Skill: seekdb

Owner: davidzhangbj

Summary: seekdb database documentation lookup. Use when users ask about seekdb features, SQL syntax, vector search, hybrid search, integrations, deployment, or any seekdb-related topics. Automatically locates relevant docs via catalog-based semantic search.

Tags: latest:0.1.5

Version history:

v0.1.5 | 2026-02-10T03:11:34.500Z | auto

seekdb-docs 0.1.5

• Improved clarity in SKILL.md workflow and guidelines, especially around when to use grep versus loading the full catalog.
• Updated instructions to clarify grep as the preferred method for most keyword queries.
• Added concise explanation for catalog search strategies and fallback handling.
• Refined examples and step ordering for better usability and quick start guidance.
• Enhanced documentation navigation by specifying where to find detailed workflow examples.

v0.1.4 | 2026-02-10T02:48:27.470Z | auto

Summary: Local-first catalog/documentation search with dynamic path resolution, remote fallback, and improved usage instructions.

• Added dynamic skill directory path resolution to reliably access local catalog and documentation regardless of working directory.
• Switched to a local-first strategy: use local catalog/docs (if present) with remote GitHub fallback for both catalog and documents.
• Catalog is now a single JSONL file (1015 entries) optimized for keyword (grep) and semantic search.
• New scripts/update_docs.sh added for optional local doc sync; full example workflows provided in references/examples.md.
• Usage guidance updated: always resolve path via reading SKILL.md, prefer grep-based queries, load full catalog only for broad/complex queries, handle both local and remote sources.

v0.0.6 | 2026-02-03T06:58:54.203Z | user

• Updated example documentation URLs to reflect new paths in the "Develop" section (e.g., vector and hybrid search examples).
• No changes to core logic or features; documentation and usage instructions remain consistent.
• Ensured alignment of example file paths with current remote documentation structure.

v0.0.5 | 2026-02-03T03:06:43.130Z | user

• Removed all local documentation files and local fallback support (952 files deleted).
• Switched documentation access to remote-only; all docs are now fetched from GitHub.
• Updated skill description and usage instructions to reflect remote-only operation.
• Added guidance for error handling when the remote documentation is unavailable.

v0.0.4 | 2026-02-02T07:10:00.781Z | user

• Improved documentation access with a remote-first, local-fallback strategy for seeking seekdb documentation.
• Clearly described catalog-driven matching: match user queries semantically to catalog descriptions and categories.
• Outlined structured processes for remote and local documentation lookup, with consistent behavior within conversations.
• Added sample usage scenarios and guidance for querying, integration, and fallback handling.
• Provided directory structure details and quick reference for local and remote documentation paths.

Archive index:

Archive v0.1.5: 4 files, 7251 bytes

Files: references/examples.md (9718b), scripts/update_docs.sh (1917b), SKILL.md (8240b), _meta.json (130b)

File v0.1.5:SKILL.md

---

name: seekdb-docs description: seekdb database documentation lookup. Use when users ask about seekdb features, SQL syntax, vector search, hybrid search, integrations, deployment, or any seekdb-related topics. Automatically locates relevant docs via catalog-based semantic search. version: "V1.1.0"

seekdb Documentation

Provides comprehensive access to seekdb database documentation through a centralized catalog system.

Quick Start

1. Locate skill directory (see Path Resolution below)
2. Load full catalog (1015 documentation entries)
3. Match query to catalog entries semantically
4. Read document from matched entry

Path Resolution (Critical First Step)

Problem: Relative paths like ./seekdb-docs/ are resolved from the current working directory, not from SKILL.md's location. This breaks when the agent's working directory differs from the skill directory.

Solution: Dynamically locate the skill directory before accessing docs.

Step-by-Step Resolution

1. Read SKILL.md itself to get its absolute path:

   read(SKILL.md)  // or any known file in this skill directory
   

2. Extract the directory from the returned path:

   If read returns: /root/test-claudecode-url/.cursor/skills/seekdb/SKILL.md
   Skill directory: /root/test-claudecode-url/.cursor/skills/seekdb/
   

3. Construct paths using this directory:

   Catalog path: <skill directory>references/seekdb-docs-catalog.jsonl
   Docs base: <skill directory>seekdb-docs/
   

Documentation Sources

Full Catalog

• Local: <skill directory>references/seekdb-docs-catalog.jsonl (1015 entries, JSONL format)
• Remote: https://raw.githubusercontent.com/oceanbase/seekdb-ecology-plugins/main/agent-skills/skills/seekdb/references/seekdb-docs-catalog.jsonl (fallback)
• Entries: 1015 documentation files
• Coverage: Complete seekdb documentation
• Format: JSONL - one JSON object per line with path and description

