Vusal Dadalov

This is a comprehensive tutorial on the Retrieval-Augmented Generation (RAG) techniques available, along with practical implementation examples. Summary of RAG Techniques ►Simple RAG – Basic retrieval and generation pipeline. ►Semantic Chunking – Break documents into semantically meaningful chunks. ►Context-Enriched Retrieval – Augment queries with additional context for better retrieval. ►Contextual Chunk Headers – Add headers to chunks to provide context during retrieval. ►Document Augmentation – Enhance documents with supplementary information. ►Query Transformation – Modify queries to improve retrieval effectiveness. ►Re-Ranker – Re-rank retrieved documents based on relevance. ►RSE (Retrieval Score Enhancement) – Boost retrieval scores for better results. ►Contextual Compression – Compress context to fit model input limitations. ►Feedback Loop RAG – Iteratively refine retrieval based on previous outputs. ►Adaptive RAG – Dynamically adjust retrieval strategies based on context. ►Self RAG – Utilize the model's own outputs for subsequent retrievals. ►Knowledge Graph RAG – Incorporate knowledge graphs into retrieval. ►Hierarchical Indices – Organize data in hierarchical structures for retrieval. ►HyDE (Hypothetical Document Embeddings) – Generate hypothetical documents to aid retrieval. ►Fusion – Combine multiple retrieval methods for improved results. ►Multi-Model RAG – Use multiple models in the retrieval and generation process. ►CRAG (Chain-of-Thought RAG) – Implement chain-of-thought reasoning in RAG. Full article - https://lnkd.in/g3NhwziU

62

2 commentaren

3x-9x Faster Parquet Metadata Parsing in Rust 🚀 Every Parquet file ends with a footer that encodes the file’s metadata - schema definitions, column-level statistics (min/max/null counts), row group offsets, and encoding information. This metadata is what enables pruning and predicate pushdown, but before any of that happens, it has to be parsed. That parsing step, powered by Apache Thrift, sometimes become a performance bottleneck - especially for wide tables or latency-sensitive workloads. To tackle this, the Arrow Rust team introduced a custom Thrift parser in the parquet crate (v57.0.0) that makes metadata parsing 3×–9× faster. Some technical details: - Direct parsing: Skips intermediate Thrift-generated structs and builds the in-memory Parquet metadata directly. - Allocation efficiency: Hand-optimized code paths cut unnecessary heap allocations and copying. - Maintainability: Rust macros generate Thrift-derived definitions (like serde), while hot paths are fine-tuned by hand for critical structures such as RowGroup and ColumnChunk. Impact: ✅ Less latency to pruning: Faster access to column statistics means engines can start pruning and scanning sooner ✅ Better fit for real-time and RAG workloads: Reduced footer parse time improves throughput for low-latency queries. ✅ Influence across ecosystems: Highlights that most Parquet readers (in Java/C++) could gain similar speedups by moving away from generic Thrift-compiler parsers. Read more in this detailed blog by the Arrow team. #dataengineering #softwareengineering

460

7 commentaren

Apache Parquet Vs New File Formats. There’s no denying the impact Apache Parquet has had on the analytics ecosystem. - Columnar layout - Strong compression & encoding - The de facto format in data lakes and warehouses for over a decade But there are also a lot of new things happening with our analytical workloads. ❌ Workloads are no longer just batch analytics - they involve AI pipelines, low-latency use cases, and hardware-accelerated compute ❌ Hardware is not only CPU but also GPU, from x86 to ARM/RISC-V, with wide SIMD instruction sets becoming the norm ❌ And performance bottlenecks are seen in decompression speed, memory pressure and lack of vectorized execution paths As a result, we’re seeing a wave of innovation around next-gen file formats. Efforts like: BtrBlocks – introduces cascaded compression with lightweight codecs Nimble (Meta) – optimized for fast scan & inference LanceDB – tailored for vector search and ML use cases And the past week I read about “Fastlanes” Here are some highlights from their VLDB paper ✅ Expression Encoding: Flexible chains of lightweight codecs (FFOR, DELTA, DICT, FSST) that outperform heavyweight compression like Zstd ✅ Multi-Column Compression: Exploits inter-column correlations (e.g. equality, one-to-one mappings) to go beyond traditional columnar compression ✅ Segmented Layout: Decompresses in small vectors (1024 values) rather than rowgroups, reducing memory pressure and improving cache efficiency ✅ Compressed Execution Support: Returns compressed vectors to engines like DuckDB/Velox for SIMD/GPU-friendly query execution Here are some numbers I picked up on Performance: - 800× faster random access than Parquet+Zstd - 41% better compression than Parquet+Snappy - 2% better compression than Parquet+Zstd (without using heavyweight codecs) - Decoding accelerates with AVX512 (up to 40% faster) I think we are going to see more of fit-for-purpose formats in the future as we tackle specific use cases around AI workloads and to target performance. Paper link in comments. #dataengineering #softwareengineering

379

16 commentaren

PostgreSQL feels slower than usual. Now what? Before reaching for a bigger instance, take a breath and check the basics. Most Postgres slowdowns follow familiar patterns. These are the first places worth looking: ✅ Query plans — Is the planner doing what you expect? ✅ Indexes — Are they aligned with current access patterns? ✅ Autovacuum — Is it keeping up with churn-heavy tables? ✅ Connections — Are you pooling efficiently under load? ✅ Memory — Are key queries spilling to disk? ✅ Locks — Is something blocking the queue? ✅ Logs — Are you logging slow queries and temp files? ✅ App patterns — Are your ORMs over-fetching or repeating work? Asking the right questions early can save hours of firefighting. Try this, before 'scaling by credit card'!

47

6 commentaren

🚨 PostgreSQL 18 Beta is here 🚨 Highlights: ⚡️ Async I/O (with io_uring) — 2–3x speedups on seq scans, vacuums 🔍 Skip scan + smarter OR/IN optimizations 🔄 Keep planner stats during major upgrades 🧬 uuidv7() and virtual generated columns 🔐 OAuth login + md5 deprecation warning 📊 EXPLAIN ANALYZE now shows I/O, CPU, WAL 🧱 Temporal constraints, LIKE on nondeterministic collation, casefolding 🧪 New wire protocol version: 3.2 (first since 2003!) Your mission, should you choose to accept it: Test the beta. Break things. Report bugs. Help shape Postgres 18.

131

8 commentaren

🚀 OCRFlux: Open-source breakthrough in document parsing OCRFlux uses a 3B vision-language model fine-tuned from Qwen 2.5-VL-3B-Instruct to convert PDFs and images into clean Markdown — even with complex layouts and split tables across pages. 🔹 Achieves 98.6% F1 on benchmarks 🔹 Outperforms larger models like olmOCR-7B on 2,000-page tests 🔹 Handles natural reading order, preserves tables, and merges fragmented content seamlessly Proves smaller, specialized models can beat bigger ones for real-world document understanding. #documentparsing #ocr #opensource #pdfparser Link: https://lnkd.in/gha6DJHy

164

5 commentaren