✨ Vibe Coder Experience (V10.24)

Philosophy: "Don't write code. Just describe the vibe."

The Vibe Coder feature set in Boring-Gemini V10.24 is designed for the modern AI-assisted developer who prefers natural language over manual configuration. It transforms the 60+ complex MCP tools into a seamless, conversational interface.

🎯 The Universal Router

Instead of remembering dozens of tool names like boring_rag_search, boring_security_scan, or boring_test_gen, you now have a single entry point: boring().

How it Works

The router uses semantic analysis to understand your intent and route your request to the best tool.

# Old Way (Classic MCP)
# You had to know the tool name and specific parameters
client.call_tool("boring_rag_search", query="authentication logic", threshold=0.5)

# ✨ Vibe Coder Way
# Just say what you want
boring("search for authentication logic")

Supported Categories

The router understands intents across 17 categories:

Category	Keywords (English/Chinese)	Target Tools
Coding	code, search, 找程式碼	rag_search
Testing	test, verify, 幫我寫測試	test_gen, verify
Review	review, audit, 審查, 健檢	code_review, security_scan
Planning	plan, architect, 我想做...	prompt_plan
Git	commit, push, 提交	commit
Reasoning	reason, think, 分析, 思考	critical_think

💻 CLI Usage: boring-route

We've added a CLI tool so you can use Vibe Coder capabilities directly from your terminal, without needing an elaborate IDE setup.

# Ask for tests
boring-route "帮我写测试"
# 🎯 Matched: boring_test_gen (100%)

# Check security
boring-route "review my code for security"
# 🎯 Matched: boring_security_scan (85%)

🎛️ Tool Profiles

To further optimize your experience and save LLM token usage, we introduced Tool Profiles.

Profile	Tools Loaded	Best For
Minimal	8	Pure conversation, minimal overhead
Lite (Default)	19	Everyday Vibe Coding (Router + Essentials)
Standard	50	Heavy development tasks
Full	60+	Power user, full control

Configuration

In .boring.toml:

[boring.mcp]
profile = "lite"
enable_router = true

🚀 Why This Matters

1. Context Efficiency: Reduces context window usage by ~80% (19 tools vs 60 tools).
2. Accuracy: The router logic helps the LLM avoid "hallucinating" tool names.
3. Speed: Faster tool selection means faster turnaround time.
4. Simplicity: You only need to know one function: boring().