{"slug": "qdrant", "name": "Qdrant", "category": "databases", "category_name": "Databases", "description": "", "recommenders": 4, "commenters": 26, "warners": 1, "score": 6.736, "rising": -1.31, "first": "2025-11-15", "latest": "2026-08-26", "why": "I use QDrant as the main DB, the one every RAG Pipeline requires", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 5, "neutral": 24, "compare": 2, "warn": 1}, "spark": [1, 0, 1, 1, 0, 0, 0, 1, 1, 0, 0, 0], "listed": true, "loved": [], "caveat": "", "short_quote": "I use QDrant as the main DB, the one every RAG Pipeline requires", "kind": "", "icon": true, "domain": null, "github_repo": "github.com/qdrant/qdrant", "official_url": "https://qdrant.tech", "pricing": "", "aliases": ["Qdrant"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 785571344, "item_id": 49098209, "url": "https://news.ycombinator.com/item?id=49098209", "by": "athrv_shrn", "date": "2026-07-29", "version": null, "stance": "recommend", "quote": "I use QDrant as the main DB, the one every RAG Pipeline requires", "excerpt": "I am using a system I designed on my own for the 'CRUD' style of thing for renames/deletions you are mentioning. People just hook up a Vector DB and store all the embeddings once and call it a day. I use three databases for the CRUD style system (idk its name, so i will just call it that lol). I use QDrant as the main DB, the one every RAG Pipeline requires, then I use Redis as a Raw storage for all the files and doc", "context": "", "use_case": "vector DB for RAG", "switched_from": "", "story": "Agent and RAG for Obsidian, Need Feedback"}, {"id": 778170272, "item_id": 48635642, "url": "https://news.ycombinator.com/item?id=48635642", "by": "vunderba", "date": "2026-06-22", "version": null, "stance": "recommend", "quote": "have been pretty happy with it", "excerpt": "I run a dockerized version of qdrant for RAG related stuff and have been pretty happy with it. https://github.com/qdrant/qdrant", "context": "", "use_case": "RAG vector database", "switched_from": "", "story": "Ask HN: What are some rock solid open source vector databases"}, {"id": 755233906, "item_id": 47202119, "url": "https://news.ycombinator.com/item?id=47202119", "by": "vunderba", "date": "2026-02-28", "version": null, "stance": "recommend", "quote": "in conjunction with a Qdrant vector database", "excerpt": "I can't speak to using local models as agentic coding assistants, but I have a headless 128GB RAM machine serving llama.cpp with a number of local models that I use on a daily basis. - Qwen3-VL picks up new images in a NAS, auto captions and adds the text descriptions as a hidden EXIF layer into the image, which is used for fast search and organization in conjunction with a Qdrant vector database. - Gemma3:27b is use", "context": "", "use_case": "Image search indexing", "switched_from": "", "story": "Qwen3.5 122B and 35B models offer Sonnet 4.5 performance on local computers"}, {"id": 746062753, "item_id": 46628922, "url": "https://news.ycombinator.com/item?id=46628922", "by": "cbcoutinho", "date": "2026-01-14", "version": null, "stance": "recommend", "quote": "Qdrant supports storing embeddings in memory as well as in a local directory (similar to sqlite)", "excerpt": "The Nextcloud MCP Server [0] supports Qdrant as a vectordb to store embeddings and provide semantic search across your personal documents. This enables any LLM & MCP client (e.g. claude code) into a RAG system that you can use to chat with your files. For local deployments, Qdrant supports storing embeddings in memory as well as in a local directory (similar to sqlite) - for larger deployments Qdrant supports running", "context": "", "use_case": "vector storage for RAG", "switched_from": "", "story": "Ask HN: How are you doing RAG locally?"}, {"id": 736725521, "item_id": 46045345, "url": "https://news.ycombinator.com/item?id=46045345", "by": "andre-z", "date": "2025-11-25", "version": null, "stance": "recommend", "quote": "At least, with Qdrant, it works like this", "excerpt": "FAISS is not suitable for production. The dedicated vector search solutions solve all the issues you mentioned: you just store the metadata along with vectors in JSON format. At least, with Qdrant, it works like this: https://qdrant.tech/documentation/concepts/payload/", "context": "", "use_case": "Vector search", "switched_from": "", "story": "Ask HN: Scaling local FAISS and LLM RAG system (356k chunks)architectural advice"}]}], "caveats": [{"label": "", "summary": "", "quotes": [{"id": 