OpenClaw + seekdb skills: Build Your Own Personal seekdb Assistant

🌟 Tip: The seekdb used in this article is the AI-native database open-sourced by OceanBase. You are welcome to try it out at https://github.com/oceanbase/seekdb — it should bring a cleaner, more efficient data management solution to your AI application development!

This article shows you how to load seekdb Agent Skills in OpenClaw, so it can answer developers’ common questions about seekdb deployment, vector search, hybrid search, integration methods, and more, based on the seekdb official documentation at any time.

What Is OpenClaw?

OpenClaw is so popular that it really needs no introduction. But for completeness, let’s give it a brief one.

OpenClaw is a personal AI assistant that runs on your own device. It talks with you on the communication channels you already use (such as WhatsApp, Telegram, Slack, Discord, Signal, iMessage, WebChat, etc.), and also supports chatting directly through a local TUI or web interface.

Feature overview:

• Local-first: Your data and conversations are under your control, and it can run entirely locally or in a self-hosted environment.
• Multi-channel: The same assistant can connect to many instant-messaging and development tools.
• Extensible: Inject domain knowledge or tool capabilities into the assistant through Skills.

What Is seekdb?

seekdb is the AI-native database released by OceanBase, capable of running on a wide range of devices.

It unifies multiple data types — relational data, vectors, full text, JSON, GIS, and more — in a single engine, and can complete hybrid operations combining vector search, full-text search, and relational query in a single SQL statement.

Feature overview:

• AI-native: Built-in embedding generation, reranking, and LLM inference — complete RAG workflows inside the database.
• Hybrid search: Vector search + full-text search + relational query, all handled in a single SQL statement.
• Lightweight and easy to deploy: Runs on as little as 1 CPU core + 2 GB of memory, and supports embedded, Docker, and RPM deployment methods.
• MySQL-compatible: Compatible with MySQL syntax and ecosystem, supports full ACID transactions, with a low learning curve.
• Open-source and free: Apache 2.0 license, with the code open-sourced on GitHub.

What Is a seekdb Agent Skill?

A seekdb Agent Skill is a set of Agent skills related to the seekdb vector database, used to enhance an AI assistant’s capabilities in seekdb scenarios. This Skill package is provided by the seekdb ecosystem plugin and supports multiple AI tools (such as OpenClaw, Claude Code, Cursor, Codex, etc.).

Currently it mainly includes three categories of skills:

1. seekdb-docs (documentation skill)
   • A built-in seekdb official documentation knowledge base, supporting content-based semantic retrieval.
   • Covers quick start, development guides (vector search, hybrid search, AI functions, etc.), SDK/API references, multi-model data, integration and deployment/operations, hands-on tutorials, and more.
   • When the assistant answers questions like “how to deploy seekdb” or “how to use vector search,” it will prefer consulting the remote documentation, and fall back to local documentation on failure.
2. importing-to-seekdb (import skill)
   • Imports CSV/Excel into seekdb, with optional column vectorization (such as all-MiniLM-L6-v2), for convenient subsequent semantic search.
   • Supports preview, batch import, and collection management.
3. querying-from-seekdb (query skill)
   • Performs scalar/hybrid search against seekdb, supports metadata filtering and RRF ranking, and can export to CSV/Excel.

This article uses OpenClaw + the seekdb documentation skill as an example to show how to install and load the skill and ask the assistant questions in natural language (such as “how to deploy seekdb”). The assistant will retrieve from the documentation based on this Skill and provide an answer.

---

Installing OpenClaw

I had Qwen Code install my OpenClaw directly for me.

Qwen Code is an alternative to Claude Code; its default Qwen3-Coder model has a free quota of 2,000 requests per day, which is plenty for normal use. I’ll recommend it here as well.

Environment requirement: Node.js ≥ 22.

