lancer

What problem does it solve?

This Skill overcomes the limitations of keyword-based search, which often fails to capture the true meaning of queries, and the challenge of effectively searching multi-modal content (text and images). It provides a powerful semantic search solution for your knowledge base.

Core Features & Use Cases

• Multi-modal Ingestion: Add diverse documents (text, markdown, PDF, SQL, various image formats) to LanceDB, with automatic chunking and vector embeddings.
• Semantic Search: Find documents by meaning and context, not just keywords, across both text and images, using advanced embedding models.
• Table & Document Management: Organize your content into LanceDB tables, get table information, and remove specific documents as needed.
• Use Case: Ingest a collection of technical documentation and architectural diagrams. Then, use a natural language query like "sunset over mountains" to find relevant images, or "how to deploy kubernetes" to find both text passages and visual diagrams.

Quick Start

Ingest all markdown documents from the './documentation/' directory into a table named 'docs', then search that table for "installation guide".