Integration Guide

What this scaffold already includes

• multi-stage site requirement collection
• base confirmation and product confirmation separation
• product source branching: upload / ds / none
• payload building
• executor mapping:
  • ecommerce -> generate_ecommerce_site
  • showcase -> generate_showcase_site
• mock generator client with preview URL

How to integrate into an existing repo

1. Copy files into your project

Recommended mapping:

• skills/site-builder/ -> your skills directory
• app/services/*.py -> your service/runtime directory
• tests/ -> your tests directory

2. Wire runtime entry

Your agent or API layer should instantiate:

1
2
runtime = SiteBuilderRuntime()
result = runtime.handle_message(session_id, user_text)

Then return:

• result["reply"]
• result["stage"]
• optional result["task"]
• optional result["build_job"]

3. Replace the mock store

Replace InMemorySiteBuilderStore with your DB-backed store if you need persistence.

4. Replace the mock generator client

Replace SiteGeneratorClient.submit_build_job() with your real job submission API.

5. Optional: strengthen extractor

The current extractor is rule-based. For production, replace it with:

• LLM structured extraction
• rule-based fallback

Known limitations

• SiteBuilderExtractor is intentionally lightweight
• style recommendation is static
• build payload shape may need adjustment to match your generator API
• no auth / tenancy / tracing included

Minimal runtime contract

handle_message(session_id, text) returns:

1
2
3
4
5
6
7
8
9
10
{
  "ok": true,
  "stage": "build_ready",
  "action": "build_ready",
  "state": {...},
  "site_blueprint": {...},
  "build_job": {...},
  "task": {...},
  "reply": "站点任务已提交，预览地址：https://..."
}