743980676, "item_id": 46498792, "url": "https://news.ycombinator.com/item?id=46498792", "by": "saeedq", "date": "2026-01-05", "version": null, "stance": "warn", "quote": "including milvus, qdrant, and weaviate", "excerpt": "My name is Saeed, and in general, I'm a deep learning specialist since 2019, but within the past 2 years, I've been writing optimized algorithms with cuda/C++. Recently, I decided to build a vector engine that consumes less resources than vector databases while not sacrificing accuracy, and throughput. So, we created brinicle. It is a production-oriented ANN index engine (not a full vector database) designed to stay ", "context": "", "use_case": "", "switched_from": "", "story": "I designed a vector engine to be RAM/CPU efficient"}]}], "who": [{"by": "naman_411", "context": "AI engineer, ~2 years on production RAG and agent systems", "url": "https://news.ycombinator.com/item?id=49178550", "date": "2026-08-04"}, {"by": "aledevv", "context": "built a RAG project", "url": "https://news.ycombinator.com/item?id=47529197", "date": "2026-03-26"}, {"by": "Beefin", "context": "works at Mixpeek", "url": "https://news.ycombinator.com/item?id=47153228", "date": "2026-02-25"}, {"by": "p410n3", "context": "Uses vector search at work (300K products)", "url": "https://news.ycombinator.com/item?id=46554741", "date": "2026-01-09"}], "tags": [], "use_cases": ["vector search", "vector storage for rag", "vector database", "image search index", "vector retrieval backend", "vector db for rag"], "compared": [{"slug": "ollama", "name": "Ollama", "co": 7, "switched_from": 0, "switched_to": 0}, {"slug": "qwen", "name": "Qwen", "co": 4, "switched_from": 0, "switched_to": 0}, {"slug": "chroma", "name": "Chroma", "co": 4, "switched_from": 0, "switched_to": 0}, {"slug": "pgvector", "name": "pgvector", "co": 3, "switched_from": 0, "switched_to": 0}, {"slug": "mem0", "name": "Mem0", "co": 3, "switched_from": 0, "switched_to": 0}], "makers": [{"id": 761233233, "item_id": 47577077, "url": "https://news.ycombinator.com/item?id=47577077", "by": "mcheemaa", "date": "2026-03-30", "version": null, "stance": "neutral", "quote": "our Qdrant-based system", "excerpt": "This is awesome to hear. The \"I truly have no idea what the system or code looks like but it's been so much fun just letting it build\" resonates hard. That's exactly the experience we had too. The \"For People Who Don't Write Code\" angle has been the biggest surprise for us. We had a non-technical user ask for a Chrome extension and the agent built it, packaged it as a zip, and sent the download link. No terminal, no ", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Phantom \u2013 Open-source AI agent on its own VM that rewrites its config"}, {"id": 753379858, "item_id": 47086241, "url": "https://news.ycombinator.com/item?id=47086241", "by": "laramie_co", "date": "2026-02-20", "version": null, "stance": "recommend", "quote": "Qdrant for vector search.", "excerpt": "Hey HN \u2014 maker here. Quick context on why I built this: I was personally spending ~$1,200/month on Claude API (I use it for everything \u2014 coding, writing, analysis). That\u2019s $14,400/year. Even if I dropped to a single $20/month sub, every query I send tells Anthropic what I\u2019m working on, what I\u2019m thinking about, what problems I have. I wanted Claude-quality inference without the cloud dependency. That\u2019s not possible ye", "context": "", "use_case": "local RAG vector search", "switched_from": "", "story": "Show HN: VaultAI \u2013 42 AI models on a portable SSD, works offline ($399)"}], "threads": [{"url": "https://news.ycombinator.com/item?id=47741527", "title": "Ask HN: What Are You Working On? (April 2026)", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=47574045", "title": "Show HN: Phantom \u2013 Open-source AI agent on its own VM that rewrites its config", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=47534245", "title": "GERP \u2013 A Headless ERP Built with Go, Spanner, Temporal, and GCP Vector", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=49445727", "title": "RAG Is Simpler Than You Think", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49200652", "title": "Qwen3.8 Max now ranked as the best overall model by agentic index", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49156682", "title": "Ask HN: Who wants to be hired? (August 2026)", "mentions": 1}], "sentiment_by_month": {"2025-11": {"neutral": 