Complete Documentation (Local-First with Remote Fallback)

Local Documentation (if available):

• Base Path: <skill directory>seekdb-docs/
• Size: 7.4M, 952 markdown files
• Document Path: Base Path + File Path

Remote Documentation (fallback):

• Base URL: https://raw.githubusercontent.com/oceanbase/seekdb-doc/V1.1.0/en-US/
• Document URL: Base URL + File Path

Strategy:

1. Locate: Determine <skill directory> using path resolution above
2. Load: Load full catalog (1015 entries) - try local first, fallback to remote
3. Search: Semantic search through all catalog entries
4. Read: Try local docs first, fallback to remote URL if missing

Workflow

Step 0: Resolve Path (Do this first!)

# Read this file to discover its absolute path
read("SKILL.md")

# Extract directory from the path
# Example: /root/.claude/skills/seekdb/SKILL.md → /root/.claude/skills/seekdb/

Step 1: Search Catalog

Start with grep for keyword searches. Only load full catalog when necessary.

Method 1: Grep Search (Preferred for 90% of queries)

Use grep to search for keywords in the catalog:

grep -i "keyword" <skill directory>references/seekdb-docs-catalog.jsonl

Examples:

# Find macOS deployment docs
grep -i "mac" references/seekdb-docs-catalog.jsonl

# Find Docker deployment docs
grep -i "docker\|container" references/seekdb-docs-catalog.jsonl

# Find vector search docs
grep -i "vector" references/seekdb-docs-catalog.jsonl

Method 2: Load Full Catalog (Only when necessary)

Load the complete catalog only when:

• Grep returns no results
• Complex semantic matching is required
• No specific keyword to search

Local: <skill directory>references/seekdb-docs-catalog.jsonl
Remote: https://raw.githubusercontent.com/oceanbase/seekdb-ecology-plugins/main/agent-skills/skills/seekdb/references/seekdb-docs-catalog.jsonl (fallback)
Format: JSONL (one JSON object per line)
Entries: 1015 documentation files

Strategy:

1. Try local catalog first: <skill directory>references/seekdb-docs-catalog.jsonl
2. If local missing, fetch from remote URL above

Catalog contents:

• Each line: {"path": "...", "description": "..."}
• All seekdb documentation indexed
• Optimized for semantic search and grep queries

Step 2: Match Query

Analyze search results to identify the most relevant documents:

For grep results:

• Review matched lines from grep output
• Extract path and description from each match
• Select documents whose descriptions best match the query
• Consider multiple matches for comprehensive answers

For full catalog:

• Parse each line as JSON to extract path and description
• Perform semantic matching on description text
• Match by meaning, not just keywords
• Return all relevant entries for comprehensive answers

Note: The catalog contains path and description fields. The description field contains topic and feature keywords, making it suitable for both keyword and semantic matching.

Step 3: Read Document

Local-First Strategy:

1. Try local first: <skill directory>seekdb-docs/[File Path]

   • If file exists → read locally (fast)
   • If file missing → proceed to step 2

2. Fallback to remote: https://raw.githubusercontent.com/oceanbase/seekdb-doc/V1.1.0/en-US/[File Path]

   • Download from GitHub

Example:

Query: "How to integrate with Claude Code?"

1. Resolve path: read(SKILL.md) → /root/.claude/skills/seekdb/SKILL.md
   Skill directory : /root/.claude/skills/seekdb/

2. Search catalog with grep:
   grep -i "claude code" /root/.claude/skills/seekdb/references/seekdb-docs-catalog.jsonl

3. Match query from grep results:
   → Found: {"path": "300.integrations/300.developer-tools/700.claude-code.md",
             "description": "This guide explains how to use the seekdb plugin with Claude Code..."}
   → This matches the query, select this document

4. Read doc:
   Try: /root/.claude/skills/seekdb/seekdb-docs/300.integrations/300.developer-tools/700.claude-code.md
   If missing: https://raw.githubusercontent.com/oceanbase/seekdb-doc/V1.1.0/en-US/300.integrations/300.developer-tools/700.claude-code.md

Guidelines

• Always resolve path first: Use the read-your-SKILL.md trick to get the absolute path
• Prefer grep for keyword queries: Load full catalog only when grep returns nothing or semantic matching is needed
• Semantic matching: Match by meaning, not just keywords
• Multiple matches: Read all relevant entries for comprehensive answers
• Local-first with remote fallback: Try local docs first, use remote URL if missing
• Optional local docs: Run scripts/update_docs.sh to download full docs locally (faster)
• Offline capable: With local docs present, works completely offline

Catalog Search Format

The catalog file is in JSONL format (one JSON object per line):

{"path": "path/to/document.md", "description": "Document description text"}

Searching the catalog:

• Keyword search: Use grep (see Step 1 examples). Each matched line contains both path and description.
• When grep is insufficient: Read the full catalog, parse each line as JSON, then do semantic matching on descriptions.

