NexusRAG brings image, table-aware RAG to Ollama

Image and table blindness is one of the most persistent pain points in production RAG — NexusRAG's approach of caption-augmented chunking is pragmatic and deployable today without exotic infrastructure.

• –Vision LLM captions are appended directly to page chunks before embedding, meaning no separate image index or retrieval path is needed — a clean architectural choice
• –Tables are exported as structured Markdown, summarized by an LLM, then injected back into the chunk — preserves both structure and semantic searchability
• –Hybrid chunking via Docling's HybridChunker respects document structure, never splitting mid-table or mid-heading
• –Three-stage retrieval (ChromaDB vector over-fetch → LightRAG knowledge graph → BGE cross-encoder reranker) is more sophisticated than most hobbyist RAG projects
• –Full Ollama support makes this fully air-gapped deployable — relevant for enterprise and privacy-sensitive use cases