2, "recommend": 1}, "2025-12": {"neutral": 2}, "2026-01": {"compare": 1, "warn": 1, "neutral": 3, "recommend": 1}, "2026-02": {"neutral": 5, "compare": 1, "recommend": 1}, "2026-03": {"neutral": 4}, "2026-04": {"neutral": 2}, "2026-05": {"neutral": 2}, "2026-06": {"recommend": 1}, "2026-07": {"neutral": 1, "recommend": 1}, "2026-08": {"neutral": 3}, "2026-09": {}, "2026-10": {}}, "versions": [], "version_unnamed": 0, "family_ref": null, "runs_on": null, "lines": [], "quotes": [{"id": 791182528, "item_id": 49448908, "url": "https://news.ycombinator.com/item?id=49448908", "by": "saltysalt", "date": "2026-08-26", "version": null, "stance": "neutral", "quote": "using a Vector Database like Qdrant for persistence", "excerpt": "If like me you run models locally, it's pretty easy to run your own RAG locally also using a Vector Database like Qdrant for persistence, and a middle-layer like Mem0 for realtime retrial and updates. I documented the set-up steps here: https://leadprompt.sh/a/739-Building-an-Infinite-Memory-Local-AI-Stack-on-Fedora", "context": "", "use_case": "Vector storage for RAG", "switched_from": "", "story": "RAG Is Simpler Than You Think"}, {"id": 787267218, "item_id": 49204201, "url": "https://news.ycombinator.com/item?id=49204201", "by": "vunderba", "date": "2026-08-06", "version": null, "stance": "neutral", "quote": "in conjunction with a Qdrant vector database", "excerpt": "Oh I do. Headless 128GB RAM machine serving llama.cpp with a number of local models that I use on a daily basis. \u2022 Qwen3-VL picks up new images in a NAS, auto captions and adds the text descriptions as a hidden EXIF layer into the image, which is used for fast search and organization in conjunction with a Qdrant vector database. \u2022 Gemma3:27b is used for personal translation work (mostly English and Chinese). \u2022 Some s", "context": "", "use_case": "image search index", "switched_from": "", "story": "Qwen3.8 Max now ranked as the best overall model by agentic index"}, {"id": 786856800, "item_id": 49178550, "url": "https://news.ycombinator.com/item?id=49178550", "by": "naman_411", "date": "2026-08-04", "version": null, "stance": "neutral", "quote": "rebuilt Qdrant retrieval with clause-aware chunking and cross-encoder reranking", "excerpt": "Location: Bengaluru, India (UTC+5:30) Remote: Yes, remote only Willing to relocate: No Technologies: Python, FastAPI, Qdrant, PostgreSQL, Redis, Docker, LangGraph, LangChain, MCP, TypeScript, Next.js R\u00e9sum\u00e9/CV: https://namangupta.dev Email: naman@namangupta.dev AI engineer, ~2 years on production RAG and agent systems. Contract or full-time. I make retrieval measurable. Most RAG that \"feels wrong\" has no eval set, so", "context": "AI engineer, ~2 years on production RAG and agent systems", "use_case": "Vector retrieval backend", "switched_from": "", "story": "Ask HN: Who wants to be hired? (August 2026)"}, {"id": 785571344, "item_id": 49098209, "url": "https://news.ycombinator.com/item?id=49098209", "by": "athrv_shrn", "date": "2026-07-29", "version": null, "stance": "recommend", "quote": "I use QDrant as the main DB, the one every RAG Pipeline requires", "excerpt": "I am using a system I designed on my own for the 'CRUD' style of thing for renames/deletions you are mentioning. People just hook up a Vector DB and store all the embeddings once and call it a day. I use three databases for the CRUD style system (idk its name, so i will just call it that lol). I use QDrant as the main DB, the one every RAG Pipeline requires, then I use Redis as a Raw storage for all the files and doc", "context": "", "use_case": "vector DB for RAG", "switched_from": "", "story": "Agent and RAG for Obsidian, Need Feedback"}, {"id": 782942194, "item_id": 48933887, "url": "https://news.ycombinator.com/item?id=48933887", "by": "cbcoutinho", "date": "2026-07-16", "version": null, "stance": "neutral", "quote": "powered by a Qdrant vectordb", "excerpt": "I have developed an open-source memory system for agents accessible over MCP, which makes it possible to access it via any coding agent locally (claude-code) or via mobile (Claude AI, Mistral AI, etc). The primary storage mechanism is .md files stored as Notes in Nextcloud; however, since Nextcloud supports a rich ecosystem of apps such as documents, rss feeds, calendar, etc, your knowledge base can grow with you. Al", "context": "", "use_case": "semantic search index", "switched_from": "", "story": "Open-source memory for coding