Semantic Search for Creative Assets

7 posts · Mar 2026 – Sep 2026

What this trail covers

Foundry turns an existing archive into a searchable creative library by scanning files, recording metadata, preparing assets for retrieval with descriptions and tags, and synchronizing this data with Weaviate. It supports keyword, semantic, and hybrid search modes, linking results back to original source files without moving or renaming them. The system uses content hashes for stable identity and change detection, and allows for filtering by project, file type, and relationship role. This post details the application of semantic search to creative asset management, focusing on reducing friction in retrieval and reuse.

All Weaviate trails

Sep 2026Building Foundry Part 3: From archive to creative search | WeaviateThis post details the implementation of Foundry's scanning process, which includes capturing source path, project, file type, size, modified date, and content hash. It also describes the preparation stage where descriptions, tags, extracted text, relationship roles, and a source URI are added to records. The synchronization step with Weaviate is explained, highlighting the creation or update of the 'Foundry' collection and the use of deterministic identifiers for updating objects. The post also contrasts keyword, semantic, and hybrid retrieval modes and mentions the possibility of adding filters for project, file type, and relationship role.
Sep 2026How to extract meaning from charts and tables in PDFs | WeaviateIntroduces a new approach to RAG for PDFs that leverages late-interaction multi-vector models to directly embed and retrieve information from charts and tables, bypassing traditional OCR and text chunking. Demonstrates this capability with Weaviate Cloud's drag-and-drop importer and the `multi2multivec-weaviate` module. Provides examples of querying financial documents (NVIDIA earnings) for visual data like revenue trends and charts. Also shows how to integrate this with the Weaviate Query Agent for synthesized answers with citations.
Aug 2026Building Foundry Part 2: Where creative workflows break | WeaviateThis post details the breakdown of traditional organization methods (folders, tags, keyword search) in creative workflows due to scaling, deadline pressure, and vocabulary mismatch. It argues for a retrieval-centric approach that combines semantic similarity, keyword matching, and metadata filtering. The post outlines an ingestion-first strategy for preparing assets with embeddings and metadata, and a retrieval-second strategy for combining multiple signals to achieve production-safe results. It uses Weaviate as an example for implementing this hybrid search.
Jul 2026Building Foundry: AI isn’t replacing creativity, it’s removing friction | WeaviateThis post details the application of semantic search to creative asset management, explaining how vector embeddings and hybrid search can solve retrieval problems in creative workflows. It highlights the ingestion process of generating embeddings and storing them alongside metadata in a vector database like Weaviate. The post contrasts keyword search with semantic search, using examples from game development, film editing, and graphic design to illustrate the benefits of searching by meaning.
Jun 2026Import & Vectorize Data with Weaviate at Scale | WeaviateIntroduces server-side batching for efficient data import, a retry-with-checkpoint pattern for robust error handling, and the `blobHash` data type for storing media embeddings while omitting raw bytes. Also discusses multimodal ingestion for PDFs and images using Weaviate Embeddings and `multi2vec-google`.
Advertisement
May 2026Your LLM Is Only as Good as What It Retrieves | WeaviateThis post details the critical role of retrieval quality in Retrieval-Augmented Generation (RAG) systems and its impact on LLM output. It identifies five specific retrieval failure modes: retrieval drift, context truncation, stale index poisoning, low-relevance top-k retrieval, and inter-agent miscommunication. It proposes four key areas for improving retrieval quality: embedding model selection, chunking architecture, retrieval strategy (including hybrid search, cross-encoder re-ranking, and relevance thresholding), and index maintenance and freshness. Finally, it outlines a practical measurement framework for evaluating retrieval quality using metrics like context precision and context recall.
Apr 2026Multimodal Embeddings and RAG: A Practical Guide | WeaviateIntroduces and details the practical application of multimodal embeddings for semantic search and RAG. Explains how natively multimodal embedding models (like Gemini Embedding 2) map text, images, audio, and video into a shared embedding space. Discusses key design decisions: native vs. bridge-based embedding, chunking strategies for non-text data (fixed-width time windows for audio/video, page-level for documents), and the use of Matryoshka Representation Learning (MRL) for dimension reduction. Provides concrete examples of building multimodal systems with Weaviate, Gemini Embedding 2, and Gemini 3 Flash, including searching audio without transcripts, searching images with text, and multimodal RAG with video.