Skill UI

This guide covers advanced retrieval strategies that significantly boost search quality beyond basic vector similarity search. Techniques include Hybrid Search (combining dense and sparse indexing), Reranking (improving result order), Relevance Feedback (guiding search with iterative user scores), and Maximal Marginal Relevance (MMR) for diversity. Use these methods when standard search results lack relevance, miss critical keywords, or exhibit poor precision.

A spectral vector search tool that enhances nearest-neighbor search with graph Laplacian features, capturing both semantic similarity and structural roles in embedding spaces.

A comprehensive guide for training or fine-tuning advanced sentence-transformer models. This skill assists in selecting the correct architecture—including Bi-encoders (SentenceTransformer), CrossEncoders (for reranking), and SparseEncoders—for various Natural Language Processing (NLP) tasks. Use cases include semantic retrieval, similarity matching, document clustering, and pair-wise classification.

This skill allows for the deterministic re-running of recorded user session paths (trajectories). It is essential for regression testing, bug reproduction, and verifying UI flow consistency after deployments. Crucially, it utilizes browser-selectors embedding similarity to intelligently recover and retry interactions even when the underlying webpage structure (DOM) has drifted, providing robust testing capabilities.

This tool simplifies the process of detecting structural drift by combining audit listing, fresh auditing, and trend diffing into a single command. It finds the most recent audit record, runs a current audit against the repository, and compares the two using structural similarity metrics. It alerts when the structural distance falls below a specified threshold, helping maintain code consistency over time.

Compute weighted similarity between two harness fingerprints using cosine, categorical agreement, and topology Jaccard. Use it to rank templates, detect drift, decide reuse vs scaffold, and generate confidence scores for recommender workflows.

This tool performs pathfinder traversal across a knowledge graph starting from a specified seed entity. It expands outward through causal edges, calculates path relevance scores, and prunes low-similarity branches. It is ideal for conducting impact analyses, tracing dependency chains, or understanding complex relationships within a structured knowledge base.

This tool ingests raw market data (OHLCV) for a specified symbol. It normalizes the data into relative OHLCV vectors, converts these vectors into high-dimensional embeddings, and then stores and indexes them using HNSW. This process prepares the data for fast, similarity-based pattern detection and advanced financial time series analysis.

This memory bridge ingests Claude Code's local auto-memory markdown files and structures them into AgentDB. It generates high-fidelity ONNX embeddings, handles deduplication using cosine similarity, and indexes the memories in a scalable namespace. This enables seamless, unified semantic search across all historical sessions and projects, significantly enhancing RAG capabilities.

A utility designed to generate high-dimensional vector embeddings (384-dim) from various sources including text, code, or documents. This process converts raw data into numerical representations optimized for semantic search, similarity comparison, and clustering, utilizing ONNX and HNSW indexing for efficient retrieval.

This plugin provides advanced vector search capabilities using two distinct paths. The large-scale path supports millions of vectors via HNSW, ideal for general corpus search. The hot-path router handles small, critical pattern routing (≤11 patterns) with sub-ms latency. Key features include RaBitQ 1-bit quantization for memory efficiency (32× reduction), hyperbolic space search for hierarchical data, and unified cross-namespace search.