- 플랫폼
- YouTube
- 길이
- 4:11
- 사용 언어
- 영어(미국)
- 조회수
- 3.2천
- 컷 간격 중앙값
- 6.87s
- 얼굴 노출
- 90%
- 어두운 프레임
- 12%
Getting Started with Supabase Vector Databases 자막 전문
브랜드가 작성한 것이 아니라, 영상의 오디오를 음성 모델이 받아쓴 것입니다. 아무 줄이나 클릭하면 그 순간으로 이동합니다.
- 0:00A lot of AI apps have the same hidden challenge.
- 0:02They need the right context before they can give a useful answer.
- 0:06That context might live in the docs, support tickets, product data, messages, images, or internal knowledge base.
- 0:12Vector databases make that data searchable by meaning.
- 0:15In this video, we'll cover what embeddings are, what vector databases do, and why Superbase is a natural place to use them.
- 0:25Before vector databases make sense, we need to understand embeddings.
- 0:29An embedding is a numerical representation of meaning.
- 0:32You give content to an embedding model and it returns a list of numbers.
- 0:37Those numbers are like coordinates in a semantic space.
- 0:40Similar ideas end up together, different ideas end up farther apart.
- 0:44So, how do I reset my password and I forgot my login may be close even though the wording is different.
- 0:51Embeddings do not match text character by character.
- 0:54They capture patterns, words, phrases, concepts, and context.
- 0:58Embeddings are vectors.
- 1:00A vector database is what makes those vectors usable in an application.
- 1:04It stores each vector alongside the original record and its metadata, such as the document,
- 1:09product, workspace, category, or permissions it belongs to.
- 1:12When a query vector comes in, the database compares it with the stored vectors, ranks
- 1:17the closest records, and returns the best matches.
- 1:21As the dataset grows, specialized vector indexes help it find nearby vectors quickly instead
- 1:26instead of comparing every vector one by one.
- 1:28And because the vectors are stored alongside the normal database records, similarity search
- 1:33can still use filters, access rules, and the rest of your application data.
- 1:38The embedding model creates the vectors, the vector databases store, indexes, filters,
- 1:43and retrieves them.
- 1:44Now how does similarity search work?
- 1:47The flow is simple.
- 1:49First, take the content you want to search, docs, support articles, products, notes, whatever.
- 1:54Generate an embedding for each of them and store it in your database.
- 1:58Then when a user performs a search, generate an embedding for their query data.
- 2:03Now compare the query embeddings against the stored embeddings.
- 2:06The closest vectors are the closest matches.
- 2:10Superbase is built on Postgres and Postgres supports Vector Search with PGVector.
- 2:15So with Superbase, Vector Search is supported natively.
- 2:19You can store embeddings, query them with SQL and keep using the database patterns you
- 2:24you already know.
- 2:25You also get the rest of Superbase platform around it,
- 2:28migrations, client libraries, edge functions, auth, storage,
- 2:31dashboard, and many more.
- 2:33What kind of apps can you actually
- 2:34build with vector databases?
- 2:36The obvious use case is semantic search.
- 2:39Let users search docs, help centers, products, categories,
- 2:43or app content with natural language.
- 2:46Another big one is RAG, Retrieval Augmented Generation.
- 2:50That's when an AI assistant retrieve
- 2:52relevant content from your database before generating an answer. You can also
- 2:56build recommendations, image search, duplicate detection, and matching systems.
- 3:01In real apps, search is rarely just find similar things. It's usually more specific,
- 3:06find similar things that the users can access in this workspace, in this category,
- 3:11or in the last 30 days. That's why having vectors in Postgres is useful. Your
- 3:16semantic search can work with permissions, filters, metadata, ownership, and
- 3:21business logic. Vectors are not replacement for every search problem. Use exact phrases when
- 3:27exactness matters. IDs, emails, usernames, codes, precise phrases. Use vector search when meaning
- 3:34matters. Questions, descriptions, support issues, docs, recommendations. And often the best answer
- 3:40is both. Keyword search gives you precision. Vector search gives you semantic understanding.
- 3:45Hybrid search combines them. And that's the core idea. Embeddings turn content into numerical
- 3:50meaning a vector database stores those embeddings and searches by similarity.
- 3:55And Superbase lets you do that right inside Postgres, next to the rest of your
- 3:59app data. Once you understand that, vector databases feel much less mysterious.
- 4:04They're a way to help your app understand what your data means, not just
- 4:08what words it contains.