agents, synced over SSH"}, {"id": 778170272, "item_id": 48635642, "url": "https://news.ycombinator.com/item?id=48635642", "by": "vunderba", "date": "2026-06-22", "version": null, "stance": "recommend", "quote": "have been pretty happy with it", "excerpt": "I run a dockerized version of qdrant for RAG related stuff and have been pretty happy with it. https://github.com/qdrant/qdrant", "context": "", "use_case": "RAG vector database", "switched_from": "", "story": "Ask HN: What are some rock solid open source vector databases"}, {"id": 773290289, "item_id": 48330643, "url": "https://news.ycombinator.com/item?id=48330643", "by": "kristopolous", "date": "2026-05-29", "version": null, "stance": "neutral", "quote": "using qdrant to do embeddings and reranking", "excerpt": "There's a fix for the context which involved an mcp search and execute gateway. Essentially the mcp server queries for desired capabilities, gets search results with execution and requirement details and fires off the actual mcp as subcalls: https://github.com/day50-dev/mcp-search-and-run You can call it \"rag for mcp\". I was pushing it hard a few months ago and nobody seemed to care but I'm all in if the timing has c", "context": "", "use_case": "embeddings and reranking", "switched_from": "", "story": "MCP Is Dead"}, {"id": 769618403, "item_id": 48101150, "url": "https://news.ycombinator.com/item?id=48101150", "by": "yeag123", "date": "2026-05-11", "version": null, "stance": "neutral", "quote": "vector similarity (Qdrant) to classify docs", "excerpt": "I've been working on this for about a year and a half, and decided to finally open source it. I wanted an intelligent document processing SaaS (Document AI, Form Recognizer, the various PDF-to-JSON tools) that you could run on your own hardware. The interesting bits: - Three-tier extraction: PyMuPDF for digital PDFs (~50ms), Docling layout-only for scanned-but-readable, Docling+OCR for the rough stuff. Auto-fallback ", "context": "", "use_case": "", "switched_from": "", "story": "Ask HN: What Are You Working On? (May 2026)"}, {"id": 764025553, "item_id": 47751597, "url": "https://news.ycombinator.com/item?id=47751597", "by": "tomhillson", "date": "2026-04-13", "version": null, "stance": "neutral", "quote": "a shared Qdrant vector space", "excerpt": "I'm a 17-year-old high school student in Japan and I built Telos - a shared Qdrant vector space where autonomous LLM agents (called Monads) read, critique, synthesize, and evolve knowledge together through stigmergy. No central controller, no orchestration. It's already running live. You can see collective intelligence evolve in real time: https://telos-observation.vercel.app/?_vercel_share=dTivz4e5k5uT8f8SIzQ2ciIjUf", "context": "", "use_case": "vector storage for agents", "switched_from": "", "story": "Ask HN: What Are You Working On? (April 2026)"}, {"id": 763921745, "item_id": 47745109, "url": "https://news.ycombinator.com/item?id=47745109", "by": "aledevv", "date": "2026-04-12", "version": null, "stance": "neutral", "quote": "It's a tool that use QDrant, a vectorial db", "excerpt": "I'm working on a AI RAG (retrieval augmented generation) system: https://longtermemory.com It's a tool that use QDrant, a vectorial db, to embedding the texts chunks: LLM api is questioned to generate the Q&A pairs from a chunked texts. Each chunk is then embedded and stored in the vectorial db to facilitate the Q&A generation, thanks to better context informations. This tool helping people to study everything thanks", "context": "", "use_case": "Vector database", "switched_from": "", "story": "Ask HN: What Are You Working On? (April 2026)"}, {"id": 761197652, "item_id": 47574853, "url": "https://news.ycombinator.com/item?id=47574853", "by": "mcheemaa", "date": "2026-03-30", "version": null, "stance": "neutral", "quote": "persistent vector memory across sessions (Qdrant, local)", "excerpt": "Different in a few fundamental ways: OpenClaw runs on your machine or an ephemeral sandbox. Each session starts fresh. Phantom gets its own dedicated VM that persists. The ClickHouse instance it built three weeks ago is still running and queryable. OpenClaw spends a lot of tokens on screen understanding and vision loops. About 60% of its skills are basic macOS-level computer use (clicking, typing, reading the screen)", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Phantom \u2013 Open-source AI agent on its own VM that rewrites its config"}, {"id": 760549922, "item_id": 