1. One-click install script

curl -fsSL https://openclaw.ai/install.sh | bash

2. Choose QuickStart first; you can configure manually later if needed

3. Choose a provider, enter the API Key in the next step, and select a model

4. Not needed for now — you can skip it first

5. Choose not to configure a skill yet — configure it later. The configuration info shown here is important:

• openclaw.json: OpenClaw’s configuration file. All configuration is written into this file, and you can manually edit this JSON file to modify the configuration.
• Workspace path: ~/.openclaw/workspace, OpenClaw’s default working directory; the skills also need to be placed under this folder later.
• sessions: a directory related to session persistence. An OpenClaw session doesn’t end when you close the terminal; when you open the terminal again, it continues the previous session, unless you enter “/new”, which starts a new session.

6. Hooks are commands OpenClaw runs when performing specific actions (such as startup or starting a new session). You can press the spacebar on each one to multi-select and enable them.

7. Here we choose to use the TUI; you can also open the Web UI at http://127.0.0.1:18789?token=…

8. Enter “what skills do you have?”, and you’ll see there are no seekdb-related skills here. Enter “/exit” to quit first.

For more details, see the official docs: Quick Start and Installation Guide.

Installing the seekdb Agent Skill via pip

The seekdb Agent Skill is published on PyPI as a Python package. After installing it with pip, run the interactive installer to install the skills into OpenClaw’s workspace.

1. Install the Python package

Note: Systems like Ubuntu may not be able to run the command below directly. If you get an error, please create a virtual environment first.

pip install seekdb-agent-skills

2. Run the interactive installer

seekdb-agent-skills

3. Operate within the installer

• Select the tool: Choose OpenClaw from the list.
• Confirm the install path: The installer will install the skills into ~/.openclaw/workspace/skills; just confirm.
• Select skills: seekdb (the documentation skill) is checked by default — you can press Enter to confirm directly.
• Use the arrow keys and Space to multi-select, and Enter to confirm. The installer will copy the corresponding skill directories into the OpenClaw workspace.

After installation, you’ll need to click “New session” in the web page, or enter “/new” in the TUI, to load the new skills (OpenClaw scans the skills under ~/.openclaw/workspace/skills).

4. Confirm whether the skill is installed — there are several ways

Confirm via the command line:

openclaw skills

The expected output is as follows:

Confirm via the Web UI:

Click the Skills tab and type seekdb to search.

Installing the seekdb Agent Skill via ClawHub

Besides pip, you can also install the seekdb Agent Skill via ClawHub.

The difference between the two installation methods is: ClawHub doesn’t support downloading a large number of files, so the seekdb Agent Skill on ClawHub only provides remote mode (pulling docs from GitHub) and does not include local documentation, so you need GitHub access before using it; whereas the version installed via PyPI supports both local mode and remote mode, and can fall back to local documentation when there’s no network or GitHub is unavailable.

1. Install clawhub

npm install -g clawhub

2. Install the seekdb-docs skill

clawhub install seekdb-docs

After installation, verifying whether the skill is installed works the same as with pip: you can run openclaw skills on the command line to view the skill list, or click the Skills tab in the Web UI and type seekdb to search.

Start the Conversation

Note: OpenClaw’s actual performance here will depend heavily on the model you choose.

1. Check What Skills Are Currently Available

Enter “what skills do you have?”, and it shows the seekdb-docs skill.

2. Open the TUI

Because OpenClaw’s Web UI page and the TUI point to the same session, for this we’ll open the TUI.

Enter the following command:

openclaw tui

We didn’t enter any question, but opening the TUI carried over the question we’d asked in the Web UI.

3. Ask How to Deploy seekdb

Enter “how to deploy seekdb?”, and it correctly answered seekdb’s deployment methods.

You can keep following up, for example:

1. “how to use vector search in seekdb?”
2. “Which AI frameworks does seekdb support integrating with?”
3. “How do I implement hybrid search in seekdb?” and so on.

As long as the question is related to the seekdb documentation, OpenClaw will prefer using this set of Skills to answer it and perform the relevant tasks.

Related Resources

• seekdb official site: https://oceanbase.ai/zh-CN/
• seekdb project repository: https://github.com/oceanbase/seekdb
• seekdb official documentation: https://www.oceanbase.ai/docs/