Retrieval, RAG & Search

Build Production-Grade RAG: Practical Choices for Vector Databases, Hybrid Search and Safe ES|QL Behavior

Retrieval, RAG & Search10 hours ago6Views 0Likes 0Comments

What Happened Elasticsearch 9.5 introduced an unmapped_fields option for ES|QL that prevents queries from failing when they reference fields missing from index mappings. The option accepts NULLIFY (return NULLs) or LOAD (read values from _source) and resolves partially unmapped non-keyword fields (PUNKs) by injecting an “unmapped” field into the query plan so the planner and…

Make Retrieval-Augmented Generation Reliable: Practical Vector‑Search Tuning, Index Choices and Deployment Patterns

Retrieval, RAG & Search2 days ago18Views 0Likes 0Comments

What Happened Teams tuning vector search often find that changing one index parameter moves an evaluation score by a small amount (for example, 0.01) and it’s unclear whether the change actually improved relevance or just reflected ranking variability. The underlying APIs define knobs such as HNSW ef, reciprocal‑rank fusion k, and quantization oversampling, but the…

How to Build Efficient, Cost‑Predictable RAG Systems: Reduce Retrieval Loops and Scale Vector Search

Retrieval, RAG & SearchAugust 14, 202663Views 0Likes 0Comments

What Happened Recent engineering results show that reducing repeated retrievals inside RAG agent loops materially improves throughput and GPU efficiency: Qdrant combined with Minima reported a 2.92× increase in agentic RAG tasks per GPU‑hour by addressing the search → inspect → retry pattern that compounds latency, context consumption and inference cost [1]. At the same…

How to Build Production-Grade RAG Search: Hybrid Retrieval, Vector Stores, and Operational Best Practices

Retrieval, RAG & SearchAugust 13, 202671Views 0Likes 0Comments

What Happened Organizations are moving from ad-hoc search to retrieval-augmented generation (RAG) backed by vector databases and hybrid retrieval. Large enterprises have proven this at scale — for example, Bayer built an enterprise-scale search engine using Qdrant to power cross-domain discovery and governance [1]. Work from creative teams shows the common failure mode: metadata and…

Retrieval, RAG & Search — August 11, 2026

Retrieval, RAG & SearchAugust 11, 2026138Views 0Likes 0Comments

What Happened Two recent operational notes illustrate complementary advances and failure modes for retrieval‑augmented generation (RAG) systems. Weaviate’s Query Agent introduced a Search Mode with an effort parameter (medium / high / ultrahigh) that scales test‑time compute for query writing and reranking. Search Mode returns ranked documents (not answers), can synthesize structured filters…

Retrieval, RAG & Search — August 10, 2026

Retrieval, RAG & SearchAugust 10, 2026101Views 0Likes 0Comments

What Happened Retrieval-augmented generation (RAG) is now the default pattern for production knowledge and assistant applications: ingest documents, create embeddings, run nearest-neighbor retrieval (often hybrid dense+sparse), and combine retrieved context with a generator LLM. The ecosystem around that pattern — orchestration libraries (LlamaIndex, LangChain, Haystack), managed vector services (Pinecone), open-source vector engines (Qdrant, Milvus), semantic…

Illustration for the Kimbodo News & Research briefing “Build Faster, Cheaper and More Accurate RAG Systems by Choosing the Right Vector Store and Search Pattern” (Retrieval, RAG & Search).

Build Faster, Cheaper and More Accurate RAG Systems by Choosing the Right Vector Store and Search Pattern

Retrieval, RAG & SearchAugust 5, 202668Views 0Likes 0Comments

What Happened Recent advances in vector database storage and memory management change the cost/latency/accuracy trade-offs for retrieval-augmented generation (RAG). Qdrant 1.19 introduces three features that are immediately relevant to production RAG deployments: a 4-bit TurboQuant compressed vector datatype that discards full-precision vectors for large storage reduction, unified memory tiers (pinned, cached, cold) for per-component placement,…

Replace Fragile Keyword Search with RAG and Vector Databases to Find Lost Assets Faster

Retrieval, RAG & SearchJuly 30, 202684Views 0Likes 0Comments

What Happened Creative and knowledge teams accumulate large, heterogeneous content over years—images, sketches, audio, notes and many iterative revisions. Traditional keyword and folder-based search breaks when naming conventions change, metadata is incomplete, or people leave, making it often faster to recreate than to locate existing work. Semantic search, powered by vector embeddings for text, images…

Retrieval, RAG & Search — July 28, 2026

Retrieval, RAG & SearchJuly 28, 202686Views 0Likes 0Comments

What Happened Several recent engineering advances and findings change practical choices for retrieval‑augmented generation (RAG) and production vector search: Elastic Agent Builder now emits full OpenTelemetry traces for every LLM call and tool execution; teams can convert those traces into token‑cost and performance dashboards in Kibana to operationalize cost and latency visibility [1].…

Illustration for the Kimbodo News & Research briefing “Cut RAG Latency and Improve Retrieval Accuracy by Adding Listwise Rerankers and Modern Vector‑store Patterns” (Retrieval, RAG & Search).

Cut RAG Latency and Improve Retrieval Accuracy by Adding Listwise Rerankers and Modern Vector‑store Patterns

Retrieval, RAG & SearchJuly 27, 202688Views 0Likes 0Comments

What Happened Jina released a new text-only listwise reranker, jina-reranker-v3.5, a ~597–600M parameter model with a 131,072-token context window that uses a “last-but-not-late” one-pass scoring approach and a modified self-attention to reduce quadratic cost and speed inference. The model was trained with three-stage self-distillation on a multilingual (52 languages) and domain-expanded corpus (legal, medical, financial,…

Deliver Accurate, Low-Latency RAG Search: Practical Architecture and Best Practices for Vector Databases

Retrieval, RAG & SearchJuly 25, 202679Views 0Likes 0Comments

What Happened Vector databases and retrieval-augmented generation (RAG) tooling have converged into repeatable patterns for production search: dense-vector candidate generation, scalar-filtered recall, and a separate ranking/merchandising layer that composes multiple signals. Tooling improvements in Qdrant, Pinecone, Milvus, Weaviate and orchestration libraries (LlamaIndex, LangChain, Haystack) make sub-100ms query paths and large-scale catalogs practical. Qdrant’s recent engineering…

Retrieval, RAG & Search — July 24, 2026

Retrieval, RAG & SearchJuly 24, 202691Views 0Likes 0Comments

What Happened Over the last few years RAG systems have moved from demos to production services. Vector databases (Pinecone, Qdrant, Milvus, Weaviate) and search engines (Elasticsearch, Vespa, Elasticsearch KNN/ANN plugins, Haystack orchestration) now offer mature features for hybrid lexical + semantic retrieval, filtering, metadata/scoring, and operational controls. Orchestration libraries (LlamaIndex, LangChain, Haystack) have standardized pipelines…

Retrieval, RAG & Search

Engineering Founded & Customer Focused