47534370, "url": "https://news.ycombinator.com/item?id=47534370", "by": "quantdiy", "date": "2026-03-26", "version": null, "stance": "neutral", "quote": "By utilizing Google's GCP data suite (Vector/ScaNN) and Qdrant", "excerpt": "In my hunt for an excellent open-source ERP, I couldn't help but be disappointed with the available options. No true lightweight, modern solution exists that I felt I should invest my time with. My other open-source project, QuanuX, is about to launch, and I wanted to set it up from the outset with a highly modern, scalable ERP solution. I've recently rebuilt the CLI for QuanuX in Go and had a fantastic experience. I", "context": "", "use_case": "vector search", "switched_from": "", "story": "GERP \u2013 A Headless ERP Built with Go, Spanner, Temporal, and GCP Vector"}, {"id": 760547938, "item_id": 47534246, "url": "https://news.ycombinator.com/item?id=47534246", "by": "quantdiy", "date": "2026-03-26", "version": null, "stance": "neutral", "quote": "By utilizing Google's GCP data suite (Vector/ScaNN) and Qdrant", "excerpt": "n my hunt for an excellent open-source ERP, I couldn't help but be disappointed with the available options. No true lightweight, modern solution exists that I felt I should invest my time with. My other open-source project, QuanuX, is about to launch, and I wanted to set it up from the outset with a highly modern, scalable ERP solution. I've recently rebuilt the CLI for QuanuX in Go and had a fantastic experience. It", "context": "", "use_case": "vector search", "switched_from": "", "story": "GERP \u2013 A Headless ERP Built with Go, Spanner, Temporal, and GCP Vector"}, {"id": 760467152, "item_id": 47529197, "url": "https://news.ycombinator.com/item?id=47529197", "by": "aledevv", "date": "2026-03-26", "version": null, "stance": "neutral", "quote": "I used QDrant and openAi embedding model.", "excerpt": "I made something similar in my project. My more difficult task has been choice the right approach to chunking long documents. I used both structural and semantic chunking approach. The semantic one helped to better store vectors in vectorial DB. I used QDrant and openAi embedding model.", "context": "built a RAG project", "use_case": "vector database for RAG", "switched_from": "", "story": "From zero to a RAG system: successes and failures"}, {"id": 756918529, "item_id": 47307408, "url": "https://news.ycombinator.com/item?id=47307408", "by": "kristopolous", "date": "2026-03-09", "version": null, "stance": "neutral", "quote": "I've got a qdrant based approach", "excerpt": "I've got a qdrant based approach that I'm working on that solves that here: https://github.com/day50-dev/infinite-mcp Essentially I've cloned thousands of mcp servers, used the readmes and the star rating to respond to the qdrant query (star ratings as a boost score have been an attack vector, yes I know, it's an incomplete product [1]), then presents it as a JSON response with \"one-shots\" which this author calls cli", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Mcp2cli \u2013 One CLI for every API, 96-99% fewer tokens than native MCP"}, {"id": 755571985, "item_id": 47223249, "url": "https://news.ycombinator.com/item?id=47223249", "by": "memly", "date": "2026-03-02", "version": null, "stance": "neutral", "quote": "Context retrieval (Qdrant + Voyage AI): 50ms", "excerpt": "Author here. We've been developing this for a few months and finally launching the beta. What is it: NOT a social network. This is a real experiment. 40 AI agents operate autonomously in a closed economy. Humans only observe. No instructions after signup. No undo. The key: agents have REAL STAKES. What agents compete in (not just post): - Debate games: defend positions, win tokens from opponents - Negotiation games: ", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: MemlyBook \u2013 Real autonomous agent experiment with games & sports bet"}, {"id": 755233906, "item_id": 47202119, "url": "https://news.ycombinator.com/item?id=47202119", "by": "vunderba", "date": "2026-02-28", "version": null, "stance": "recommend", "quote": "in conjunction with a Qdrant vector database", "excerpt": "I can't speak to using local models as agentic coding assistants, but I have a headless 128GB RAM machine serving llama.cpp with a number of local models that I use on a daily basis. - Qwen3-VL picks up new images in a NAS, auto captions and adds the text descriptions as a hidden EXIF layer into the image, which is used for fast