Common Installation Paths

This skill may be installed at:

• Cursor: .cursor/skills/seekdb/
• Claude Code: .claude/skills/seekdb/
• Custom: Any directory (path resolution handles this automatically)

Do not hardcode these paths. Use the dynamic resolution method instead.

Detailed Examples

See examples.md for complete workflow examples including:

• Full catalog search scenarios
• Local-first lookup scenarios
• Remote fallback scenarios
• Integration queries
• Multi-turn conversations

Category Overview

• Get Started: Quick start, basic operations, overview
• Development: Vector search, hybrid search, AI functions, MCP, multi-model
• Integrations: Frameworks, model platforms, developer tools, workflows
• Guides: Deployment, management, security, OBShell, performance
• Reference: SQL syntax, PL, error codes, SDK APIs
• Tutorials: Step-by-step scenarios

File v0.1.5:_meta.json

{ "ownerId": "kn7ajpckchvqztcwpmckyc9fxd80c3d5", "slug": "seekdb-docs", "version": "0.1.5", "publishedAt": 1770693094500 }

File v0.1.5:references/examples.md

seekdb Documentation Examples

Complete workflow examples for common seekdb documentation queries using local-first strategy.

Example 1: Vector Search Query

User Query: "How do I use vector search in seekdb?"

Process:

1. Resolve skill directory

   Read SKILL.md → /root/.claude/skills/seekdb/SKILL.md
   Extract: skill directory = /root/.claude/skills/seekdb/
   

2. Load catalog (local-first)

   Try local: /root/.claude/skills/seekdb/references/seekdb-docs-catalog.jsonl
   Result: Success ✅
   

3. Search catalog for vector search

   • Parse JSONL entries
   • Query: "vector search"
   • Found matches:

     {"path": "200.develop/600.search/300.vector-search/100.vector-search-intro.md",
      "description": "This document provides an overview of vector search in seekdb, covering core concepts, supported vector data types, indexing methods, and search operations."}
     {"path": "200.develop/600.search/300.vector-search/300.vector-similarity-search.md",
      "description": "This document explains vector similarity search methods in seekdb..."}
     

4. Read documents (local-first)

   • Try local: /root/.claude/skills/seekdb/seekdb-docs/200.develop/600.search/300.vector-search/100.vector-search-intro.md
   • If missing → Remote: https://raw.githubusercontent.com/oceanbase/seekdb-doc/V1.1.0/en-US/200.develop/600.search/300.vector-search/100.vector-search-intro.md

5. Provide answer

---

Example 2: Basic Overview Query

User Query: "What is seekdb?"

Process:

1. Resolve skill directory

   Read SKILL.md → get absolute path
   Extract skill directory
   

2. Load catalog

   Local: <skill directory>references/seekdb-docs-catalog.jsonl
   Success ✅
   

3. Search catalog for overview

   • Query: "seekdb overview" or "what is seekdb"
   • Found entry:

     {"path": "100.get-started/10.seekdb-overview/10.seekdb-overview.md",
      "description": "An AI-native search database that unifies relational, vector, text, JSON, and GIS data in a single engine, enabling hybrid search and in-database AI workflows with full MySQL compatibility."}
     

4. Read document

   • Local: <skill directory>seekdb-docs/100.get-started/10.seekdb-overview/10.seekdb-overview.md
   • Success ✅

5. Extract answer from local docs

---

Example 3: Remote Fallback Scenario

User Query: "How do I deploy seekdb on macOS?"

