The accordo cache system with semantic search enables agents to enhance analysis phases by leveraging historical context from similar past work. This guide shows practical examples of how agents can use semantic search to create richer, more informed analysis.
- Every workflow execution is automatically stored with embeddings
- Task descriptions, implementation details, and outcomes become searchable
- Agents build an organizational knowledge base over time
- Agents query past work using natural language descriptions
- Vector embeddings find semantically similar tasks and contexts
- Results include similarity scores, metadata, and historical patterns
- Historical context informs current analysis
- Proven patterns reduce risk and accelerate development
- Past failures provide risk mitigation insights
Scenario: Agent needs to optimize slow database queries
Traditional Analysis:
Current task: Optimize slow database queries
- Analyze current query performance
- Identify bottlenecks
- Plan optimization approach
Enhanced Analysis with Semantic Search:
1. Baseline Analysis:
Current task: Optimize slow database queries affecting user experience
2. Historical Context Discovery:
workflow_semantic_search(
query="database performance optimization slow queries",
analysis_type="similar_tasks",
max_results=5
)
Results:
- Project Alpha: 87% similarity - "PostgreSQL query optimization for user dashboard"
- Project Beta: 76% similarity - "MySQL performance tuning slow reporting queries"
- Project Gamma: 65% similarity - "Database indexing strategy for large datasets"
3. Lessons Learned Analysis:
workflow_semantic_search(
query="database optimization performance",
analysis_type="lessons_learned",
max_results=3
)
Results:
- Completed: "Adding composite indexes improved query time by 85%"
- Completed: "Connection pooling reduced timeout issues significantly"
- Completed: "Query caching at application level provided 60% performance boost"
4. Enhanced Analysis Output:
Based on historical analysis:
- Follow proven indexing strategy from Project Alpha (87% similar)
- Implement connection pooling based on successful pattern from Project Beta
- Consider query caching approach that showed 60% improvement
- Avoid ORM optimization approach that failed in 2 similar projects
- Plan for 2-week timeline based on historical data from similar complexity
Scenario: Agent building a new user dashboard component
Enhanced Analysis Process:
1. Current Task Analysis:
"Build responsive user dashboard with charts and real-time data updates"
2. Similar Tasks Discovery:
workflow_semantic_search(
query="React dashboard component responsive charts real-time data",
analysis_type="similar_tasks"
)
Historical Insights:
- 89% similar: "React analytics dashboard with Chart.js integration"
- 76% similar: "Real-time data visualization dashboard using React hooks"
- 71% similar: "Responsive admin panel with multiple chart types"
3. Technology Context Analysis:
workflow_semantic_search(
query="React hooks useEffect useState real-time data WebSocket",
analysis_type="related_context"
)
Related Patterns:
- WebSocket connection management in React components
- State management patterns for real-time updates
- Performance optimization for frequent data updates
- Mobile responsiveness patterns for dashboard layouts
4. Risk Assessment from Historical Data:
- Memory leaks in useEffect cleanup (found in 3 similar projects)
- WebSocket reconnection handling (failed in 2 related implementations)
- Chart performance with large datasets (optimization needed in 4 similar cases)
5. Evidence-Based Implementation Plan:
- Use proven WebSocket management pattern from 89% similar project
- Implement Chart.js integration following successful 76% similar approach
- Add memory leak prevention based on lessons from 3 related failures
- Plan for mobile-first responsive design using patterns from admin panel project
Scenario: Implementing authentication for a new API
Semantic-Enhanced Analysis:
1. Security Context Discovery:
workflow_semantic_search(
query="API authentication JWT OAuth2 security implementation",
analysis_type="similar_tasks"
)
Security Patterns Found:
- 91% similar: "JWT authentication with refresh token implementation"
- 84% similar: "OAuth2 integration with Google and GitHub providers"
- 78% similar: "API rate limiting and security middleware setup"
2. Security Lessons Analysis:
workflow_semantic_search(
query="JWT token security vulnerability",
analysis_type="lessons_learned"
)
Critical Security Lessons:
- JWT secret rotation strategy prevented security breach
- Token expiry handling crucial for user experience
- CORS configuration errors caused production issues in 2 projects
