Zvec Agent Skills
The Zvec skill guides an agent through the embedded vector database: collection schema, hybrid search, RAG and multimodal queries
- 21 ★ on the repository
- Last commit
- Apache-2.0
Minimal access
Why this level
- The skill is made of reference files and code samples and runs nothing itself
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/zvec-agent-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill is made of reference files and code samples and runs nothing itself
This is third-party code. Review the repository files before installing.
What it does
The `zvec` skill is built from the documentation of the Zvec vector database and targets Python 3.10 or newer and Node.js 16 or newer. Zvec runs inside the application process on Alibaba's Proxima engine and needs no separate server. The agent gets references for creating a collection, working with documents, searching by one or several vectors, hybrid search with filters, building RAG and searching by image together with text. Every topic has separate examples in Python and in TypeScript. Shared files describe the data model, hold an API cheatsheet and cover common problems. Name the language and the scenario, for example a failure log of 100 thousand records, and the agent picks a schema, an index and a ranking method to match.
Who it is for. Python and Node.js developers who build document search or recommendations on the embedded Zvec database and want an agent that knows its API.
Good fit when
- A support knowledge base in Markdown gets a RAG pipeline with semantic search
- An online shop searches products by the meaning of the description and trims results by price, category and brand
- Search runs over an image and text at once, and `WeightedReRanker` blends the scores with set weights
- Failure logs are stored as vectors to find similar cases
- A Zvec query fails with an error, and the agent works through it with `troubleshooting.md`
Not a fit when
- You work with another vector database such as Qdrant or Pinecone: the skill knows Zvec only
- You need a distributed cluster with a separate server: Zvec runs inside the application process
- The project is neither Python nor Node.js: examples exist for these two only
Example request
Build a Python search over a Markdown knowledge base with Zvec: a collection schema with a 768-dimension vector, loading and a query with a category filterLimitations
- The skill holds documentation and examples, while the Zvec package installs separately: `pip install zvec` for Python and `npm install @zvec/zvec` for Node.js.
- Examples are in Python and TypeScript, with no other languages.
- Tell the agent the language and the scenario up front, or the advice stays generic.
- The last commit in the repository dates from April 2026, and recent API changes may be missing from the references.
- The skill is built from the `src/` directory with a Bun command, and the finished output lives in `skills/zvec/`.
How to disable. Delete the `zvec` folder from the agent's skills directory. The Zvec package in the project is removed separately: `pip uninstall zvec` or `npm uninstall @zvec/zvec`.
FAQ
How do I install the Zvec skill in Claude Code?
Run `npx skills add github:zvec-ai/zvec-agent-skills --skill "zvec"`. Then install the package itself: `pip install zvec` for Python or `npm install @zvec/zvec` for Node.js. The skill only explains to the agent how to work with the database.
Does Zvec need a separate server?
No. Zvec runs inside the application process as a library and installs as a single package with no setup. That suits local tools and services where a separate database is overkill.
Which scenarios does the skill cover?
RAG for documents, semantic search, recommendations, search by image together with text, and a hybrid of BM25 keyword search with embeddings. Each has an example in Python and in TypeScript.
How do I combine vector search with filtering?
A query takes a vector and a `filter` condition such as `price >= 100 AND price <= 500 AND in_stock == true`, and the database returns the `topk` nearest documents among the matching ones. The hybrid search section covers it in detail.
How is the skill different from the Zvec documentation itself?
It is the same documentation packed into a skill format with scenario examples. The agent loads the file that matches the topic of the request. For the full API reference look at zvec.org.
Discussion
No questions or reviews yet.
Related
SkillsZvec MCP: vector database for agents
Zvec MCP Server
17 MCP tools let an agent create Zvec collections, write documents and search by meaning with HNSW, IVF and FLAT indexes
Official Weaviate skills: vector database search, data import and ready-made RAG app blueprints for Claude Code and Cursor
Nine Pinecone skills: the agent creates a vector index, loads data, searches it and builds an assistant for documents
A skill that picks a scientific chart from nearly 100 templates for CSV or Excel data and outputs PNG, SVG, PDF and Python code.
Voxel51 skills for computer vision: import datasets, find duplicates, run inference and evaluate models in FiftyOne
VisActor Skills: charts, dashboards and posters
VisActor Skills
Nine VisActor skills: business charts, data stories, map dashboards and posters, all assembled by your agent
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Zvec Agent Skills (this card) | yes | 21 | ||
| zvec-grep | no | not required | 4,229 | a local `zg` search tool for the workspace, while this skill teaches an agent to write code on the Zvec database itself |
| Qdrant Skills | no | 267 | advice for the Qdrant vector database, while here are references and examples for the embedded Zvec database with no server |
README in short
The page explains what Zvec is, lists scenarios and gives the install commands for the skill and the packages. Short Python and Node.js examples follow, then the skill folder tree, build instructions and sample requests to the agent. Topics by language and documentation links close the page.