Opik MCP Server

/** * Tracing information module for Opik Comet * Provides detailed information about Opik's tracing capabilities */ interface TracingTopicInfo { title: string; description: string; keyFeatures: string[]; useCases: string[]; } interface TracingOverviewInfo { title: string; description: string; availableTopics: string[]; } type TracingInfo = TracingTopicInfo | TracingOverviewInfo; /** * Get information about Opik's tracing capabilities * @param topic Optional topic to get specific information about * @returns Information about the specified topic or an overview if no topic is provided */ export function getTracingInfo(topic?: string): TracingInfo | null { // If no topic is provided, return an overview if (!topic) { return { title: 'Opik Tracing Capabilities', description: 'Opik provides comprehensive tracing capabilities to help you understand and analyze your LLM applications. Traces capture the full context of LLM interactions, including inputs, outputs, and metadata.', availableTopics: ['traces', 'spans', 'feedback', 'search', 'visualization'], }; } // Return information based on the requested topic switch (topic.toLowerCase()) { case 'traces': return { title: 'Traces', description: 'Traces are complete records of LLM interactions, capturing the full context of a request including inputs, outputs, and metadata. They provide a comprehensive view of how your LLM application is performing.', keyFeatures: [ 'Automatic capture of inputs and outputs', 'Metadata collection for context', 'Hierarchical structure with spans', 'Long-term storage and retrieval', 'Integration with evaluation metrics', ], useCases: [ 'Debugging LLM application issues', 'Auditing LLM responses', 'Compliance and governance', 'Performance analysis', 'Training data collection', ], }; case 'spans': return { title: 'Spans', description: 'Spans are individual units within a trace that represent discrete operations or steps in your LLM application. They help break down complex interactions into manageable pieces for analysis.', keyFeatures: [ 'Hierarchical relationship (parent-child)', 'Timing information for performance analysis', 'Custom attributes for context', 'Support for nested operations', 'Automatic correlation with parent traces', ], useCases: [ 'Performance bottleneck identification', 'Detailed step-by-step analysis', 'Tracking complex multi-step LLM workflows', 'Measuring time spent in different components', 'Correlating errors with specific operations', ], }; case 'feedback': return { title: 'Feedback', description: 'Feedback allows you to annotate traces with human or automated evaluations. This helps you build a dataset of evaluated responses that can be used for model improvement and analysis.', keyFeatures: [ 'Binary (thumbs up/down) feedback', 'Structured feedback with categories', 'Free-form text comments', 'Support for multiple feedback sources', 'Integration with evaluation metrics', ], useCases: [ 'Building labeled datasets for fine-tuning', 'Identifying patterns in problematic responses', 'Measuring user satisfaction', 'Comparing model versions', 'Creating ground truth for automated evaluations', ], }; case 'search': return { title: 'Search', description: 'Opik provides powerful search capabilities to help you find specific traces based on content, metadata, or performance characteristics. This makes it easy to analyze patterns and identify issues.', keyFeatures: [ 'Full-text search across inputs and outputs', 'Metadata filtering', 'Time-based queries', 'Performance metric filtering', 'Support for complex boolean queries', ], useCases: [ 'Finding examples of specific behaviors', 'Identifying patterns in errors', 'Analyzing performance trends', 'Auditing responses for compliance', 'Building targeted datasets for evaluation', ], }; case 'visualization': return { title: 'Visualization', description: 'Opik provides visualization tools to help you understand your traces and spans. These visualizations make it easier to identify patterns, bottlenecks, and issues in your LLM applications.', keyFeatures: [ 'Timeline views of traces and spans', 'Performance dashboards', 'Metric trend analysis', 'Comparison views for A/B testing', 'Custom visualization options', ], useCases: [ 'Performance monitoring over time', 'Comparing model versions', 'Identifying usage patterns', 'Tracking evaluation metrics', 'Presenting insights to stakeholders', ], }; default: return null; } }