Process:

1. Resolve skill directory

   Read SKILL.md → /path/to/skills/seekdb/SKILL.md
   Skill directory = /path/to/skills/seekdb/
   

2. Load catalog (local-first)

   Try local: /path/to/skills/seekdb/references/seekdb-docs-catalog.jsonl
   Result: File not found ❌
   

3. Fallback to remote catalog

   Fetch: https://raw.githubusercontent.com/oceanbase/seekdb-ecology-plugins/main/agent-skills/skills/seekdb/references/seekdb-docs-catalog.jsonl
   Result: Success ✅
   

4. Search catalog

   • Query: "macOS", "deploy", "systemd"
   • Found entry:

     {"path": "400.guides/400.deploy/700.server-mode/100.deploy-by-systemd.md",
      "description": "This document provides instructions for deploying and managing the seekdb database using systemd on RPM-based, DEB-based, and macOS systems, covering installation, configuration, connection, and removal procedures."}
     

5. Read document (remote only, since local missing)

   URL: https://raw.githubusercontent.com/oceanbase/seekdb-doc/V1.1.0/en-US/400.guides/400.deploy/700.server-mode/100.deploy-by-systemd.md
   Fetch and parse
   

6. Provide answer

---

Example 4: Integration Query

User Query: "I want to integrate seekdb with Claude Code"

Process:

1. Resolve path & load catalog (local success)

2. Search catalog for integration

   • Query: "Claude Code", "integration", "AI coding"
   • Found entry:

     {"path": "100.get-started/200.ai-coding/200.building-an-ai-programming-assistant-that-understands-vector-databases-with-seekdb-claude-code.md",
      "description": "This tutorial explains how to build a vector-aware AI coding assistant by integrating the seekdb plugin with Claude Code, enabling accurate documentation retrieval and hybrid search application development in Python."}
     

3. Read document (local)

4. Extract integration steps and provide code examples

---

Example 5: SQL Syntax Query

User Query: "What's the syntax for creating a vector column?"

Process:

1. Resolve path & load catalog (local success)

2. Search for SQL/vector column syntax

   • Query: "vector column", "CREATE TABLE", "vector type"
   • Found entry:

     {"path": "200.develop/600.search/300.vector-search/150.vector-data-type.md",
      "description": "This document describes seekdb's vector data types for AI vector search, covering storage of dense and sparse floating-point arrays, index creation, and hybrid search capabilities."}
     

3. Read document (local)

4. Provide syntax with examples

---

Example 6: Hybrid Search Query

User Query: "How does hybrid search work?"

Process:

1. Resolve path & load catalog (local success)

2. Search for hybrid search

   • Query: "hybrid search", "combined search"
   • Found entry:

     {"path": "200.develop/600.search/500.hybrid-search.md",
      "description": "This document describes hybrid search in seekdb, combining vector-based semantic search and full-text keyword search for more accurate results through integrated ranking."}
     

3. Read document (local)

4. Explain hybrid search concepts with usage examples

---

Example 7: Multiple Matches Query

User Query: "Tell me about seekdb indexes"

Process:

1. Resolve path & load catalog (local success)

2. Search for all index types

   • Query: "index"
   • Found multiple matches:

     {"path": "200.develop/600.search/300.vector-search/200.vector-index/100.vector-index-overview.md", "description": "...vector index types..."}
     {"path": "200.develop/200.design-database-schema/35.multi-model/300.char-and-text/300.full-text-index.md", "description": "This document details the creation and use of full-text indexes in seekdb, covering supported column types, tokenizers, and complex DML operations on indexed tables."}
     {"path": "200.develop/200.design-database-schema/40.create-index-in-develop.md", "description": "...creating indexes..."}
     

3. Read all relevant documents (local)

4. Provide comprehensive answer covering:

   • Vector indexes (HNSW, IVF)
   • Full-text indexes
   • Index creation syntax
   • Best practices

---

Example 8: Python SDK Query

User Query: "How do I use seekdb with Python?"

Process:

1. Resolve path & load catalog (local success)

2. Search for Python SDK

   • Query: "Python", "SDK", "pyseekdb"
   • Found entries:

     {"path": "100.get-started/50.use-seekdb-with-sdk/25.using-seekdb-in-python-sdk.md",
      "description": "This document provides a step-by-step guide to using OceanBase's pyseekdb Python SDK for embedded seekdb operations..."}
     {"path": "200.develop/100.connect-to-seekdb/300.sample-program/200.python/100.mysqlclient-connection-to-seekdb-sample-program.md", "description": "This document provides a sample program demonstrating how to connect to seekdb using mysqlclient in Python, including steps for creating tables, inserting data, and querying data."}
     

3. Read documents (local)

4. Provide Python setup and code examples

---

Example 9: Deployment Query

User Query: "How do I deploy seekdb in production?"