- Rate limiting prevented DOS attacks in 3 similar APIs
3. Cross-Domain Security Insights:
workflow_semantic_search(
query="authentication security best practices",
analysis_type="related_context"
)
Broader Security Context:
- Password hashing strategies from user management systems
- Session management patterns from web applications
- Multi-factor authentication approaches from enterprise systems
- Audit logging patterns from financial applications
4. Risk-Informed Implementation:
- Follow JWT implementation pattern with 91% similarity success rate
- Implement refresh token rotation based on security lessons
- Add rate limiting using proven middleware from 78% similar project
- Plan comprehensive CORS testing based on 2 historical failures
- Include audit logging using patterns from financial domain
workflow_semantic_search(
query="description of current task or context",
analysis_type="similar_tasks|related_context|lessons_learned",
max_results=5,
min_similarity=0.3
)Parameters:
query: Natural language description of current task/problemanalysis_type: Type of analysis neededsimilar_tasks: Find similar past work and approachesrelated_context: Find related domains and technologieslessons_learned: Find completed similar workflows with outcomes
max_results: Number of results to return (1-20)min_similarity: Similarity threshold (0.0-1.0)
# Analysis phase with semantic enhancement
analysis_phase:
goal: "Analyze requirements with historical context"
instructions: |
1. Perform baseline analysis of current requirements
2. Use workflow_semantic_search() to find relevant past work
3. Integrate historical insights into analysis
4. Document synthesis of current + historical understanding
acceptance_criteria:
baseline_analysis: "Current requirements fully understood"
historical_context: "Relevant past work discovered and analyzed"
integrated_insights: "Historical lessons integrated into current analysis"- Leverage proven patterns instead of starting from scratch
- Reduce analysis time by building on past work
- Focus effort on novel aspects not covered historically
- Learn from historical failures and cancelled projects
- Avoid known pitfalls and problematic approaches
- Plan mitigation strategies based on past experience
- Reference successful patterns and best practices
- Validate approaches against historical success rates
- Ensure consistency with organizational standards
- Maintain institutional knowledge across projects
- Enable cross-team learning and collaboration
- Build organizational expertise over time
- Identify areas where proven patterns apply
- Focus innovation on truly novel requirements
- Build confidently on validated foundations
- Use specific, descriptive language in queries
- Include key technologies, domains, and objectives
- Try multiple query variations for comprehensive coverage
- Always perform baseline analysis first
- Use semantic search to enhance, not replace, critical thinking
- Document both historical insights and novel considerations
- Don't copy patterns blindly - adapt to current context
- Understand why historical approaches succeeded or failed
- Combine insights from multiple similar projects for robustness
- Document current work with clear, searchable descriptions
- Record decision rationale for future agents
- Contribute lessons learned back to the knowledge base
To enable semantic analysis capabilities:
# Start MCP server with cache mode
python -m accordo-mcp.server \
--enable-cache-mode \
--cache-embedding-model all-MiniLM-L6-v2 \
--cache-db-path .accordo/cache \
--cache-max-results 50- Start with broad queries, then narrow based on results
- Cross-reference multiple search types for comprehensive view
- Refine understanding as domain knowledge increases
- Search related but different domains for innovative approaches
- Adapt successful patterns from adjacent technologies
- Find creative solutions by exploring broader contexts
- Compare current results with similar past projects
- Use historical metrics for validation and quality assessment
- Reference performance data from comparable implementations
- Track pattern evolution across multiple projects
- Identify emerging best practices and standards
- Build domain expertise through accumulated experience
Semantic search transforms analysis phases from isolated, individual efforts into knowledge-driven processes that leverage organizational learning. Agents can deliver higher quality analysis faster by building on proven patterns while contributing to collective knowledge for future projects.
The cache system automatically captures and organizes workflow knowledge, enabling continuous improvement and organizational learning at scale.