VetCopilot-DAPT v3.1 部署

包含目录: vetcopilot-app/ — 前端 (React/TypeScript + Vite + PWA) vetcopilot-backend/ — 后端 (FastAPI + SQLAlchemy + LCPS Client) vetnetv2.0/ — DAPT AI 引擎 (Qwen2.5-3B + LCPS 2.0 + FAISS RAG) deploy_dapt.sh — 一键部署脚本

快速启动: chmod +x deploy_dapt.sh && ./deploy_dapt.sh

手动启动:

启动 AI 引擎: cd vetnetv2.0 VET_MODEL=./vet-qwen3b-v3-dapt PORT=8000 python server_gpu.py
启动后端: cd vetcopilot-backend PORT=8080 python -m uvicorn app.main:app --host 0.0.0.0
启动前端 (开发): cd vetcopilot-app npm install && npm run dev

API 端点: GET :8000/health — AI 引擎健康检查 POST :8000/api/v1/diagnose — LCPS 诊断 (VetCopilot 标准接口) POST :8000/run — LCPS 诊断 (兼容旧接口) POST :8000/chat — 兽医自由问答 GET :8080/api/health — 应用健康检查 POST :8080/api/consultations/start — 开始问诊

模型下载:

从 ModelScope 下载 DAPT 模型权重

pip install modelscope python -c "from modelscope.hub.snapshot_download import snapshot_download; snapshot_download('WWSCA123/vet-qwen3b-v3-dapt-inference', local_dir='./vet-qwen3b-v3-dapt')"

下载 tokenizer (从 HF mirror)

python -c "from huggingface_hub import snapshot_download; import shutil, os; d=snapshot_download('Qwen/Qwen2.5-3B-Instruct', allow_patterns=['tokenizer*','vocab*','merges*','generation_config*']); [shutil.copy(os.path.join(d,f), f'./vet-qwen3b-v3-dapt/{f}') for f in os.listdir(d)]"