Process:

1. Resolve path & load catalog (local success)

2. Search deployment guides

   • Query: "production deployment", "install"
   • Found entries:

     {"path": "400.guides/400.deploy/50.deploy-overview.md", "description": "This documentation outlines seekdb's two deployment modes: embedded mode and server mode, and details the specific deployment methods available for each mode."}
     

3. Read deployment guides (local)

4. Extract deployment steps:

   • Prerequisites
   • Installation methods
   • Configuration
   • Best practices

---

Key Patterns

Path Resolution (Always First)

1. Read SKILL.md to get absolute path
2. Extract skill directory from path
3. Construct catalog path: <skill directory>references/seekdb-docs-catalog.jsonl

Catalog Loading (Local-First)

1. Try local catalog: <skill directory>references/seekdb-docs-catalog.jsonl
2. If missing, fetch remote: https://raw.githubusercontent.com/oceanbase/seekdb-ecology-plugins/main/agent-skills/skills/seekdb/references/seekdb-docs-catalog.jsonl
3. Parse JSONL (one JSON object per line)

Searching

• Semantic matching: Match query meaning to description text
• Multiple results: Read all relevant entries for comprehensive answers
• Fields available: path and description only

Document Reading (Local-First)

1. Try local: <skill directory>seekdb-docs/[File Path]
2. If missing, fetch remote: https://raw.githubusercontent.com/oceanbase/seekdb-doc/V1.1.0/en-US/[File Path]

Best Practices

• ✅ Always resolve path first - Never hardcode paths
• ✅ Local-first for speed - Try local before remote
• ✅ Semantic over keyword - Understand meaning, not just match words
• ✅ Read multiple sources - Comprehensive answers from all relevant docs
• ✅ Parse JSONL correctly - Each line is a valid JSON object with path and description

Archive v0.1.4: 4 files, 7114 bytes

Files: references/examples.md (9092b), scripts/update_docs.sh (1917b), SKILL.md (8522b), _meta.json (130b)

File v0.1.4:SKILL.md

---

name: seekdb-docs description: seekdb database documentation lookup. Use when users ask about seekdb features, SQL syntax, vector search, hybrid search, integrations, deployment, or any seekdb-related topics. Automatically locates relevant docs via catalog-based semantic search. version: "V1.1.0"

seekdb Documentation

Provides comprehensive access to seekdb database documentation through a centralized catalog system.

Quick Start

1. Locate skill directory (see Path Resolution below)
2. Load full catalog (1015 documentation entries)
3. Match query to catalog entries semantically
4. Read document from matched entry

Path Resolution (Critical First Step)

Problem: Relative paths like ./seekdb-docs/ are resolved from the current working directory, not from SKILL.md's location. This breaks when the agent's working directory differs from the skill directory.

Solution: Dynamically locate the skill directory before accessing docs.

Step-by-Step Resolution

1. Read SKILL.md itself to get its absolute path:

   read(SKILL.md)  // or any known file in this skill directory
   

2. Extract the directory from the returned path:

   If read returns: /root/test-claudecode-url/.cursor/skills/seekdb/SKILL.md
   Skill directory =: /root/test-claudecode-url/.cursor/skills/seekdb/
   

3. Construct paths using this directory:

   Catalog path =: <skill directory>references/seekdb-docs-catalog.jsonl
   Docs base =: <skill directory>seekdb-docs/
   

Documentation Sources

Full Catalog

• Local: <skill directory>references/seekdb-docs-catalog.jsonl (1015 entries, JSONL format)
• Remote: https://raw.githubusercontent.com/oceanbase/seekdb-ecology-plugins/main/agent-skills/skills/seekdb/references/seekdb-docs-catalog.jsonl (fallback)
• Entries: 1015 documentation files
• Coverage: Complete seekdb documentation
• Format: JSONL - one JSON object per line with path and description

Complete Documentation (Local-First with Remote Fallback)

Local Documentation (if available):

• Base Path: <skill directory>seekdb-docs/
• Size: 7.4M, 952 markdown files
• Document Path: Base Path + File Path

Remote Documentation (fallback):

• Base URL: https://raw.githubusercontent.com/oceanbase/seekdb-doc/V1.1.0/en-US/
• Document URL: Base URL + File Path

Strategy:

1. Locate: Determine <skill directory> using path resolution above
2. Load: Load full catalog (1015 entries) - try local first, fallback to remote
3. Search: Semantic search through all catalog entries
4. Read: Try local docs first, fallback to remote URL if missing

Workflow

Step 0: Resolve Path (Do this first!)

# Read this file to discover its absolute path
read("SKILL.md")

# Extract directory from the path
# Example: /root/.claude/skills/seekdb/SKILL.md → /root/.claude/skills/seekdb/

Step 1: Search Catalog

Start with grep for keyword searches. Only load full catalog when necessary.

