Does this skill support scanned PDF documents?

Yes, by setting a MISTRAL_API_KEY in your configuration, the skill can leverage Mistral OCR to extract text from scanned PDFs that standard parsers might miss.

How can I improve the quality of my search results?

You can improve results by using more specific queries, adjusting chunk sizes (default 512), increasing the search limit with the --limit flag, or ensuring documents are properly indexed using the list command.

What embedding models does the rag-research skill support?

The skill utilizes FastEmbed with ONNX Runtime, supporting BGE-small-en-v1.5 (default), BGE-base, and BGE-large models for efficient CPU-based vector generation.

Where are the indexed vectors stored?

Vectors and metadata are persisted locally in a Qdrant database located in the ~/.rag-research directory on your machine.

RAG Research & Semantic Search

Name: RAG Research & Semantic Search
Author: docutray

bydocutray

•

Data Science & ML

Optimizes document retrieval and semantic search workflows using RAG best practices, vector databases, and advanced chunking strategies.

About

The rag-research skill empowers Claude to efficiently manage Retrieval-Augmented Generation workflows by providing domain-specific guidance on document indexing, embedding generation, and vector storage. It streamlines the process of converting PDFs, Markdown, and text files into searchable vectors using FastEmbed and Qdrant, offering actionable advice on chunking sizes, model selection, and search quality optimization to ensure high-relevance retrieval during the development process.

Key Features

Advanced text chunking and overlap strategies for context preservation
2 GitHub stars
Support for Mistral OCR to process and index scanned PDF documents
Automated document indexing for PDF, Markdown, and Text files
Integrated vector storage management with Qdrant persistence
Configurable embedding generation using FastEmbed BGE models

Use Cases

Building a local knowledge base from technical documentation for AI retrieval
Troubleshooting and refining embedding models for domain-specific data
Optimizing search relevance and retrieval quality in RAG-based applications

About

Key Features

Advanced text chunking and overlap strategies for context preservation
2 GitHub stars
Support for Mistral OCR to process and index scanned PDF documents
Automated document indexing for PDF, Markdown, and Text files
Integrated vector storage management with Qdrant persistence
Configurable embedding generation using FastEmbed BGE models

Use Cases

Building a local knowledge base from technical documentation for AI retrieval
Troubleshooting and refining embedding models for domain-specific data
Optimizing search relevance and retrieval quality in RAG-based applications