        case 'stats': {
+          const words = text.split(/\s+/).filter(Boolean);
+          const sentences = text.split(/[.!?。！？]+/).filter(Boolean);
+          const readingTimeMin = Math.ceil(words.length / 200);
+          return {
+            content: JSON.stringify({
+              characters: text.length,
+              words: words.length,
+              sentences: sentences.length,
+              paragraphs: text.split(/\n\s*\n/).filter(Boolean).length,
+              readingTimeMinutes: readingTimeMin,
+            }, null, 2),
+          };
+        }
+
+        case 'keywords': {
+          const topN = Number(input.topN ?? 10);
+          const words = text.toLowerCase().replace(/[^\w\s\u4e00-\u9fff]/g, '').split(/\s+/).filter(Boolean);
+          const stopWords = new Set(['the', 'a', 'an', 'is', 'are', 'was', 'were', 'in', 'on', 'at', 'to', 'for', 'of', 'and', 'or', 'but', 'not', 'with', 'this', 'that', 'it', 'be', 'has', 'have', 'had', 'do', 'does', 'did', 'will', 'would', 'can', 'could', 'should', 'may', 'might', 'from', 'by', 'as', 'if', 'so', 'than']);
+          const freq: Record<string, number> = {};
+          for (const w of words) {
+            if (w.length < 2 || stopWords.has(w)) continue;
+            freq[w] = (freq[w] ?? 0) + 1;
+          }
+          const sorted = Object.entries(freq).sort((a, b) => b[1] - a[1]).slice(0, topN);
+          return { content: JSON.stringify(sorted.map(([word, count]) => ({ word, count })), null, 2) };
+        }
+
+        case 'sentiment': {
+          // Basic lexicon-based sentiment
+          const positiveWords = new Set(['good', 'great', 'excellent', 'amazing', 'wonderful', 'fantastic', 'love', 'happy', 'best', 'awesome', 'perfect', 'brilliant', 'outstanding', 'superb', 'beautiful', 'nice', 'enjoy', 'like', 'positive', 'success']);
+          const negativeWords = new Set(['bad', 'terrible', 'awful', 'horrible', 'hate', 'worst', 'poor', 'ugly', 'fail', 'wrong', 'sad', 'angry', 'frustrating', 'disappointing', 'negative', 'annoying', 'broken', 'useless', 'boring', 'painful']);
+
+          const words = text.toLowerCase().split(/\s+/);
+          let pos = 0, neg = 0;
+          for (const w of words) {
+            if (positiveWords.has(w)) pos++;
+            if (negativeWords.has(w)) neg++;
+          }
+          const total = pos + neg || 1;
+          const score = (pos - neg) / total; // -1 to 1
+          const label = score > 0.1 ? 'positive' : score < -0.1 ? 'negative' : 'neutral';
+          return {
+            content: JSON.stringify({ score: Math.round(score * 100) / 100, label, positive: pos, negative: neg }, null, 2),
+          };
+        }
+
+        case 'detect_language': {
+          // Simple heuristic
+          const hasChinese = /[\u4e00-\u9fff]/.test(text);
+          const hasJapanese = /[\u3040-\u309f\u30a0-\u30ff]/.test(text);
+          const hasKorean = /[\uac00-\ud7af]/.test(text);
+          const hasArabic = /[\u0600-\u06ff]/.test(text);
+          const lang = hasChinese ? 'zh' : hasJapanese ? 'ja' : hasKorean ? 'ko' : hasArabic ? 'ar' : 'en';
+          return { content: JSON.stringify({ detected: lang, confidence: 'heuristic' }) };
+        }
+
+        case 'truncate': {
+          const maxLen = Number(input.maxLength ?? 100);
+          const truncated = text.length > maxLen ? text.slice(0, maxLen) + '...' : text;
+          return { content: truncated };
+        }
+
+        case 'split_sentences': {
+          const sentences = text.match(/[^.!?。！？]+[.!?。！？]+/g) ?? [text];
+          return { content: JSON.stringify(sentences.map((s) => s.trim()), null, 2) };
+        }
+
+        default:
+          return { content: `Unknown operation: ${op}`, isError: true };
+      }
+    } catch (err) {
+      return { content: `Error: ${(err as Error).message}`, isError: true };
+    }
+  },
+};