Method 1: Grep Search (Preferred for 90% of queries)

Use grep to search for keywords in the catalog:

grep -i "keyword" <skill directory>references/seekdb-docs-catalog.jsonl

Examples:

# Find macOS deployment docs
grep -i "mac" references/seekdb-docs-catalog.jsonl

# Find Docker deployment docs
grep -i "docker\|container" references/seekdb-docs-catalog.jsonl

# Find vector search docs
grep -i "vector" references/seekdb-docs-catalog.jsonl

Method 2: Load Full Catalog (Only when necessary)

Load the complete catalog only when:

• Grep returns no results
• Complex semantic matching is required
• No specific keyword to search

Local: <skill directory>references/seekdb-docs-catalog.jsonl
Remote: https://raw.githubusercontent.com/oceanbase/seekdb-ecology-plugins/main/agent-skills/skills/seekdb/references/seekdb-docs-catalog.jsonl (fallback)
Format: JSONL (one JSON object per line)
Entries: 1015 documentation files

Strategy:

1. Try local catalog first: <skill directory>references/seekdb-docs-catalog.jsonl
2. If local missing, fetch from remote URL above

Catalog contents:

• Each line: {"path": "...", "description": "..."}
• All seekdb documentation indexed
• Optimized for semantic search and grep queries

Step 2: Match Query

Analyze search results to identify the most relevant documents:

For grep results:

• Review matched lines from grep output
• Extract path and description from each match
• Select documents whose descriptions best match the query
• Consider multiple matches for comprehensive answers

For full catalog:

• Parse each line as JSON to extract path and description
• Perform semantic matching on description text
• Match by meaning, not just keywords
• Return all relevant entries for comprehensive answers

Note: The catalog contains path and description fields. The description field contains topic and feature keywords, making it suitable for both keyword and semantic matching.

Step 3: Read Document

Local-First Strategy:

1. Try local first: <skill directory>seekdb-docs/[File Path]

   • If file exists → read locally (fast)
   • If file missing → proceed to step 2

2. Fallback to remote: https://raw.githubusercontent.com/oceanbase/seekdb-doc/V1.1.0/en-US/[File Path]

   • Download from GitHub

Example:

Query: "How to integrate with Claude Code?"

1. Resolve path: read(SKILL.md) → /root/.claude/skills/seekdb/SKILL.md
   Skill directory =: /root/.claude/skills/seekdb/

2. Search catalog with grep:
   grep -i "claude code" references/seekdb-docs-catalog.jsonl

3. Match query from grep results:
   → Found: {"path": "300.integrations/300.developer-tools/700.claude-code.md",
             "description": "This guide explains how to use the seekdb plugin with Claude Code..."}
   → This matches the query, select this document

4. Read doc:
   Try: /root/.claude/skills/seekdb/seekdb-docs/300.integrations/300.developer-tools/700.claude-code.md
   If missing: https://raw.githubusercontent.com/oceanbase/seekdb-doc/V1.1.0/en-US/300.integrations/300.developer-tools/700.claude-code.md

Guidelines

• Always resolve path first: Use the read-your-SKILL.md trick to get the absolute path
• Always use full catalog: Load complete catalog for comprehensive search (1015 entries)
• Semantic matching: Match by meaning, not just keywords
• Multiple matches: Read all relevant entries for comprehensive answers
• Local-first with remote fallback: Try local docs first, use remote URL if missing
• Optional local docs: Run scripts/update_docs.sh to download full docs locally (faster)
• Offline capable: With local docs present, works completely offline

Catalog Search Format

The catalog file is in JSONL format (one JSON object per line):

{"path": "path/to/document.md", "description": "Document description text"}

Searching the catalog:

Preferred method - Read the full catalog:

• Read the entire catalog file
• Parse each line as JSON to extract path and description
• Perform semantic matching on description text

Quick keyword search - Use grep:

grep -i "keyword" references/seekdb-docs-catalog.jsonl

Each matched line contains both path and description together.

Why JSONL is better:

• ✅ One line per document - grep shows complete information
• ✅ Structured data - easy to parse with JSON tools
• ✅ No special flags needed - works naturally with grep, jq, etc.

Common Installation Paths

This skill may be installed at:

• Cursor: .cursor/skills/seekdb/
• Claude Code: .claude/skills/seekdb/
• Custom: Any directory (path resolution handles this automatically)

Do not hardcode these paths. Use the dynamic resolution method instead.

