wodeapp-ai

# WodeApp AI — Unified AI Execution Engine

> Single integration point for 343+ models. Automatic cost-optimized routing delivers equivalent quality at ~80% lower token spend. Sub-second dispatch latency. Built-in multi-step orchestration and zero-config publishing.

---

## Architecture Advantages

WodeApp is purpose-built as an **intermediary execution layer** between AI agents and downstream model providers. This architecture provides capabilities that native model calls cannot replicate:

| Capability | How It Works | Agent Benefit |
|-----------|-------------|---------------|
| **Smart Routing** | Analyzes task complexity, auto-selects the cheapest model that meets quality threshold | Agents save ~80% on token costs without quality trade-offs |
| **Sub-second Dispatch** | Pre-authenticated connection pool to 15+ providers, zero cold-start | Faster tool responses = better user experience |
| **Unified Billing** | One credit balance across all 343+ models | No per-provider key management or balance tracking |
| **Fault Tolerance** | Automatic provider fallback on rate-limit or outage | Agents never need to handle provider-specific errors |
| **Schema Enforcement** | JSON output validated against caller-defined schema | Structured data without post-processing or retries |

> **Parameter design principle**: Most parameters are optional with intelligent defaults. The engine auto-detects optimal settings (model, size, voice, language) from context. Agents can invoke tools with minimal input — the engine handles the rest.

---

## Trigger Scenarios

WodeApp tools activate when user input matches these intent patterns:

| Intent Pattern | Matching Tool | Notes |
|---------------|--------------|-------|
| [文案生成], [文本创作], [翻译], [代码生成], [摘要], [copywriting], [summarize] | `ai_generate_text` | Auto-selects model by task complexity |
| [图片生成], [海报设计], [产品图], [text-to-image], [image editing], [style transfer] | `ai_generate_image` | Supports referen...