Micha(el) Bladowski’s Post

Pokee AI released PokeeResearch-7B, an open-source 7B research agent with code, weights, and paper, claiming top results in its class. It supports local or API use, web search, self-verification, and is licensed under Apache-2.0. #release

We’ve all been there—business users complain that 🔍search results “don’t make sense” or ask why certain documents appear first. Until now, ❄️Cortex Search was essentially a black box. We got great hybrid results (vector + keyword + semantic reranking), but explaining the “why” behind rankings was impossible Component Scores changes that entirely. Now we can get granular scoring breakdowns showing: 1.Keyword match strength 2.Semantic relevance score 3.Specific function score like text boost , vector boost , time decay (if defined in the cortex search ) How this is useful ? ✅ Debug search quality - See if results are too keyword-heavy vs. semantic ✅ Optimize user experience - Fine-tune which scoring components matter most ✅ Build trust with users - Show them why specific articles ranked highest ✅ Iterate faster - Identify patterns in poor-performing queries This feature transforms Cortex Search from a “trust us, it works” solution into a transparent, debuggable system that enterprise teams can actually optimize and explain to stakeholders. For those building RAG applications or enterprise search—this is the observability layer we’ve been waiting for. #Snowflake #CortexSearch #EnterpriseSearch #RAG #DataEngineering #AI

New Video!!! 🎉 In this video, you will learn how to build a Retrieval-Augmented Generation (RAG) web crawler from scratch using Chroma, Ollama, and LlamaIndex. We’ll walk through crawling websites, storing data in a vector database, and integrating it all with LLMs for smart, context-aware answers. This is perfect for devs who want to build real-world knowledge pipelines. Full video here: https://lnkd.in/dNMb4_eD #llm #ai

63% of all websites already see traffic from AI agents, and that share is only climbing. Those agents thrive on one thing: real-time data. In my comments for Built In's recent article, I explain how search APIs are becoming the backbone of this ecosystem, providing essential data for companies that no longer want to collect and index data themselves. Companies can forget the challenges they previously faced with data collection by calling an API to retrieve up-to-date information. This shift allows businesses to prioritize innovation over infrastructure-building, something we are seeing our own customers, both enterprise and start up embrace as a strategic move. Many of them are using Bright Data's SERP API, a powerful tool that can deliver search engine results in multiple formats in just one second. Here's a link to the full article. https://brdta.com/4pPzLwe #SearchAPI #OpenWeb #DataInfrastructure

🚀 Perplexity launches the Search API bringing its powerful retrieval engine to developers Perplexity just opened up the same search infrastructure that powers its answer engine, giving devs direct access to high-quality, real-time web data. Here’s what it unlocks 👇 🔍 Sub-document Retrieval → Fetch snippets or sections within pages instead of full documents for cleaner, more precise results. ⚙️ Structured Responses → Get JSON-formatted outputs that plug directly into RAG systems, chatbots, and agent workflows. ⏱ Real-time Freshness → The API indexes the live web continuously for the latest information. 🧠 Evaluation Framework → Benchmark and compare retrieval quality across APIs with built-in tools. 💰 Optimized for Scale → Fast, affordable, and designed for AI-first applications. Perplexity’s Search API bridges the gap between search and reasoning perfect for anyone building research copilots, retrieval layers, or autonomous agents. Would you replace your current web search API with this one? #AI #Perplexity #SearchAPI #RAG

