Knowledge Share
Technical articles, tutorials, and insights
RAG Series (23): Multimodal RAG — Images and Tables Can Be Retrieved Too
Part 23 of the RAG series. 30–50% of the information in real documents lives in images and tables — invisible to text-only RAG. Three approaches: extract and textualize (most mature), CLIP multimodal embeddings (text and images in the same vector space), ColPali (process each PDF page as an image directly, bypassing text extraction entirely — the 2024 breakthrough). When to use each, and a practical decision guide.
RAG Series (24): Code RAG — Teaching AI to Understand Your Codebase
The final article in the RAG series. Code is not a document — it has structure (functions/classes), semantics (docstrings), and call relationships (call graph). Plain text chunking destroys all three. This article uses AST to parse the llm-in-action codebase (22 Python files, 225 code units), builds a call graph (168 edges), and implements semantic code search plus call chain traversal. Real results: build_self_rag_graph's full downstream chain, main's 54 direct calls, build_index's call tree — all found correctly.
RAG Series (22): Long Context vs RAG — Do We Even Need RAG?
Part 22 of the RAG series. Gemini 1.5 Pro and Claude's million-token context windows have led some to declare RAG obsolete. This article unpacks the actual cost of each approach — long context is not free lunch on cost and latency; RAG's retrieval is imperfect and incomplete. A decision framework across four dimensions (document volume, update frequency, query volume, latency requirements) plus a hybrid strategy that uses both.