Morphik

Documents full of diagrams, tables, and screenshots are where text-only RAG breaks down, and Morphik attacks the problem by embedding full pages visually. Its ColPali-based retrieval treats each page as an image, so charts and layouts are searched directly instead of being flattened through lossy OCR, and metadata extraction adds bounding boxes and classification on top. The core engine, morphik-core, self-hosts via Docker or direct install with ingestion, deep and shallow search, transformation, and document management APIs, plus MCP support and integrations for Google Workspace, Slack, and Confluence, while a hosted cloud with a free tier is the recommended path since self-hosting receives limited community support. Licensing is Business Source License 1.1: free for personal and indie use and for commercial deployments under 2,000 dollars a month in revenue, with each release converting to Apache-2.0 four years after it ships, so the code is source-available rather than strictly open source. The developer project and its docs live at dev.morphik.ai, and the repository has about 3,600 stars with active 2026 commits.