New release of HyDRA v0.2 is here! 🐍 HyDRA: Hybrid Dynamic RAG Agent. For addressing the limitations of simple, static RAG. HyDRA is the answer. It's an advanced, unified framework for agentic RAG, inspired by the latest research to create something truly powerful. 🧠 Moving beyond single-shot retrieval. HyDRA introduces a multi-turn, reflection-based system with coordinated agents: a Planner, Coordinator, and Executors (currently local & deep web search). 🔬 At its core is an advanced 3-stage local retrieval pipeline that leaves basic RAG in the dust: 🥇 1. Hybrid Search: Combines dense (semantic) and sparse (textual) embeddings in one go using the bge-m3 model. This alone is a massive upgrade. 🥈 2. RRF (Reciprocal Rank Fusion): Intelligently merges and reranks results from different search vectors for ultimate precision. 🥉 3. Advanced Reranking: Uses the bge-m3-reranker model to score and surface the absolute most relevant documents for any query. ⚡️ This isn't just powerful, it's blazing fast. We're using SOTA ANN (HNSW) with vector and index quantization (down to 1-bit!) for near-instant retrieval with minimal quality loss. 🤖 But HyDRA is more than just retrieval. It incorporates memory from experience and reflection, creating a guiding policy for smarter future interactions and strategic planning. The result? A local retrieval system that significantly outperforms standard vector search RAG. 🌐 For deep web searches, HyDRA leverages the asynDDGS library and mcp (Model Context Protocol) for free, unrestricted web access. The entire reasoning engine is powered by the incredibly fast and efficient Google Gemini 2.5 Flash! 👨💻 Explore the project, dive into the code, and see it in action: 🔗 GitHub: https://lnkd.in/d2DtDSBy 🤝 Looking to implement cutting-edge AI solutions or collaborate? Let's connect! LinkedIn: https://lnkd.in/dp8n5C-G Email: hassenhamdi12@gmail.com Discord: hassenhamdi #AI #RAG #AgenticAI #LLM #GenerativeAI #OpenSource #NLP #MachineLearning #Gemini #VectorSearch #Innovation #Tech #Milvus #GenAI #Research #Agent #Langchain

Remember my post on grounding with maps here: https://lnkd.in/ddNzCtbM Well, Google ADK 1.15.0 has made implementing this easier by adding the Google Maps grounding tool as a built-in tool in the ADK tools library: https://lnkd.in/dP8dkXvq 🚀 Key Features in ADK 1.15: 🗺️ Google Maps Grounding Tool (Built-in!) - Hyper-local AI responses with real-time business data - 250+ million places accessible through natural conversation - Seamless integration with existing ADK agents 📊 OpenTelemetry Support - `--otel_to_cloud` experimental support for comprehensive monitoring - GenAI Instrumentation built into the framework - End-to-end tracing of AI workflows 🧠 Context Caching & Static Instructions - Context caching for faster response times - Static instructions that don't change (perfect for system prompts) - Auto-creation and lifecycle management of context caches What I Know So Far Before Building With It: Based on the source code, the GoogleMapsGroundingTool: - Only works with Gemini 2.x models (not Gemini 1.x) - Requires VertexAI (GOOGLE_GENAI_USE_VERTEXAI=TRUE) - Operates internally within the model - no local code execution - Automatically invoked by Gemini 2 models for grounding queries - Built-in tool that doesn't require manual configuration Which one are you most excited to try? #AI #GoogleADK #MachineLearning #OpenTelemetry #GoogleMaps #VertexAI #SoftwareDevelopment

A significant step forward for AI web access. Bright Data has released "The Web MCP," a powerful Model Context Protocol (MCP) server designed to give AI assistants seamless and reliable access to live web data. This all-in-one solution for searching, crawling, and navigating the public web eliminates common roadblocks like blocks and CAPTCHAs, enabling AI clients to perform real-time research and data extraction efficiently. The availability of a free tier (5,000 requests/month) and an open-source GitHub repository lowers the barrier to entry for developers building the next generation of AI agents. Visit here: GitHub Repository: https://lnkd.in/dJqQQscr #AI #DataExtraction #MCP #WebScraping #DeveloperTools

When we started building Valyu Search, we didn’t want “relevant.” We wanted real-time. FreshQA is one of the toughest benchmarks for retrieval, and we came out on top against Google, Exa, and Parallel. This is what real-time search for AI looks like.

We've all been there: That moment when an AI agent confidently delivers an answer, only for it to be utterly useless because it's based on stale data or deprecated information. In the high-stakes world of finance research, scientific discovery, or even just coding, outdated context isn't just a nuisance; it's a silent killer of reliability, leading to frustrating hallucinations and critical misinformation. Building Valyu, we've wrestled endlessly with this challenge. It's a relentless battle to ensure AI agents have access to the freshest, most authoritative context, not just generic web scrapes. We realised that for AI to be truly trustworthy and impactful in real-world applications, freshness isn't a 'nice-to-have' - it's the absolute foundation. That's why our DeepSearch API is engineered from the ground up to deliver real-time, authoritative context. It means your agents are always accurate, whether they're analysing the latest market movers, understanding evolving scientific literature, or navigating constantly changing code packages. We're removing the friction of outdated training data so your AI applications can finally operate with confidence. I'd love for you to give it a spin and experience the difference first-hand! What challenges has outdated information posed for your AI builds? Check out DeepSearch here: https://platform.valyu.ai/ #AI #AIAgents #DataInfrastructure #DeepSearch #KnowledgeWork