Detailed Examples

See examples.md for complete workflow examples including:

• Full catalog search scenarios
• Local-first lookup scenarios
• Remote fallback scenarios
• Integration queries
• Multi-turn conversations

Category Overview

• Get Started: Quick start, basic operations, overview
• Development: Vector search, hybrid search, AI functions, MCP, multi-model
• Integrations: Frameworks, model platforms, developer tools, workflows
• Guides: Deployment, management, security, OBShell, performance
• Reference: SQL syntax, PL, error codes, SDK APIs
• Tutorials: Step-by-step scenarios

File v0.1.4:_meta.json

{ "ownerId": "kn7ajpckchvqztcwpmckyc9fxd80c3d5", "slug": "seekdb-docs", "version": "0.1.4", "publishedAt": 1770691707470 }

File v0.1.4:references/examples.md

seekdb Documentation Examples

Complete workflow examples for common seekdb documentation queries using local-first strategy.

Example 1: Vector Search Query

User Query: "How do I use vector search in seekdb?"

Process:

1. Resolve skill directory

   Read SKILL.md → /root/.claude/skills/seekdb/SKILL.md
   Extract: skill directory = /root/.claude/skills/seekdb/
   

2. Load catalog (local-first)

   Try local: /root/.claude/skills/seekdb/references/seekdb-docs-catalog.jsonl
   Result: Success ✅
   

3. Search catalog for vector search

   • Parse JSONL entries
   • Query: "vector search"
   • Found matches:

     {"path": "200.develop/600.search/300.vector-search/100.vector-search-intro.md",
      "description": "This document provides an overview of vector search in seekdb, covering core concepts, supported vector data types, indexing methods, and search operations."}
     {"path": "200.develop/600.search/300.vector-search/300.vector-similarity-search.md",
      "description": "This document explains vector similarity search methods in seekdb..."}
     

4. Read documents (local-first)

   • Try local: /root/.claude/skills/seekdb/seekdb-docs/200.develop/600.search/300.vector-search/100.vector-search-intro.md
   • If missing → Remote: https://raw.githubusercontent.com/oceanbase/seekdb-doc/V1.1.0/en-US/200.develop/600.search/300.vector-search/100.vector-search-intro.md

5. Provide answer

---

Example 2: Basic Overview Query

User Query: "What is seekdb?"

Process:

1. Resolve skill directory

   Read SKILL.md → get absolute path
   Extract skill directory
   

2. Load catalog

   Local: <skill directory>references/seekdb-docs-catalog.jsonl
   Success ✅
   

3. Search catalog for overview

   • Query: "seekdb overview" or "what is seekdb"
   • Found entry:

     {"path": "100.get-started/10.seekdb-overview/10.seekdb-overview.md",
      "description": "An AI-native search database that unifies relational, vector, text, JSON, and GIS data in a single engine, enabling hybrid search and in-database AI workflows with full MySQL compatibility."}
     

4. Read document

   • Local: <skill directory>seekdb-docs/100.get-started/10.seekdb-overview/10.seekdb-overview.md
   • Success ✅

5. Extract answer from local docs

---

Example 3: Remote Fallback Scenario

User Query: "How do I deploy seekdb on macOS?"

Process:

1. Resolve skill directory

   Read SKILL.md → /path/to/skills/seekdb/SKILL.md
   Skill directory = /path/to/skills/seekdb/
   

2. Load catalog (local-first)

   Try local: /path/to/skills/seekdb/references/seekdb-docs-catalog.jsonl
   Result: File not found ❌
   

3. Fallback to remote catalog

   Fetch: https://raw.githubusercontent.com/oceanbase/seekdb-ecology-plugins/main/agent-skills/skills/seekdb/references/seekdb-docs-catalog.jsonl
   Result: Success ✅
   

4. Search catalog

   • Query: "macOS", "deploy", "systemd"
   • Found entry:

     {"path": "400.guides/200.deployment/500.deploy-with-systemd.md",
      "description": "This document provides instructions for deploying and managing the seekdb database using systemd on RPM-based, DEB-based, and macOS systems..."}
     

5. Read document (remote only, since local missing)

   URL: https://raw.githubusercontent.com/oceanbase/seekdb-doc/V1.1.0/en-US/400.guides/200.deployment/500.deploy-with-systemd.md
   Fetch and parse
   

6. Provide answer

---

Example 4: Integration Query

User Query: "I want to integrate seekdb with Claude Code"

Process:

1. Resolve path & load catalog (local success)

2. Search catalog for integration

   • Query: "Claude Code", "integration", "AI coding"
   • Found entry:

     {"path": "100.get-started/200.ai-coding/200.building-an-ai-programming-assistant-that-understands-vector-databases-with-seekdb-claude-code.md",
      "description": "This tutorial explains how to build a vector-aware AI coding assistant by integrating the seekdb plugin with Claude Code, enabling accurate documentation retrieval and hybrid search application development in Python."}
     

3. Read document (local)

4. Extract integration steps and provide code examples

---

Example 5: SQL Syntax Query

User Query: "What's the syntax for creating a vector column?"