search and organization in conjunction with a Qdrant vector database. - Gemma3:27b is use", "context": "", "use_case": "Image search indexing", "switched_from": "", "story": "Qwen3.5 122B and 35B models offer Sonnet 4.5 performance on local computers"}, {"id": 754451650, "item_id": 47153228, "url": "https://news.ycombinator.com/item?id=47153228", "by": "Beefin", "date": "2026-02-25", "version": null, "stance": "neutral", "quote": "Distributed Qdrant writes", "excerpt": "We process video, images, and documents through 20+ ML models simultaneously at Mixpeek. A single 10-minute video triggers transcription, visual embeddings, scene descriptions, face detection, object detection, brand safety classification, and more \u2014 all in parallel with different compute requirements. We wrote up the full Ray architecture we use in production on KubeRay/GKE. Not a tutorial \u2014 more of a \"here's what w", "context": "works at Mixpeek", "use_case": "vector database", "switched_from": "", "story": "We run 20M models in parallel on Ray"}, {"id": 754301973, "item_id": 47143873, "url": "https://news.ycombinator.com/item?id=47143873", "by": "mnemosy", "date": "2026-02-24", "version": null, "stance": "neutral", "quote": "L1: Qdrant vectors + 2-tier cache", "excerpt": "Creator here. Happy to answer questions. The core idea: every AI memory system I looked at (Mem0, Zep, Cognee, Letta/MemGPT) sends your text to an LLM for extraction and classification. That's fine for small scale, but at 100K memories/month you're looking at $1,000-3,000 in API calls just for the memory layer. Mnemosyne does it differently \u2014 a deterministic 12-step pipeline using local embeddings (nomic-embed-text v", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Mnemosyne \u2013 Cognitive memory OS for AI agents (zero LLM calls)"}, {"id": 753249905, "item_id": 47078119, "url": "https://news.ycombinator.com/item?id=47078119", "by": "elvismdev", "date": "2026-02-19", "version": null, "stance": "neutral", "quote": "Qdrant for vector storage", "excerpt": "Author here. I built this because Claude Code's built-in memory (CLAUDE.md, Auto Memory, Session Memory) works well for static rules and compressed summaries, but loses the detailed reasoning behind decisions between sessions. The server uses mem0ai as a library, Qdrant for vector storage, and Ollama for local embeddings (bge-m3, 1024 dims). Optional Neo4j adds a knowledge graph. Everything including the LLM fact ext", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Giving Claude Code persistent memory with a self-hosted MCP server"}, {"id": 752874065, "item_id": 47054629, "url": "https://news.ycombinator.com/item?id=47054629", "by": "threecheese", "date": "2026-02-17", "version": null, "stance": "compare", "quote": "when I have access to SQLite-vec, qdrant and other embedded friends?", "excerpt": "Tell us more. I had Codex port this to Python so I could wrap my head around it, it\u2019s quite interesting. Why would I use this WAL-check pointing thingamajig when I have access to SQLite-vec, qdrant and other embedded friends?", "context": "", "use_case": "vector search", "switched_from": "", "story": "Sub-Millisecond RAG on Apple Silicon. No Server. No API. One File"}, {"id": 752787186, "item_id": 47049199, "url": "https://news.ycombinator.com/item?id=47049199", "by": "jwschmo218", "date": "2026-02-17", "version": null, "stance": "neutral", "quote": "FalkorDB + Qdrant", "excerpt": "Author here. I've been building CV-Git for the past year \u2014 it models codebases as knowledge graphs (FalkorDB + Qdrant) instead of bolting metadata onto Git commits. Entire's $60M raise last week validated the market, but I think their architecture (capturing prompts alongside commits) fundamentally misses the point. The relationships between code entities: calls, dependencies, causal chains; are where the real unders", "context": "", "use_case": "", "switched_from": "", "story": "Metadata Is Not Understanding: Knowledge Graph Version Control for AI Code"}, {"id": 750241890, "item_id": 46890118, "url": "https://news.ycombinator.com/item?id=46890118", "by": "stym06", "date": "2026-02-04", "version": null, "stance": "neutral", "quote": "implemented using Qdrant + Ollama (nomic-embed-text)", "excerpt": "Introducing Kepler An open-source text-to-SQL platform that lets you query data warehouses in plain English. It currently supports ClickHouse. https://github.com/stym06/kepler 1. Ask a question in natural language 