RelaMind

<div align="center"> # Tool List  > Record growth, understand yourself, and become a better you </div> RelaMind is an AI-based personal growth trajectory analysis system that helps users better understand themselves and achieve continuous growth by recording life, reviewing history, and analyzing patterns. [](https://star-history.com/#El-12stu/RelaMind&type=Date) ## 📸 Effect Display ### Homepage  *The homepage displays three core functions: intelligent dialogue, growth diary, and task assistant, using a modern gradient design and animation effects* ### Intelligent Dialogue - Intelligent Routing Function The intelligent dialogue system supports automatic identification of user intentions and intelligent routing to the most appropriate processing method: #### Ordinary Chat Mode  *Daily companion chat, emotional support, and guided recording* #### RAG Historical Query Mode  *Automatically retrieve historical records and answer questions like "What did I write in my diary at this time last year?"* #### Tool Call Mode  *Automatically call tools such as search, download, and file generation to complete tasks* ### Intelligent Routing Recognition Process The system will automatically identify user intentions and route to the corresponding mode: #### Ordinary Chat Routing  *The system identifies as ordinary chat intention and uses daily companion mode* #### RAG Query Routing  *The system identifies as historical query intention and automatically uses RAG retrieval mode* #### Tool Call Routing  *The system identifies as tool call intention and automatically calls the corresponding tool* ### Growth Diary  *Growth diary recording interface, supporting mood selection, tag management, content recording, and automatic storage to vector database* ### Manus Task Assistant 🤖 Manus is an advanced intelligent agent of RelaMind, based on ReAct (Reasoning and Acting) mode, which can independently plan complex tasks and execute multi-step operations. #### Task Execution Process Display **Step 1: Task Analysis and Planning**  *Manus analyzes user needs, formulates an execution plan, and displays the thinking process and tools called* **Step 2: Information Collection and Processing**  *Manus automatically calls tools such as search and web crawling to collect information and display execution progress in real-time* **Step 3: Content Generation and File Creation**  *Manus organizes information and generates files, supporting txt, PDF and other formats, and provides download links* **Step 4: Task Completion and Result Display**  *After the task is completed, Manus displays the final result and generated file, and the user can directly download* **Core Features**: - **Streaming Output**: Immediately display after each execution step, real-time feedback task progress - **Visualized Thinking Process**: Clearly display AI's thinking content and decision-making process - **Transparent Tool Calling**: Display the name of the tool called, hiding technical details - **Automatic File Generation**: Support generating txt, PDF and other format files, and provide download links - **Multi-Step Collaboration**: Automatically coordinate multiple tools to complete complex tasks ### Sensitive Word Detection 🛡️ The system has built-in sensitive word detection function to ensure content safety and compliance.  *When the user input contains sensitive words, the system will immediately intercept and prompt, without calling large model API, to protect system safety* **Technical Features**: - **Trie Tree Algorithm**: Based on prefix tree to achieve efficient sensitive word matching - **Zero API Calls**: Directly intercept when detecting sensitive words, without consuming large model API resources - **Real-Time Detection**: Detect in real-time before sending the request to the large model - **Configurable Word Library**: Support loading sensitive word library from configuration file ## ✨ Product Positioning RelaMind is positioned as "your AI growth partner" to help users: - **Daily Companion**: Listen to user voice, give emotional support and understanding - **Guided Recording**: Encourage users to record life bits and pieces, develop recording habits - **Growth Analysis**: Analyze growth patterns based on historical