Process:

1. Resolve path & load catalog (local success)

2. Search for SQL/vector column syntax

   • Query: "vector column", "CREATE TABLE", "vector type"
   • Found entry:

     {"path": "200.develop/600.search/300.vector-search/150.vector-data-type.md",
      "description": "This document describes seekdb's vector data types for AI vector search, covering storage of dense and sparse floating-point arrays, index creation, and hybrid search capabilities."}
     

3. Read document (local)

4. Provide syntax with examples

---

Example 6: Hybrid Search Query

User Query: "How does hybrid search work?"

Process:

1. Resolve path & load catalog (local success)

2. Search for hybrid search

   • Query: "hybrid search", "combined search"
   • Found entry:

     {"path": "200.develop/600.search/500.hybrid-search.md",
      "description": "This document describes hybrid search in seekdb, combining vector-based semantic search and full-text keyword search for more accurate results through integrated ranking."}
     

3. Read document (local)

4. Explain hybrid search concepts with usage examples

---

Example 7: Multiple Matches Query

User Query: "Tell me about seekdb indexes"

Process:

1. Resolve path & load catalog (local success)

2. Search for all index types

   • Query: "index"
   • Found multiple matches:

     {"path": "200.develop/600.search/300.vector-search/200.vector-index/100.vector-index-overview.md", "description": "...vector index types..."}
     {"path": "200.develop/300.read-and-write-data/200.read-data/300.full-text-index.md", "description": "...full-text indexes..."}
     {"path": "200.develop/200.design-database-schema/40.create-index-in-develop.md", "description": "...creating indexes..."}
     

3. Read all relevant documents (local)

4. Provide comprehensive answer covering:

   • Vector indexes (HNSW, IVF)
   • Full-text indexes
   • Index creation syntax
   • Best practices

---

Example 8: Python SDK Query

User Query: "How do I use seekdb with Python?"

Process:

1. Resolve path & load catalog (local success)

2. Search for Python SDK

   • Query: "Python", "SDK", "pyseekdb"
   • Found entries:

     {"path": "100.get-started/50.use-seekdb-with-sdk/25.using-seekdb-in-python-sdk.md",
      "description": "This document provides a step-by-step guide to using OceanBase's pyseekdb Python SDK for embedded seekdb operations..."}
     {"path": "200.develop/100.connect-to-seekdb/300.sample-program/200.python/...", "description": "..."}
     

3. Read documents (local)

4. Provide Python setup and code examples

---

Example 9: Deployment Query

User Query: "How do I deploy seekdb in production?"

Process:

1. Resolve path & load catalog (local success)

2. Search deployment guides

   • Query: "production deployment", "install"
   • Found entries:

     {"path": "400.guides/200.deployment/...", "description": "...deployment instructions..."}
     

3. Read deployment guides (local)

4. Extract deployment steps:

   • Prerequisites
   • Installation methods
   • Configuration
   • Best practices

---

Key Patterns

Path Resolution (Always First)

1. Read SKILL.md to get absolute path
2. Extract skill directory from path
3. Construct catalog path: <skill directory>references/seekdb-docs-catalog.jsonl

Catalog Loading (Local-First)

1. Try local catalog: <skill directory>references/seekdb-docs-catalog.jsonl
2. If missing, fetch remote: https://raw.githubusercontent.com/oceanbase/seekdb-ecology-plugins/main/agent-skills/skills/seekdb/references/seekdb-docs-catalog.jsonl
3. Parse JSONL (one JSON object per line)

Searching

• Semantic matching: Match query meaning to description text
• Multiple results: Read all relevant entries for comprehensive answers
• Fields available: path and description only

Document Reading (Local-First)

1. Try local: <skill directory>seekdb-docs/[File Path]
2. If missing, fetch remote: https://raw.githubusercontent.com/oceanbase/seekdb-doc/V1.1.0/en-US/[File Path]

Best Practices

• ✅ Always resolve path first - Never hardcode paths
• ✅ Local-first for speed - Try local before remote
• ✅ Semantic over keyword - Understand meaning, not just match words
• ✅ Read multiple sources - Comprehensive answers from all relevant docs
• ✅ Parse JSONL correctly - Each line is a valid JSON object with path and description