package/templates/Dockerfile

@@ -1,15 +1,15 @@
-FROM node:22-alpine AS base
-WORKDIR /app
-
-# Install dependencies
-COPY package.json package-lock.json* ./
-RUN npm ci --production 2>/dev/null || npm install --production
-
-# Copy agent files
-COPY oad.yaml ./
-COPY .env* ./
-COPY prompts/ ./prompts/ 2>/dev/null || true
-
-EXPOSE 3000
-
-CMD ["npx", "opc", "run"]
+FROM node:22-alpine AS base
+WORKDIR /app
+
+# Install dependencies
+COPY package.json package-lock.json* ./
+RUN npm ci --production 2>/dev/null || npm install --production
+
+# Copy agent files
+COPY oad.yaml ./
+COPY .env* ./
+COPY prompts/ ./prompts/ 2>/dev/null || true
+
+EXPOSE 3000
+
+CMD ["npx", "opc", "run"]

package/templates/code-reviewer/README.md

@@ -1,27 +1,27 @@
-# Code Reviewer Template
-
-AI-powered code reviewer that catches bugs and suggests improvements.
-
-## Features
-- **Bug Detection**: Finds potential bugs and security issues
-- **Style Checks**: Enforces coding best practices
-- **Improvement Suggestions**: Actionable refactoring advice
-- **Severity Ratings**: 🔴 Critical | 🟡 Warning | 🔵 Info
-
-## Quick Start
-```bash
-opc init my-reviewer --template code-reviewer
-cd my-reviewer
-opc run
-```
-
-## Usage
-Send code via the web or WebSocket channel:
-```json
-{
-  "message": "Review this:\n```python\ndef calc(x):\n    return eval(x)\n```"
-}
-```
-
-## Configuration
-Customize the system prompt in `oad.yaml` to focus on specific languages or coding standards.
+# Code Reviewer Template
+
+AI-powered code reviewer that catches bugs and suggests improvements.
+
+## Features
+- **Bug Detection**: Finds potential bugs and security issues
+- **Style Checks**: Enforces coding best practices
+- **Improvement Suggestions**: Actionable refactoring advice
+- **Severity Ratings**: 🔴 Critical | 🟡 Warning | 🔵 Info
+
+## Quick Start
+```bash
+opc init my-reviewer --template code-reviewer
+cd my-reviewer
+opc run
+```
+
+## Usage
+Send code via the web or WebSocket channel:
+```json
+{
+  "message": "Review this:\n```python\ndef calc(x):\n    return eval(x)\n```"
+}
+```
+
+## Configuration
+Customize the system prompt in `oad.yaml` to focus on specific languages or coding standards.

