ThoughtMCP

# Building Knowledge **Time**: 10 minutes **Difficulty**: Beginner **Tools Used**: `remember`, `recall` ## The Scenario You want ThoughtMCP to learn from your experiences and remember important information for future use. This creates a personalized knowledge base that improves over time. ## Storing Memories ### Episodic Memory (Experiences) Store specific experiences with context and emotional significance. **Tool Call:** ```json { "tool": "remember", "arguments": { "content": "Had a great meeting with the client today. They loved our prototype and want to move forward with the full project. Sarah from their team will be our main contact.", "type": "episodic", "importance": 0.8, "emotional_tags": ["satisfaction", "success", "relief"] } } ``` **What Happens:** - Stores the experience with emotional context - Associates with current time and situation - Tags with importance level for future retrieval - Creates connections to related memories **When to Use Episodic Memory:** - Meeting outcomes - Project milestones - Learning experiences - Personal insights - Problem solutions ### Semantic Memory (Facts & Knowledge) Store general knowledge and facts that aren't tied to specific experiences. **Tool Call:** ```json { "tool": "remember", "arguments": { "content": "React hooks were introduced in version 16.8. useState and useEffect are the most commonly used hooks for state management and side effects.", "type": "semantic", "importance": 0.7, "context": { "domain": "web-development", "technology": "React" } } } ``` **What Happens:** - Stores factual information - Creates conceptual associations - Links to related knowledge - Available for future reasoning **When to Use Semantic Memory:** - Technical facts - Best practices - Definitions - Procedures - General knowledge ## Retrieving Memories ### Basic Recall Find relevant memories using search cues. **Tool Call:** ```json { "tool": "recall", "arguments": { "cue": "client meeting prototype" } } ``` **Expected Response:** ```json { "memories": [ { "content": "Had a great meeting with the client today...", "type": "episodic", "relevance_score": 0.92, "timestamp": "2024-01-15T14:30:00Z", "emotional_context": ["satisfaction", "success"] } ], "search_time_ms": 45, "total_memories_searched": 1247 } ``` ### Advanced Recall with Filters **Tool Call:** ```json { "tool": "recall", "arguments": { "cue": "React hooks useState", "type": "semantic", "max_results": 5, "threshold": 0.6 } } ``` **What's Different:** - Searches only semantic memories - Returns up to 5 results - Uses lower similarity threshold - More comprehensive search ### Context-Aware Recall **Tool Call:** ```json { "tool": "recall", "arguments": { "cue": "project management challenges", "context": { "domain": "software-development", "urgency": 0.8 } } } ``` **What's Different:** - Considers current context - Prioritizes urgent situations - Domain-specific search - More relevant results ## Building Knowledge Over Time ### Learning from Experiences **Scenario**: You've completed a challenging project and want to capture lessons learned. **Step 1: Store the Experience** ```json { "tool": "remember", "arguments": { "content": "Completed the e-commerce project 2 weeks late due to scope creep. Client kept adding features without adjusting timeline. Next time, need to establish change control process upfront.", "type": "episodic", "importance": 0.9, "emotional_tags": ["frustration", "learning", "determination"], "context": { "project_type": "e-commerce", "domain": "project-management" } } } ``` **Step 2: Store the Lesson** ```json { "tool": "remember", "arguments": { "content": "Always establish a formal change control process at project start. Document that scope changes require timeline and budget adjustments. Get client agreement in writing.", "type": "semantic", "importance": 0.8, "context": { "domain": "project-management", "category": "best-practices" } } } ``` ### Connecting Related Knowledge **Tool Call:** ```json { "tool": "remember", "arguments": { "content": "The Agile methodology emphasizes responding to change over following a plan, but this doesn't mean accepting unlimited scope creep. Change requests should still go through proper evaluation.", "type": "semantic", "importance": 0.7, "context": { "domain": "project-management", "methodology": "agile", "relates_to": "scope-management" } } } ``` ## Using Memory with Thinking ### Memory-Enhanced Reasoning Combine recall with thinking for better decisions. **Step 1: Recall Relevant