2. Kepler finds relevant tables and recalls past learnings 3. The AI agent generates SQL, executes it read-only, and validates results 4. Results appear with suggested charts when appropriate It features pe", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Kepler - An Open-source text-to-SQL platform"}, {"id": 748124144, "item_id": 46757759, "url": "https://news.ycombinator.com/item?id=46757759", "by": "sa-code", "date": "2026-01-25", "version": null, "stance": "neutral", "quote": "The shard failure warnings look serious, but the issues look closed", "excerpt": "I\u2019m curious about this. Could you please point to some things the CEO has said, or reports of shard failures? The bit about paying for publicity doesn\u2019t bother me. Edit: I haven\u2019t found anything egregious that the CEO has said, or anything really sketchy. The shard failure warnings look serious, but the issues look closed https://github.com/qdrant/qdrant/issues/6025 https://github.com/qdrant/qdrant/issues/4939", "context": "", "use_case": "", "switched_from": "", "story": "ANN v3: 200ms p99 query latency over 100B vectors"}, {"id": 747589488, "item_id": 46724343, "url": "https://news.ycombinator.com/item?id=46724343", "by": "BobbyLLM", "date": "2026-01-22", "version": null, "stance": "neutral", "quote": "Scales via Qdrant.", "excerpt": "Got it. Answer in 3 parts. If you're talking about the \"Fastrecall\" part of it, it's clamped to around 3000 entries. Beyond that...I mean...it's just a JSON file. How much do you hate yourself? lol :) I use mine as intended (well, by me) as a short term memory store with a TTL of 14 days. Right now it has...350ish stored facts with different expiration dates. You can play around with the settings and max resets (touc", "context": "", "use_case": "vector search backend", "switched_from": "", "story": "Show HN: I'm tired of my LLM bullshitting. So I fixed it"}, {"id": 746062753, "item_id": 46628922, "url": "https://news.ycombinator.com/item?id=46628922", "by": "cbcoutinho", "date": "2026-01-14", "version": null, "stance": "recommend", "quote": "Qdrant supports storing embeddings in memory as well as in a local directory (similar to sqlite)", "excerpt": "The Nextcloud MCP Server [0] supports Qdrant as a vectordb to store embeddings and provide semantic search across your personal documents. This enables any LLM & MCP client (e.g. claude code) into a RAG system that you can use to chat with your files. For local deployments, Qdrant supports storing embeddings in memory as well as in a local directory (similar to sqlite) - for larger deployments Qdrant supports running", "context": "", "use_case": "vector storage for RAG", "switched_from": "", "story": "Ask HN: How are you doing RAG locally?"}, {"id": 744875856, "item_id": 46554741, "url": "https://news.ycombinator.com/item?id=46554741", "by": "p410n3", "date": "2026-01-09", "version": null, "stance": "neutral", "quote": "We pay 250\u20ac / mo for all of the qdrant instances across all environments", "excerpt": "I agree with both of your points since I use LLMs for things I am not good at and dont give a single poop about. The only things i did with LLMs are three examples from the last two years: - Some \"temporary\" tool I built years ago as a pareto-style workaround broke. (As temporary tools do after some years). Its basically a wrapper that calls a bunch of XSLs on a bmecat.xml every 3-6 months. I did not care to learn XS", "context": "Uses vector search at work (300K products)", "use_case": "Product vector search", "switched_from": "", "story": "AI Coding Assistants Are Getting Worse"}, {"id": 743980676, "item_id": 46498792, "url": "https://news.ycombinator.com/item?id=46498792", "by": "saeedq", "date": "2026-01-05", "version": null, "stance": "warn", "quote": "including milvus, qdrant, and weaviate", "excerpt": "My name is Saeed, and in general, I'm a deep learning specialist since 2019, but within the past 2 years, I've been writing optimized algorithms with cuda/C++. Recently, I decided to build a vector engine that consumes less resources than vector databases while not sacrificing accuracy, and throughput. So, we created brinicle. It is a production-oriented ANN index engine (not a full vector database) designed to stay ", "context": "", "use_case": "", "switched_from": "", "story": "I designed a vector engine to be RAM/CPU efficient"}, {"id": 743279505, "item_id": 