records, provide personalized insights - **Active Care**: Remember user's important information, actively care in time ## 🎯 Core Functions ### 1. Intelligent Dialogue 💬 **Function Features**: - **Intelligent Routing**: Automatically identify user intentions (ordinary chat / historical query / tool call), choose the most appropriate processing method - **Daily Companion**: Multi-round dialogue, emotional support, guided recording - **Historical Retrieval**: Based on RAG technology to retrieve historical records, answer questions like "What was I doing at this time last year?" - **Tool Calls**: Automatically call tools such as search, download, and file generation - **Dialogue Memory**: Store dialogue history based on Cassandra, realize long-term memory **Usage Scenarios**: - Daily chat: "Today's mood is good" - Historical query: "What was I doing at this time last year?" - Tool calls: "Help me search for today's weather" ### 2. Growth Diary 📝 **Function Features**: - **Mood Recording**: 8 mood options (happy, excited, calm, tired, sad, anxious, angry, confused) - **Tag Management**: Custom tags + common tags, easy to classify and retrieve - **RAG Storage**: Automatically store to vector database, support semantic retrieval - **Metadata Enrichment**: Automatically record date, time, mood, tags and other information **Usage Scenarios**: - Record daily thoughts and feelings - Build personal growth knowledge base - Provide data source for subsequent RAG retrieval ### 3. Manus Task Assistant 🤖 **Function Features**: - **Autonomous Planning**: Based on ReAct (Reasoning and Acting) mode, autonomous reasoning and action - **Multi-Tool Collaboration**: Automatically call multiple tools to complete complex tasks (search, crawling, file generation, etc.) - **Multi-Step Execution**: Support complex multi-step task planning, up to 10 steps - **Streaming Output**: Immediately display after each execution step, real-time feedback task progress - **Visualized Thinking**: Clearly display AI's thinking process and decision-making logic - **File Generation**: Automatically generate txt, PDF and other format files, and provide download links - **Transparent Tool Calling**: Display the name of the tool called, hiding technical details **Usage Scenarios**: - Develop a complete plan (e.g., "Help me develop a weekend plan") - Information collection and organization (e.g., "Help me search for fitness suggestions and return in PDF format") - Tasks that require multi-step operations (e.g., "Help me search for today's news and organize into a document") - Complex information processing (e.g., "Help me analyze a topic and generate a report") **Execution Process**: 1. **Thinking Stage**: Analyze user needs, formulate execution plan 2. **Action Stage**: Call corresponding tools (search, crawling, file operation, etc.) 3. **Iterative Execution**: Continue thinking and executing next step based on results 4. **Task Completion**: Generate final result and file, provide download link ### 4. Intelligent Intention Identification 🧠 **Technical Features**: - **LLM-Based Classification**: Use structured output for intention classification, improve accuracy - **Rule Matching**: Fast rule matching (default), can switch to LLM classification - **Confidence Mechanism**: When recognition confidence is low, actively ask user to confirm - **Continuous Optimization**: Record recognition accuracy, continuously optimize model ### 5. Sensitive Word Detection 🛡️ **Function Features**: - **Trie Tree Algorithm**: Based on prefix tree to achieve efficient sensitive word matching, time complexity O(n) - **Zero API Calls**: Directly intercept when detecting sensitive words, without consuming large model API resources - **Real-Time Detection**: Detect in real-time before sending the request to the large model - **Configurable Word Library**: Support loading sensitive word library from `sensitive_words.txt` configuration file - **Advisor Interception**: Through Spring AI Advisor mechanism, intercept in the early stage of call chain **Technical Implementation**: - Use `SensitiveWordAdvisor` as the highest priority Advisor (order = -100) - Based on Trie tree multi-pattern matching algorithm - Support Chinese and English sensitive word