package/templates/code-reviewer/oad.yaml

@@ -1,41 +1,41 @@
-apiVersion: opc/v1
-kind: Agent
-metadata:
-  name: code-reviewer
-  version: 1.0.0
-  description: "AI code reviewer that reviews code and suggests improvements"
-  author: Deepleaper
-  license: Apache-2.0
-  marketplace:
-    certified: false
-    category: developer-tools
-spec:
-  provider:
-    default: deepseek
-    allowed: [openai, deepseek, qwen]
-  model: deepseek-chat
-  systemPrompt: |
-    You are an expert code reviewer. When given code:
-    1. Check for bugs, security issues, and performance problems
-    2. Suggest improvements for readability and maintainability
-    3. Follow language-specific best practices
-    4. Be constructive and explain your reasoning
-    Rate severity: 🔴 Critical | 🟡 Warning | 🔵 Info
-  skills:
-    - name: code-analysis
-      description: "Analyze code for bugs and improvements"
-    - name: style-check
-      description: "Check coding style and best practices"
-  channels:
-    - type: web
-      port: 3000
-    - type: websocket
-      port: 3002
-  memory:
-    shortTerm: true
-    longTerm: false
-  dtv:
-    trust:
-      level: sandbox
-    value:
-      metrics: [reviews_completed, issues_found, suggestions_accepted]
+apiVersion: opc/v1
+kind: Agent
+metadata:
+  name: code-reviewer
+  version: 1.0.0
+  description: "AI code reviewer that reviews code and suggests improvements"
+  author: Deepleaper
+  license: Apache-2.0
+  marketplace:
+    certified: false
+    category: developer-tools
+spec:
+  provider:
+    default: deepseek
+    allowed: [openai, deepseek, qwen]
+  model: deepseek-chat
+  systemPrompt: |
+    You are an expert code reviewer. When given code:
+    1. Check for bugs, security issues, and performance problems
+    2. Suggest improvements for readability and maintainability
+    3. Follow language-specific best practices
+    4. Be constructive and explain your reasoning
+    Rate severity: 🔴 Critical | 🟡 Warning | 🔵 Info
+  skills:
+    - name: code-analysis
+      description: "Analyze code for bugs and improvements"
+    - name: style-check
+      description: "Check coding style and best practices"
+  channels:
+    - type: web
+      port: 3000
+    - type: websocket
+      port: 3002
+  memory:
+    shortTerm: true
+    longTerm: false
+  dtv:
+    trust:
+      level: sandbox
+    value:
+      metrics: [reviews_completed, issues_found, suggestions_accepted]

package/templates/customer-service/README.md

@@ -1,22 +1,22 @@
-# Customer Service Agent Template
-
-A ready-to-use customer service agent with:
-- **FAQ Lookup** — Matches common questions to predefined answers
-- **Human Handoff** — Detects when a customer wants a real person
-- **Web Channel** — HTTP API for chat integration
-
-## Quick Start
-
-```bash
-opc init my-service --template customer-service
-cd my-service
-opc run
-```
-
-## Customization
-
-Edit `oad.yaml` to:
-- Change the system prompt
-- Add custom FAQ entries
-- Switch LLM provider
-- Adjust trust level
+# Customer Service Agent Template
+
+A ready-to-use customer service agent with:
+- **FAQ Lookup** — Matches common questions to predefined answers
+- **Human Handoff** — Detects when a customer wants a real person
+- **Web Channel** — HTTP API for chat integration
+
+## Quick Start
+
+```bash
+opc init my-service --template customer-service
+cd my-service
+opc run
+```
+
+## Customization
+
+Edit `oad.yaml` to:
+- Change the system prompt
+- Add custom FAQ entries
+- Switch LLM provider
+- Adjust trust level

package/templates/customer-service/oad.yaml

@@ -1,36 +1,36 @@
-apiVersion: opc/v1
-kind: Agent
-metadata:
-  name: customer-service
-  version: 1.0.0
-  description: "A customer service agent with FAQ lookup and human handoff"
-  author: Deepleaper
-  license: Apache-2.0
-  marketplace:
-    certified: false
-    category: customer-service
-spec:
-  provider:
-    default: deepseek
-    allowed: [openai, deepseek, qwen]
-  model: deepseek-chat
-  systemPrompt: |
-    You are a friendly and professional customer service agent.
-    You help customers with their questions about products, orders, shipping, and returns.
-    Be concise, helpful, and empathetic.
-  skills:
-    - name: faq-lookup
-      description: "Look up FAQ answers"
-    - name: human-handoff
-      description: "Hand off to human agent when needed"
-  channels:
-    - type: web
-      port: 3000
-  memory:
-    shortTerm: true
-    longTerm: false
-  dtv:
-    trust:
-      level: sandbox
-    value:
-      metrics: [response_time, satisfaction_score]
+apiVersion: opc/v1
+kind: Agent
+metadata:
+  name: customer-service
+  version: 1.0.0
+  description: "A customer service agent with FAQ lookup and human handoff"
+  author: Deepleaper
+  license: Apache-2.0
+  marketplace:
+    certified: false
+    category: customer-service
+spec:
+  provider:
+    default: deepseek
+    allowed: [openai, deepseek, qwen]
+  model: deepseek-chat
+  systemPrompt: |
+    You are a friendly and professional customer service agent.
+    You help customers with their questions about products, orders, shipping, and returns.
+    Be concise, helpful, and empathetic.
+  skills:
+    - name: faq-lookup
+      description: "Look up FAQ answers"
+    - name: human-handoff
+      description: "Hand off to human agent when needed"
+  channels:
+    - type: web
+      port: 3000
+  memory:
+    shortTerm: true
+    longTerm: false
+  dtv:
+    trust:
+      level: sandbox
+    value:
+      metrics: [response_time, satisfaction_score]

