Guidelines Analysis Workflow
The guidelines analysis workflow is the core feature of Better Context, enabling AI assistants to analyze repository code and generate contextual development guidelines.
It combines repository analysis, AI processing, and structured guidelines generation to create actionable development documentation tailored to specific contexts like testing, architecture, or documentation standards.
Run the Analysis
Make sure you have at least one context set up for the current project and Better Context is running. Prompt the following:
Using better context MCP, run the tool guidelines_analysis to initiate the coding guidelines analysis workflow for this project.
Workflow Architecture
Workflow Steps
1. Analysis Initiation
The workflow begins when an AI assistant calls the guidelines_analysis
MCP tool without a contextId:
- Repository Detection: Identifies the current repository using Git origin URL or working directory path
- Context Loading: Retrieves all defined guidelines contexts for the repository
- Queue Creation: Creates a processing queue with all contexts that need analysis
- Initial Prompt: Returns a structured prompt instructing the AI to begin with the first context
2. Context Processing Loop
The workflow processes each context sequentially through a loop mechanism:
Per-Context Steps:
Context Analysis Setup
- AI calls
guidelines_analysis
with the current context ID - Tool returns context-specific analysis instructions and file patterns
- AI calls
Repository Analysis
- AI scans relevant files based on context configuration
- Analyzes code patterns, structures, and conventions
- Identifies context-specific guidelines opportunities
Guidelines Generation
- AI generates actionable guidelines as an array of strings
- Guidelines are tailored to the specific context (testing, architecture, etc.)
Guidelines Storage
- AI calls
guidelines_save
with generated guidelines and context ID - System checks if guidelines already exist for this context
- New Guidelines: Stored directly in the database
- Existing Guidelines: Creates a pending version for review
- AI calls
Loop Continuation:
- Context Queue Check: System determines if more contexts remain unprocessed
- Continue Prompt: If contexts remain, generates a continuation prompt for the next context
- Loop Iteration: Process repeats from step 1 with the next context
- Completion: When all contexts are processed, the analysis workflow completes
3. Real-time Updates
Throughout the loop process:
- HTTP Streaming Events: HTTP streaming provides real-time progress updates.
- Web Interface: Updates reflect current analysis status and completed contexts.
- Diff Synchronization: When a diff is created (pending version), the interface immediately displays the comparison and pending state.
4. Analysis Completion
Once all contexts have been processed:
- Final Update: Web interface shows all generated guidelines
- Database Consistency: All guidelines are properly stored and associated
- Ready for Use: Guidelines are immediately available for AI assistants and developers
Key Features
Sequential Processing
The loop-based architecture ensures:
- Focused Analysis: Each context receives dedicated attention
- Resource Management: Processing happens one context at a time
- Error Isolation: Issues with one context don't affect others
Prompt-Driven Workflow
The system uses intelligent prompting with templating:
- Dynamic Instructions: Each prompt is tailored to the current context and remaining queue using template variables
- Template Engine: Supports variable substitution (
{{variable}}
) and array iteration ({{#each array}}
) for dynamic content generation - Self-Managing: AI assistants follow structured instructions to complete the entire workflow
- Flexible Continuation: Supports interruption and resumption of analysis
Context-Aware Guidelines
Each context produces specialized guidelines.
Guidelines Versioning and Diff Management
When generating new guidelines for a context that already has existing guidelines, Better Context creates a pending version. Your current guidelines are never overwritten until you validate or cancel the pending changes. This ensures you always control what is kept or discarded.
Automatic Version Detection
The system automatically detects when guidelines already exist for a context:
- First-time Generation: Guidelines are stored directly without versioning.
- Subsequent Generations: A pending version is created and shown alongside your current guidelines. You must validate or cancel before making further changes to that context.
- Protection: Prevents immediate overwrite of existing work.
Diff Visualization
When pending versions are created, the web interface always shows a clear visual diff:
- Side-by-side Comparison: Current guidelines vs. pending changes.
- Line-by-line Analysis: Visual indicators for added, removed, and unchanged content.
- Context Information: Contexts with pending changes are highlighted in the interface.
User Actions
For each pending version, you must choose to validate or cancel before continuing:
Validate Changes
- Action: Replaces current guidelines with the pending version.
- Effect: Pending version becomes the new current guidelines.
- Use Case: When the new analysis provides better or updated guidelines.
Cancel Changes
- Action: Discards the pending version.
- Effect: Current guidelines remain unchanged.
- Use Case: When current guidelines are preferred or new analysis is incorrect.
Workflow Considerations
Multiple Contexts
- Each context manages its own versions independently.
- Partial validation is supported (validate some contexts, cancel others).
- Analysis can continue with remaining contexts after resolving diffs.
Error Prevention
- You cannot generate new guidelines for a context until you have validated or canceled any pending changes.
- Clear error messages guide you to resolve pending changes first.
- Ensures data consistency and prevents confusion.
Web Interface Integration
- Contexts with pending changes are visually highlighted.
- All changes and pending updates are instantly synchronized for all users working on the same repository.
- Streamlined workflow for reviewing and managing multiple diffs.
Best Practices
- Review Before Validating: Always examine the diff to understand what's changing
- Context-Specific Decisions: Each context may warrant different actions based on analysis quality
- Incremental Updates: Consider validating high-quality changes and canceling questionable ones
- Regular Maintenance: Address pending diffs promptly to keep the workflow smooth
Need Help?
- Check the troubleshooting guide
- Report issues on GitHub