[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"guide-what-is-a-vector-database::en":3,"guide-related-what-is-a-vector-database::en":18},{"slug":4,"title":5,"excerpt":6,"body":7,"meta_title":8,"meta_description":9,"keywords":10,"category":16,"published_at":17,"updated_at":17},"what-is-a-vector-database","What Is a Vector Database? And When You Need One","A vector database stores embeddings and finds the closest matches to a query. This guide covers what that enables, what it costs, and why most teams need one later than they think.","\u003Ch2>What it stores\u003C\u002Fh2>\n\u003Cp>A vector database stores embeddings — lists of numbers that represent the meaning of a piece of text, an image, or audio — and answers one question quickly: which stored items are closest to this one? Closeness in that numeric space corresponds to similarity in meaning, which is why a query about \"cancelling my plan\" can retrieve a document titled \"ending your subscription\" without sharing a single keyword. That is the entire value proposition, and everything else is engineering around it.\u003C\u002Fp>\n\u003Ch2>Why ordinary search is not enough\u003C\u002Fh2>\n\u003Cp>Keyword search matches strings. It is excellent when the user types the words your documents use, and it fails when they do not — which is most of the time in support, documentation, and internal knowledge. Semantic search closes that gap. It also introduces a new failure: it will always return the nearest items even when nothing relevant exists, so a query about something absent from your corpus returns whatever is least unrelated, with a similarity score that looks reassuring. Handling that empty case explicitly is the difference between a search that says \"I don't know\" and one that confidently retrieves the wrong document.\u003C\u002Fp>\n\u003Ch2>How the search works\u003C\u002Fh2>\n\u003Cp>Comparing a query against every stored vector is exact and too slow past a few hundred thousand items, so these systems use approximate nearest-neighbour indexes — most commonly HNSW, a layered graph you descend to reach a near-best answer in a fraction of the time. The word approximate is not decorative: you are trading a small amount of recall for a large amount of speed, and the index parameters control that trade. Compression techniques shrink the vectors further, because ten million full-precision embeddings occupy tens of gigabytes of memory.\u003C\u002Fp>\n\u003Ch2>You may not need a dedicated one\u003C\u002Fh2>\n\u003Cp>This is the part vendors skip. Postgres with the pgvector extension handles hundreds of thousands of vectors comfortably, keeps your embeddings in the database you already operate and back up, and lets you filter by tenant, date, or permission in the same query. A dedicated vector database earns its place at genuinely large scale, with very high query volume, or when you need features it specialises in. Starting with a separate system means running, securing, and syncing two stores before you know whether you needed to.\u003C\u002Fp>\n\u003Ch2>The parts that actually determine quality\u003C\u002Fh2>\n\u003Cp>Teams tend to attribute poor retrieval to the database and then switch databases, which almost never helps. Retrieval quality is set upstream: how you split documents into chunks, which embedding model you use, and whether you re-rank the top results with a more expensive model before using them. Chunking is the most consequential and the least discussed — split too small and passages lose the context that made them meaningful, too large and the embedding averages several topics into a vector that matches nothing precisely.\u003C\u002Fp>\n\u003Ch2>The operational trap\u003C\u002Fh2>\n\u003Cp>Embeddings from different models are not comparable. Change your embedding model and every stored vector becomes meaningless against new queries, which means a full re-embedding of the entire corpus — a migration that is easy to underestimate and impossible to do halfway. Plan for it: record which model produced each vector, and treat a model upgrade as a data migration rather than a config change.\u003C\u002Fp>\n\u003Ch2>The short version\u003C\u002Fh2>\n\u003Cp>A vector database is a specialised index, not a source of truth. Keep your documents where they live, treat the vectors as a derived artefact you can rebuild, and start with the database you already run. Reach for a dedicated system when you have measured a reason to, not when you start.\u003C\u002Fp>","What Is a Vector Database?","What a vector database is: storage for embeddings with fast similarity search. What it enables, what it costs, and why Postgres is often enough to start.",[11,12,13,14,15],"vector database","embeddings","semantic search","rag","similarity search","fundamentals","2026-08-07T03:45:02+00:00",[19,24,28,32,37,42],{"slug":20,"title":21,"excerpt":22,"updated_at":23},"ai-tool-pricing-models-seat-vs-usage-vs-credits","AI Tool Pricing Models: Seat-Based vs Usage-Based vs Credits","The three common ways AI tools charge — per seat, per usage, and by credits — and how to reason about which one will actually be cheaper for the way your team works.","2026-08-05T14:32:26+00:00",{"slug":25,"title":26,"excerpt":27,"updated_at":23},"how-ai-image-generators-differ-diffusion-vs-the-rest","How AI Image Generators Differ: Diffusion vs the Rest, in Plain Terms","A non-technical explanation of how AI image generators work, why the diffusion approach became dominant, and what practical differences to expect between tools.",{"slug":29,"title":30,"excerpt":31,"updated_at":23},"how-to-automate-your-workflow-without-code","How to Automate Your Workflow Without Code","A practical sequence for building automations that survive: picking the right process, mapping it before touching a tool, and handling the failure cases that break most first attempts.",{"slug":33,"title":34,"excerpt":35,"updated_at":36},"how-to-build-a-chatbot-without-coding","How to Build a Chatbot Without Coding","A practical route to a working chatbot using no-code tools: deciding scope, connecting your own content, handling the questions it cannot answer, and knowing what it will cost.","2026-08-05T14:32:27+00:00",{"slug":38,"title":39,"excerpt":40,"updated_at":41},"how-to-change-a-prompt-without-breaking-production","How to Change a Prompt Without Breaking Production","Prompts get edited in a text box and shipped in seconds, which is why they break things quietly: no compiler, no stack trace, no obvious moment of failure. Give them the release discipline code gets.","2026-08-24T03:30:02+00:00",{"slug":43,"title":44,"excerpt":45,"updated_at":23},"how-to-choose-an-ai-writing-assistant","How to Choose an AI Writing Assistant","A practical framework for picking an AI writing tool — matching it to the kind of writing you actually do, checking editing controls, and avoiding tools that produce confident but generic copy."]