kavak-documentation（卡瓦克文档）

Kavak Documentation

Search Kavak's internal documentation first via platform_docs_search and search_resource MCP tools, then complement with other sources.

Related Skills: - backend-development - After finding Kavak patterns, implement using backend best practices - Check references/ folder for detailed Kavak implementation guides

Core Principle

Search Kavak docs first, always. Internal platform knowledge reduces agent errors and ensures proper Kavak patterns.

When to Use

• Planning new features or architecture
• Architecture, kbroker, STS, SDKs, databases
• Creating new services/workloads (Go, TypeScript, Python, Java)
• GitLab CI/CD pipelines and configurations
• Docker/container configurations
• Development workflow, logs, metrics best practices
• Kavak CLI usage/installation

Quick Start

1. Verify kavak-platform MCP is configured
2. Use platform_docs_search for documentation queries and search_resource for workload/infrastructure info
3. Complement results with other tools as needed

Workflow

1. Check MCP Availability

Ensure kavak-platform MCP server is enabled. If MCP is unavailable, fall back to: searching kavak-it repos via gitlab_code_search/gitlab_file_content, or checking the project's existing patterns as the next best source of truth.

2. Query Documentation

Rules for platform_docs_search:

• One question per call (avoid compound questions)
• Use llm: true for synthesized answers, llm: false for raw doc chunks
• Run multiple queries in parallel when you have several questions

Good queries:

"How to create a new Go workload?"
"What is kbroker and how to publish events?"
"How does STS authentication work?"
"What are the logging best practices?"
"How to configure GitLab CI pipeline for a Go service?"
"What is the standard Dockerfile for TypeScript services?"
"How to set up container registry in GitLab?"

Bad queries:

"Tell me about architecture, kbroker, and how to create services"  # Too many topics
"Explain everything about the platform"  # Too broad

3. Interpret Results

The response comes from an LLM - extract key information and apply to your context.

4. Complement if Needed

After Kavak docs, use additional sources:

• Context7 for external library docs
• Web search for general patterns
• Codebase search for existing implementations

Topics & References

CI/CD & Docker Queries

When working with pipelines or containers, query these topics:

"What is the standard .gitlab-ci.yml template?"
"How to configure multi-stage Docker builds?"
"What are the required pipeline stages for production?"
"How to configure environment variables in CI?"
"What base images should I use for Go/TypeScript/Python?"
"How to set up automated testing in GitLab CI?"
"How to configure deployment to staging/production?"

Anti-Patterns

• Skipping Kavak docs and going straight to external sources
• Asking multiple questions in one query
• Writing overly long or vague queries
• Ignoring platform-specific patterns in favor of generic solutions
• Using generic Dockerfile templates instead of Kavak standards
• Copying CI configs from external projects without checking Kavak patterns

References