detection - Automatically load and initialize sensitive word library **Usage Scenarios**: - Prevent users from inputting improper content - Protect system safety, avoid malicious requests - Save API call cost ### 6. Tool Call Capability 🔧 - **Internet Search**: Real-time search for the latest information - **File Operation**: Save, read user files - **Web Crawling**: Extract web content - **Resource Download**: Download pictures, documents and other resources - **PDF Generation**: Generate personalized reports - **Terminal Operation**: Execute code scripts ### 7. MCP Service Integration 🌐 - **Map Service**: Find nearby locations, plan routes - **Picture Search**: Search for pictures from specific websites - **Extended Capability**: Support custom MCP service ## 🛠️ Technical Stack ### Backend Technology - **Java 21** + **Spring Boot 3** - **Spring AI**: AI application development framework - **LangChain4j**: AI application development tool library - **Cassandra**: Dialogue memory persistence storage - **PgVector**: Vector database for RAG retrieval - **Spring AI ChatClient**: Multi-round dialogue, Advisor, ChatMemory - **Spring AI RAG**: Knowledge base retrieval enhancement generation - **Spring AI Tool Calling**: Tool call capability - **Spring AI MCP**: Model context protocol support ### AI Capability - **Large Model Access**: Support multiple AI large models (Tongyi Qianwen, GPT, etc.) - **Prompt Engineering**: Optimized prompt word design - **Structured Output**: Use Entity for structured data extraction - **Streaming Response**: SSE streaming transmission, improve user experience - **Multi-Modal Support**: Support text, pictures and other multi-modal input ### Tools and Libraries - **Kryo**: High-performance serialization - **Jsoup**: Web crawling - **iText**: PDF generation - **Knife4j**: API document generation ### Frontend Technology - **Vue 3**: Frontend framework - **Vite**: Build tool - **SSE**: Server push event, realize streaming response ### Deployment - **Docker**: Containerized deployment - **Serverless**: Support Serverless deployment ## 📁 Project Structure ``` RelaMind/ ├── src/main/java/io/el12stu/RelaMind/ │ ├── app/ # Core application logic │ │ └── RelaMindApp.java # Intelligent routing, RAG, tool calls │ ├── agent/ # AI intelligent agent │ │ ├── Manus.java # Autonomous planning intelligent agent │ │ └── ReActAgent.java # ReAct mode intelligent agent │ ├── advisor/ # Spring AI Advisor │ │ ├── SensitiveWordAdvisor.java │ │ ├── IntentDetectionService.java │ │ └── ... │ ├── chatmemory/ # Dialogue memory │ │ └── CassandraBasedChatMemory.java │ ├── rag/ # RAG related │ │ ├── RelaMindAppDocumentLoader.java │ │ ├── QueryRewriter.java │ │ └── ... │ ├── service/ # Business service │ │ ├── IntentDetectionService.java # Intention identification │ │ ├── DiaryService.java # Diary storage service │ │ └── SensitiveWordService.java │ ├── tools/ # Tool definition │ │ ├── WebSearchTool.java │ │ ├── FileOperationTool.java │ │ ├── PDFGenerationTool.java │ │ └── ... │ └── controller/ # API interface │ ├── AiController.java # Intelligent dialogue interface │ └── DiaryController.java # Diary record interface ├── src/main/resources/ │ ├── application.yml # Configuration file │ └── document/ # Knowledge base document └── ai-agent-frontend/ # Frontend project └── src/ ├── views/ │ ├── Home.vue # Homepage (three functional cards) │ ├── RelaMind.vue # Intelligent dialogue page │ ├── Diary.vue # Growth diary page │ └── SuperAgent.vue # Task assistant page ├── components/ │ └── ChatRoom.vue # Chat room component └── api/ └── index.js # API call encapsulation ``` ## 🚀 Quick Start ### Environment Requirements - JDK 21+ - Maven 3.6+ - PostgreSQL (support PgVector) - Cassandra (optional, for dialogue memory) ### Configuration Description 1. Configure AI large model API Key (in `application.yml`) 2. Configure database connection (PostgreSQL, Cassandra) 3. Configure vector database (PgVector) ### Run Project #### Method 1: Local Development and Operation ```bash # Backend mvn spring-boot:run # Frontend cd ai-agent-frontend npm install npm run dev ``` #### Method 2: Docker Deployment (Recommended) 🐳 **Difficulty Assessment**: ⭐⭐☆☆☆ (simple to medium) Dockerization has been completed, supporting