62줄 · 650단어
More from Supabase
All of Supabase →The rest of their work we have taken apart.
Cut at the same speed
Films whose median shot runs closest to 6.87s.
144K viewsheyclicky
never touch me again
0:5525 frames6.9s cut4% face
3K viewsWMAR-2 News
Amazon Dating launches as a pre-Valentine's Day joke
0:5324 frames6.8s cut79% face
8.2M viewsVodafone Australia
Switch to Vodafone this summer
0:158 frames6.8s cut75% face
133K viewsLuma
Stop guessing. Start directing. Ray3 Modify is now in Dream Machine. ...
0:4421 frames6.8s cut29% face
6K viewsRaycast
Settings in Raycast
2:35108 frames6.9s cut12% face
907K viewsAnthropic
We’re building tools to support research in the life sciences, from e...
1:2029 frames6.8s cut0% face
48K viewsCursor
Cloud agents and automations now support multi-repo environments. Co...
0:4419 frames6.8s cut0% face
21K views1X
Logs | Lowrider
1:4549 frames6.8s cut22% face
134K viewsRay-Ban | Meta
How to Set Up Your Ray-Ban Meta Smart Glasses | FULL Step-by-Step Tutorial
10:23166 frames7.0s cut35% face
226K viewsAirtime
Meet Airtime Creator | How to Create Professional Video Presentations in 2025
2:19108 frames7.0s cut100% face
44 viewsFireCut
AI YouTube Chapters: How to Automate Timestamps & Animated Intros
1:2130 frames7.0s cut100% face
791K viewsPeloton
Such a great App.
1:0330 frames6.8s cut37% face
Lit and framed like this
Closest on how much of the running time has a face in it and how dark the frames are.
Launch43K viewsD-ID AI Video Platform
Introducing chat.D-ID! Giving a face to ChatGPT
0:4718 frames89% face
68K viewsSubah Wadhwani
We went so deep on AEO, we should lowkey charge for people to view th...
60:18483 frames16.0s cut87% face
163K viewsAdil Mania
september was wild.
0:5951 frames1.9s cut88% face
61 viewsHeadshotPro
Free Profile Picture Generator
0:178 frames88% face
Ali Abdaal
7 Habits that Save Me 3+ Hours a Day
11:27171 frames4.5s cut92% face
290K viewsLinear
After the IDE (2026) A short film about @Coinbase and the work of bui...
6:00451 frames5.1s cut89% face
88K viewsSketch Together
hilos: a place where people and AI agents ship projects
0:5241 frames2.2s cut93% face
1K viewsDescript
Just Hit Record. Descript Does the Rest.
1:1759 frames1.4s cut90% face
124K viewsProductive Dude
Claude Design FULL Tutorial: Beginner to Pro
37:11285 frames9.9s cut91% face
108K viewsRunway
Learn how to create your own Runway Character via the Runway API. And...
4:34421 frames2.3s cut86% face
44K viewsNotion
Talent = Capability x Taste x Agency @ivanhzhao & @tim at @LDNTe...
0:4421 frames5.0s cut86% face
103K viewsSilicon Mania
last week in tech was coa-chella.
0:3567 frames0.6s cut87% face
More social
Same category, other makers, most watched first.
622M viewsBase44
Your ideas, your apps—Base44
0:5651 frames1.1s cut6% face
538M viewsBase44
Turn any idea into an app with Base44
0:5538 frames2.2s cut50% face
154M viewsBase44
Build Your Website with One Sentence | Base44
0:4835 frames0.9s cut31% face
81M viewsNike Football
Rip The Script | Nike Football
6:00310 frames1.1s cut60% face
75M viewsPetSmart
PetSmart | "Pickup in...STORE!"
0:1511 frames1.4s cut64% face
69M views1X
NEO The Home Robot Order Today
9:53271 frames2.8s cut25% face
69M viewsApple
Apple Education: Ready for every learning opportunity
1:4687 frames1.3s cut40% face
64M viewsBland
Phone calls suck. So, we bought a giant telephone car, raised $65,00...
1:4854 frames4.3s cut81% face
39M viewsFigma
Oh wait there’s more: we’re building 3D Transforms, here’s a preview ...
1:0423 frames4.0s cut78% face
36M viewshims
Getting Help is Simple, With Hims
0:3023 frames2.0s cut70% face
36M viewsFigma
Made in Figma (2025) Like this ❤️ we designed it for you
0:3872 frames0.5s cut7% face
35M viewsApple
Hello, MacBook Neo
3:50291 frames67.5s cut1% face