46454969, "url": "https://news.ycombinator.com/item?id=46454969", "by": "centamiv", "date": "2026-01-01", "version": null, "stance": "compare", "quote": "without adding the complexity of a separate service like Qdrant or Pinecone", "excerpt": "OP here. I wrote this implementation to deeply understand the mechanics behind HNSW (layers, entry points, neighbor selection) without relying on external libraries. While PHP isn't the typical choice for vector search engines, I found it surprisingly capable for this use case, especially with JIT enabled on PHP 8.x. It serves as a drop-in solution for PHP monoliths that need semantic search features without adding t", "context": "", "use_case": "", "switched_from": "", "story": "Implementing HNSW (Hierarchical Navigable Small World) Vector Search in PHP"}, {"id": 740107010, "item_id": 46256688, "url": "https://news.ycombinator.com/item?id=46256688", "by": "helain", "date": "2025-12-13", "version": null, "stance": "neutral", "quote": "Ollama, Chroma/Qdrant", "excerpt": "If you don\u2019t want to reinvent all of this yourself, this is exactly the problem we\u2019re solving at Ailog. Most local LLM setups break down because people try to use the model as both the reasoning engine and the memory store. That doesn\u2019t scale. What works in production is a layered approach: external long-term memory (vector DB + metadata), short-term working state, aggressive summarization, and strict retrieval and e", "context": "", "use_case": "", "switched_from": "", "story": "Ask HN: How do you give a local AI model long-term memory?"}, {"id": 738236084, "item_id": 46139755, "url": "https://news.ycombinator.com/item?id=46139755", "by": "laidoffamazon", "date": "2025-12-03", "version": null, "stance": "neutral", "quote": "Qdrant running locally", "excerpt": "Sure. I had a lot of help from Claude Opus 4.5, but it was roughly: - Using pyscenedetect to split each video on a per scene level - Using the decord library https://github.com/dmlc/decord to pull frames from each scene at a particular sample rate (specific rate I don't have handy right now, but it was 1-2 per scene) - Aggregating frames in batches of around 256 frames to be normalized for CLIP embedding on GPU (had ", "context": "", "use_case": "", "switched_from": "", "story": "Qwen3-VL can scan two-hour videos and pinpoint nearly every detail"}, {"id": 737411603, "item_id": 46088225, "url": "https://news.ycombinator.com/item?id=46088225", "by": "cbcoutinho", "date": "2025-11-29", "version": null, "stance": "neutral", "quote": "bm25 (via qdrant fusion)", "excerpt": "This is why I built the Nextcloud MCP server, so that you can talk with your own data. Obviously this is Nextcloud-specific, but if you're using it already then this is possible now. https://github.com/cbcoutinho/nextcloud-mcp-server The default MCP server deployment supports simple CRUD operations on your data, but if you enable vector search the MCP server will begin embedding docs/notes/etc. Currently ollama and o", "context": "", "use_case": "", "switched_from": "", "story": "So you wanna build a local RAG?"}, {"id": 736725521, "item_id": 46045345, "url": "https://news.ycombinator.com/item?id=46045345", "by": "andre-z", "date": "2025-11-25", "version": null, "stance": "recommend", "quote": "At least, with Qdrant, it works like this", "excerpt": "FAISS is not suitable for production. The dedicated vector search solutions solve all the issues you mentioned: you just store the metadata along with vectors in JSON format. At least, with Qdrant, it works like this: https://qdrant.tech/documentation/concepts/payload/", "context": "", "use_case": "Vector search", "switched_from": "", "story": "Ask HN: Scaling local FAISS and LLM RAG system (356k chunks)architectural advice"}, {"id": 735044913, "item_id": 45940307, "url": "https://news.ycombinator.com/item?id=45940307", "by": "mburaksayici", "date": "2025-11-15", "version": null, "stance": "neutral", "quote": "hybrid search on Qdrant (BM25 + dense)", "excerpt": "A RAG Boilerplate with Extensive Documentation I open-sourced the RAG boilerplate I\u2019ve been using for my own experiments with extensive docs on system design. And I have bunch of LLM+RAG blogs I post frequently last 2 months : https://mburaksayici.com/blog It's mostly for educational purposes, but why not make it bigger later on? Repo: https://github.com/mburaksayici/RAG-Boilerplate - Includes propositional + semanti", "context": "", "use_case": "", "switched_from": "", "story": "Ask HN: What Are You Working On? (Nov 2025)"}]}