two deployment methods: use pre-built images (recommended) or local construction. **Prerequisites**: - Install Docker and Docker Compose - Configure AI large model API Key --- ### 🚀 Method A: Use Pre-Built Images (Recommended, Simplest) **Advantages**: No need to build environment, directly pull and use, quick start ```bash # 1. Clone repository git clone https://github.com/your-username/RelaMind.git cd RelaMind # 2. Create configuration file cp env.example .env # 3. Edit .env file, configure the following: # - DASHSCOPE_API_KEY (required): Your Alibaba Cloud DashScope API Key # - GITHUB_USER (required): Your GitHub username # - GITHUB_REPO (required): Repository name (usually RelaMind) # - POSTGRES_PASSWORD (optional, recommended modification): Database password # - CASSANDRA_PASSWORD (optional, recommended modification): Cassandra password # 4. Modify docker-compose.yml image address # Replace your-username with your GitHub username # For example: ghcr.io/your-username/RelaMind/backend:latest # 5. Start all services docker-compose -f docker-compose.yml up -d # 6. View service status docker-compose -f docker-compose.yml ps # 7. View logs docker-compose -f docker-compose.yml logs -f backend ``` **Access Application**: - Frontend: http://localhost - Backend API: http://localhost:8123/api - Swagger documentation: http://localhost:8123/api/swagger-ui.html --- ### 🔨 Method B: Local Build Image **Advantages**: Can customize build parameters, suitable for development and debugging ```bash # 1. Create .env file (optional, used to configure sensitive information) cp env.example .env # Edit .env file, configure at least DASHSCOPE_API_KEY # 2. Use development configuration to start (will automatically build image) docker-compose up -d --build # 3. View service status docker-compose ps # 4. View logs docker-compose logs -f backend # 4. Restart service docker-compose restart backend # 1. Modify .env file # 2. Restart service docker-compose restart backend ``` --- ### 📋 Configuration Description All configurations are injected through environment variables, **images do not contain any sensitive information**. **Required Configuration**: - `DASHSCOPE_API_KEY`: Alibaba Cloud DashScope API Key **Optional Configuration** (default value): - `POSTGRES_USER`: default `relamind` - `POSTGRES_PASSWORD`: default `relamind123` (️ Production environment recommended modification) - `CASSANDRA_USERNAME`: default `cassandra` - `CASSANDRA_PASSWORD`: default `cassandra` (️ Production environment recommended modification) For detailed configuration description, please see [CONFIGURATION.md](CONFIGURATION.md) --- ### 🛠️ Common Commands ```bash # Start service docker-compose -f docker-compose.yml up -d # Stop service docker-compose -f docker-compose.yml down # Stop and delete data volume (⚠️ Will delete all data) docker-compose -f docker-compose.yml down -v # View logs docker-compose -f docker-compose.yml logs -f backend # Restart service docker-compose -f docker-compose.yml restart backend # Modify configuration and restart # 1. Modify .env file # 2. Restart service docker-compose -f docker-compose.yml restart backend ``` ### 📦 Docker File Description - `Dockerfile`: Backend application image (multi-stage construction, optimize image size) - `docker-compose.yml`: Development environment configuration (local build image) - `docker-compose.prod.yml`: Production environment configuration (use pre-built images) - `.dockerignore`: Optimize construction speed, exclude unnecessary files --- ### ⚠️ Notes 1. **First-time use of pre-built image**: - Need to modify the image address in `docker-compose.prod.yml` - Replace `your-username` with your GitHub username - Ensure the image has been published to GitHub Container Registry 2. **Port occupancy**: - Ensure ports 80, 8123, 5432, and 9042 are not occupied 3. **Data persistence**: - Data is stored in Docker volumes - Deleting containers will not lose data - Using `docker-compose down -v` will delete all data 4. **Configuration modification**: - After modifying the `.env` file, restart the service to take effect - No need to rebuild the image --- ### 📚 More Documentation - [CONFIGURATION.md](CONFIGURATION.md) - Detailed configuration instructions - [DEPLOYMENT.md](DEPLOYMENT.md) - Complete