NLP Academy · 课时

使用向量存储进行向量搜索

检索最相关的段落

使用向量存储进行向量搜索 是 CoddyKit 上的免费 NLP Academy 课时。这是第 3 节课，共 4 节。你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。这是 NLP Academy 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 NLP Academy 课程共包含 4 节课。

本课时的部分内容尚未翻译，以英文显示。

What a Vector Store Does

A vector store holds your chunk embeddings and finds the closest ones to a query fast, even across millions of vectors.

Searching by Similarity

Instead of matching keywords, you match meaning. The store returns chunks whose vectors sit nearest to the query vector.

Cosine Similarity

The common distance measure is cosine similarity. It compares the angle between two vectors, so closer in meaning scores higher.

Embed the Query the Same Way

The question must use the same model as your chunks. Mismatched embeddings live in different spaces and cannot be compared.

query_vec = model.encode(["How do I reset my password?"])

Top-K Retrieval

You rarely want every match. Ask the store for the top-k nearest chunks, where k is a small number like 3 or 5.

A Simple In-Memory Store

For small data, FAISS keeps vectors in memory and searches in milliseconds. You build an index, then add your vectors.

import faiss
index = faiss.IndexFlatIP(384)
index.add(vectors)

Running a Search

Call search with the query vector and k. It returns the scores and the indexes of the best-matching chunks.

scores, ids = index.search(query_vec, k=3)

Mapping Results Back to Text

The store gives you indexes, not text. Use them to look up the original chunks you stored alongside the vectors.

hits = [chunks[i] for i in ids[0]]

Persistent Vector Databases

For real apps, a vector database like Chroma or Pinecone stores vectors on disk, scales out, and survives restarts.

Filtering With Metadata

Attach tags such as author or date to each vector. Then you can filter the search, like only chunks from one document.

Why Approximate Search Helps

Exact search gets slow at scale. Approximate nearest-neighbor indexes trade a tiny bit of accuracy for huge speed gains.

Quick Check

Recall how a vector store finds relevant chunks.

Recap

A vector store embeds the query, finds the top-k nearest chunks by similarity, and maps results back to text. ✅

常见问题解答

「使用向量存储进行向量搜索」课时是免费的吗？

是的 — 「使用向量存储进行向量搜索」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 NLP Academy 课程的其余内容，请升级到 CoddyKit PRO。 NLP Academy 课程共包含 4 节课。

「使用向量存储进行向量搜索」这节课中我会学到什么？

检索最相关的段落你通过在浏览器中直接运行的动手代码来练习 NLP Academy，全天候 AI 导师会在你学习这节课的过程中回答你的问题。

学习 NLP Academy 需要有经验吗？

无需任何先前经验。CoddyKit 上的 NLP Academy 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。这是第 3 节课，共 4 节。

「使用向量存储进行向量搜索」课时需要多长时间？

大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。

我能在这节 NLP Academy 课中编写并运行代码吗？

能。每节 NLP Academy 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。

此课程中的所有课时

← 返回 NLP Academy