StackMap
Subscribe

morphik-core vs PixelRAG

Multimodal retrieval engine for visually rich documents: ingestion, visual-first search over charts, tables and diagrams, knowledge graphs and cache-augmented generation — one engine, not a pipeline. — versus — Berkeley's visual RAG: render pages and PDFs to screenshot tiles and retrieve with a VLM embedder — tables, charts and layout survive. pixelshot CLI plus a hosted 8.28M-page Wikipedia index.

The curated verdict

Both bet on visual retrieval for documents: PixelRAG is Berkeley's research recipe (screenshot tiles + VLM embedder); Morphik Core is the production engine with storage, KG and an API around the same idea.

morphik-corePixelRAG
Stars3.7k9.3k
Forks322788
LanguagePythonPython
LicenseNOASSERTIONApache-2.0
Last activity17 days ago9 days ago
Topicsragrag, vision
Curated connections23

morphik-core — the curator's take

Built on the observation that RAG dies on visual documents — charts become word soup — so retrieval is visual-first end to end, and the whole extract-embed-store-retrieve chain is one system instead of duct tape. When NOT: license is NOASSERTION, and the company has pivoted to nursing-home back-office AI with Core as the engine underneath — gauge maintenance commitment before betting your stack; for text-only corpora a plain vector store is simpler and cheaper.

PixelRAG — the curator's take

The paper's claim — screenshots beat parsed text for RAG — matters when your answers live in visual structure: tables, charts, infographics, layout-heavy PDFs that text chunkers flatten into noise. Zero-setup on-ramp is real: a hosted 8.28M-page Wikipedia index with no API key, and the pixelbrowse Claude Code plugin gives the agent screenshot-based page reading (pixelshot via Playwright/CDP, no MCP server). NOT a drop-in for text RAG stacks: your reader must be a VLM, tile indexes cost more storage/compute than text embeddings, and retrieval quality rides on their LoRA-tuned Qwen3-VL-Embedding model. Research codebase — expect pipeline assembly, not a product.