package/templates/docker-compose.yml

@@ -1,21 +1,21 @@
-version: '3.8'
-
-services:
-  agent:
-    build: .
-    ports:
-      - "3000:3000"
-    env_file:
-      - .env
-    volumes:
-      - ./oad.yaml:/app/oad.yaml:ro
-      - agent-data:/app/data
-    restart: unless-stopped
-    healthcheck:
-      test: ["CMD", "wget", "-q", "--spider", "http://localhost:3000/health"]
-      interval: 30s
-      timeout: 5s
-      retries: 3
-
-volumes:
-  agent-data:
+version: '3.8'
+
+services:
+  agent:
+    build: .
+    ports:
+      - "3000:3000"
+    env_file:
+      - .env
+    volumes:
+      - ./oad.yaml:/app/oad.yaml:ro
+      - agent-data:/app/data
+    restart: unless-stopped
+    healthcheck:
+      test: ["CMD", "wget", "-q", "--spider", "http://localhost:3000/health"]
+      interval: 30s
+      timeout: 5s
+      retries: 3
+
+volumes:
+  agent-data:

package/templates/knowledge-base/README.md

@@ -1,28 +1,28 @@
-# Knowledge Base Template
-
-RAG-powered knowledge base agent that answers questions from your company docs.
-
-## Features
-- **Semantic Search**: Uses DeepBrain for intelligent document retrieval
-- **Document Ingestion**: Add docs to build your knowledge base
-- **Source Citations**: Cites sources in answers
-
-## Quick Start
-```bash
-opc init my-kb --template knowledge-base
-cd my-kb
-# Optional: install deepbrain for full RAG support
-npm install deepbrain
-opc run
-```
-
-## How It Works
-1. Ingest company documents into DeepBrain
-2. User asks a question
-3. Agent searches for relevant document chunks
-4. LLM generates answer based on retrieved context
-
-## Configuration
-- `memory.longTerm.provider: deepbrain` — enables semantic memory
-- `memory.longTerm.collection` — DeepBrain collection name
-- Falls back to in-memory if deepbrain is not installed
+# Knowledge Base Template
+
+RAG-powered knowledge base agent that answers questions from your company docs.
+
+## Features
+- **Semantic Search**: Uses DeepBrain for intelligent document retrieval
+- **Document Ingestion**: Add docs to build your knowledge base
+- **Source Citations**: Cites sources in answers
+
+## Quick Start
+```bash
+opc init my-kb --template knowledge-base
+cd my-kb
+# Optional: install deepbrain for full RAG support
+npm install deepbrain
+opc run
+```
+
+## How It Works
+1. Ingest company documents into DeepBrain
+2. User asks a question
+3. Agent searches for relevant document chunks
+4. LLM generates answer based on retrieved context
+
+## Configuration
+- `memory.longTerm.provider: deepbrain` — enables semantic memory
+- `memory.longTerm.collection` — DeepBrain collection name
+- Falls back to in-memory if deepbrain is not installed