Vision RAG: Enabling Search on Any Documents

Summary

Traditional retrieval-augmented generation relies on text parsing and optical character recognition to extract data from complex visual documents. These legacy techniques introduce high engineering effort, inconsistent accuracy, and heavy scaling costs. Vision RAG resolves these challenges by embedding entire visual documents directly through multimodal models like Voyage AI's voyage-multimodal-3.5, removing the need for text extraction pipelines. In this pipeline, dense vector embeddings map both text queries and visual assets into a unified representation space. At query time, the system retrieves relevant document images and supplies them alongside the text prompt to Anthropic's Claude 3.5 Sonnet to generate visually grounded responses.

Context

Enterprise knowledge frequently resides in complex documents, slide decks, PDFs, and diagrams containing interleaved text and visuals. Traditional text RAG depends on OCR and parsing tools to extract this data, resulting in brittle setups, high operational costs, and low extraction fidelity.

Approach / What changed

Implement a Vision RAG pipeline using Voyage AI's voyage-multimodal-3.5 model to create unified vector embeddings directly from document images and Anthropic's Claude 3.5 Sonnet to answer queries from retrieved visual assets.

Takeaways

• Unlike CLIP-based architectures that use dual encoders, Voyage AI's voyage-multimodal-3 and 3.5 models use a single encoder for text and images to prevent modality bias.
• Vision RAG indexes document pages and charts directly as images, bypassing complex optical character recognition and parsing pipelines.
• At runtime, retrieved images are converted to base64 and passed directly alongside user queries to a vision-capable large language model for generation.

Related reading

• MongoDB · Jan 15, 2026

  MongoDB.local San Francisco 2026: Ship Production AI, Faster

  MongoDB announced several features aimed at streamlining AI application development across its database and search ecosystem. The Voyage 4 embedding model family is now generally available, introducing cross-model compatibility within a shared embedding space alongside the video-capable voyage-multimodal-3.5 model. For in-database operations, MongoDB Community Edition introduced Automated Embedding in public preview to eliminate external synchronization pipelines for vector generation. Vector Search gained Lexical Prefilters, replacing the knnBeta operator with a dedicated vectorSearch operator and vector data type inside $search aggregations. Additionally, MongoDB made its intelligent assistant generally available in Compass and released the source code for its mongot search engine under the SSPL.

  Peter Richards

• MongoDB · Oct 1, 2025

  Smarter AI Search, Powered by MongoDB Atlas and Pureinsights