deployment guide ## 📝 Usage Examples ### Intelligent Conversation Example **Ordinary Chat**: ``` User: I'm feeling good today AI: Great! Can you share what's making you feel so good? Do you want to record it so you can look back on these happy moments later? ``` **History Record Query (Automatic use of RAG)**: ``` User: What was I doing this time last year? AI: Based on your records from March 2023, you were preparing for an important project and also recorded some thoughts about work pressure... ``` **Tool Call (Automatic tool invocation)**: ``` User: Help me search for today's weather AI: [Automatically invoke search tool] According to the search results, today's weather is sunny, with a temperature of... ``` ### Growth Diary Example Users can record through the interface: - Select mood: 😊 Happy - Add tags: work, thinking - Input content: I completed an important project today and feel a great sense of accomplishment... After saving, these records will be automatically stored in the vector database, and can be retrieved through intelligent conversation later. ### Manus Task Assistant Example **Example 1: Generate a learning suggestion document** ``` User: I want to learn about fitness, give me some suggestions in txt format. Manus Execution Process: ━━━ Step 1 ━━━ 💭 Thinking: The user wants to learn about fitness and get some suggestions, finally returned in txt format. Need to collect effective suggestions for learning fitness, then generate a txt file. 🔧 Tools called: searchWeb ━━━ Step 2 ━━━ 💭 Thinking: Search results have been obtained, need to grab relevant webpage content for detailed information. 🔧 Tools called: scrapeWebPage ━━━ Step 3 ━━━ 💭 Thinking: Enough information has been collected, now need to organize and generate a txt file. 🔧 Tools called: writeFile ━━━ Step 4 ━━━ 💭 Thinking: The file has been successfully generated, task completed. ✅ Task completed, tools called: doTerminate 📁 Generated file: 1. Fitness Suggestions.txt [Click to download] ``` **Example 2: Generate a PDF report** ``` User: Help me search for today's news and organize it into a PDF format. Manus Execution Process: 1. Search for news → 2. Grab webpage content → 3. Organize information → 4. Generate PDF → 5. Task completed 📁 Generated file: Today's News.pdf [Click to download] ``` ### Sensitive Word Detection Example ``` User input: Content with sensitive words System response: [System security prompt]: Your input contains sensitive content, and the request has been intercepted. Features: - Zero API calls: Direct interception, no consumption of large model resources - Real-time response: Millisecond-level detection speed - Safe and reliable: Protect the system from improper content ``` ## 🎨 Product Features 1. **Intelligent Routing**: Automatically identify user intent, no need for user selection, and the system automatically chooses the most appropriate processing method 2. **Long-term Memory**: Based on vector databases to achieve long-term memory, truly understand the user's growth trajectory 3. **Personalization**: Provide personalized suggestions based on user historical data, rather than general talk 4. **Integrated Experience**: Intelligent conversation, growth diary, and task assistant functions are seamlessly connected 5. **Continuous Optimization**: Record accuracy and continuously optimize intent recognition models ## 🔄 Functional Process ### Complete Usage Process 1. **Recording Stage**: Users record thoughts and moods in the "Growth Diary" 2. **Storage Stage**: Diary automatically stored in vector database (RAG) 3. **Conversation Stage**: Users communicate with AI in "Intelligent Conversation" 4. **Retrieval Stage**: When users ask about history, the system automatically retrieves relevant records through RAG 5. **Analysis Stage**: AI provides personalized analysis and suggestions based on historical data ### Intelligent Routing Process ``` User input ↓ Intent recognition (rules/LLM) ↓ ┌─────────┬──────────┬──────────┐ │ Ordinary Chat │ RAG Query │ Tool Calls │ └─────────┴──────────┴──────────┘ ↓ Intelligent reply ``` ## 📄 License This project is licensed under the [MIT License](LICENSE). **RelaMind** - Record growth, understand yourself, and become a better you