Experience** ```json { "tool": "recall", "arguments": { "cue": "project timeline scope changes" } } ``` **Step 2: Think with Memory Context** ```json { "tool": "think", "arguments": { "input": "A new client wants to add social media integration to our web app project. How should I handle this request?", "mode": "deliberative", "context": { "previous_memories": "Retrieved memories about scope creep challenges" } } } ``` **What Happens:** - ThoughtMCP considers past experiences - Applies learned lessons to current situation - Provides more informed recommendations - Avoids repeating past mistakes ## Memory Management ### Importance Levels Use importance to prioritize what gets remembered: - **0.9-1.0**: Critical information, never forget - **0.7-0.9**: Important, remember long-term - **0.5-0.7**: Useful, remember medium-term - **0.3-0.5**: Interesting, remember short-term - **0.0-0.3**: Trivial, may be forgotten quickly ### Emotional Tags Help with memory retrieval and context: - **Positive**: "success", "satisfaction", "joy", "pride" - **Negative**: "frustration", "disappointment", "concern" - **Learning**: "insight", "discovery", "understanding" - **Social**: "collaboration", "conflict", "support" ### Context Tags Organize memories by domain and category: ```json { "context": { "domain": "web-development", "technology": "React", "project": "e-commerce-site", "phase": "development" } } ``` ## Try It Yourself ### Experiment 1: Personal Knowledge Base Build a knowledge base about your current project: 1. **Store project facts** (semantic memory) 2. **Store daily experiences** (episodic memory) 3. **Store lessons learned** (semantic memory) 4. **Recall information** when making decisions ### Experiment 2: Learning Loop Create a learning loop: 1. **Try something new** 2. **Store the experience** with emotional context 3. **Extract the lesson** as semantic knowledge 4. **Apply the lesson** in future situations ### Experiment 3: Memory-Enhanced Decisions For your next decision: 1. **Recall relevant experiences** first 2. **Think about the decision** with memory context 3. **Store the outcome** for future reference ## Common Patterns ### Project Documentation ```json // Store project milestone { "tool": "remember", "arguments": { "content": "Completed user authentication module. Used JWT tokens with refresh mechanism. Took 3 days longer than estimated due to security requirements.", "type": "episodic", "importance": 0.7, "context": { "project": "current-project", "module": "authentication", "phase": "development" } } } // Store technical knowledge { "tool": "remember", "arguments": { "content": "JWT refresh tokens should be stored in httpOnly cookies for security. Access tokens can be in memory. Refresh token rotation prevents replay attacks.", "type": "semantic", "importance": 0.8, "context": { "domain": "security", "technology": "JWT" } } } ``` ### Learning from Mistakes ```json // Store the mistake { "tool": "remember", "arguments": { "content": "Deployed to production without running full test suite. Broke user registration. Had to rollback and fix. Embarrassing client call.", "type": "episodic", "importance": 0.9, "emotional_tags": ["embarrassment", "learning", "regret"] } } // Store the lesson { "tool": "remember", "arguments": { "content": "Never deploy to production without running the complete test suite. Create deployment checklist to prevent this mistake.", "type": "semantic", "importance": 0.9, "context": { "domain": "deployment", "category": "best-practices" } } } ``` ## Key Takeaways ### Memory Types - **Episodic**: Specific experiences with emotional context - **Semantic**: General knowledge and facts - **Both**: Work together to create rich understanding ### Best Practices 1. **Be specific** in memory content 2. **Use appropriate importance** levels 3. **Add emotional context** for better recall 4. **Include domain tags** for organization 5. **Connect related memories** through context ### Memory Benefits - **Personalized responses** based on your experience - **Continuous learning** from successes and failures - **Context-aware decisions** using past knowledge - **Improved reasoning** with relevant background ## Next Steps - **[Quality Checking](quality-checking.md)** - Improve reasoning quality - **[Complete Workflow](complete-workflow.md)** - Use all tools together - **[Real-World Examples](../real-world/)** - See complex applications --- _Ready to improve reasoning quality? Learn about [Quality Checking](quality-checking.md)._