Open-source multi-modal RAG for building AI apps over private knowledge
Top 16.3% on sourcepulse
Morphik provides a multi-modal Retrieval Augmented Generation (RAG) framework for building AI applications over private knowledge bases, targeting developers who need to process and query complex, visual documents. It offers a unified approach to ingest, search, transform, and manage unstructured and multimodal data, enabling advanced search capabilities and knowledge graph construction.
How It Works
Morphik employs techniques like ColPali for multimodal search, allowing queries across images, PDFs, videos, and more via a single endpoint. It facilitates the creation of domain-specific knowledge graphs with a single line of code, leveraging battle-tested system prompts or custom ones. The system also offers fast and scalable metadata extraction, including bounding boxes and classification, and features a cache-augmented generation mechanism for faster responses by creating persistent KV-caches of documents.
Quick Start & Requirements
pip install morphik
) or Docker.Highlighted Details
Maintenance & Community
The project welcomes contributions via GitHub issues and pull requests. Focus areas include speed improvements, tool integrations, and research paper integration. Community support is available via Discord.
Licensing & Compatibility
Features outside the ee
namespace are open-source under the MIT Expat license. Features within the ee
namespace have a different license and are not available in the open-source version.
Limitations & Caveats
Full support for open-source deployments is limited due to resource constraints, with community support available via Discord. Certain features, like the Morphik Console, are exclusive to the paid/hosted version.
2 days ago
1 day