🔧 Fusion360 CAM × Ollama 本地大模型智能数控工艺推荐系统

Fusion360内置脚本 → 自动识别3D模型特征 → 本地FastAPI中转 → 本地Ollama大模型API → 多步工艺流程 + 刀路策略推荐 + 切削参数 → 自动弹窗展示

📖 Table of Contents

Overview
System Architecture
Features
Quick Start
Project Structure
API Reference
Built-in Knowledge Base
Configuration
Fusion360 Integration
Troubleshooting
Changelog
License

🎯 Overview

A production-ready CNC process recommendation system that bridges Autodesk Fusion360 CAM with a local Ollama LLM (通义千问 Qwen). The system provides AI-generated cutting parameters (tool, spindle speed, feed rate, depth of cut) for machining operations directly inside the Fusion360 CAM workspace.

Key Design Decisions

Decision	Rationale
🖥️ Local LLM via Ollama	No cloud API costs, data stays local, no network dependency
🔧 Configurable model	Default `qwen2.5:7b-instruct-q4_K_M`; can switch to any Ollama model
📐 Fixed output format	`Tool
🌡️ `temperature=0.1`	Minimizes hallucination; parameters are deterministic
📚 Embedded knowledge base	8 features × 14 materials with shop-floor-validated baseline params
⚡ Zero middleware	No Redis, no database, no Docker — Python + FastAPI only

🏗 System Architecture

┌──────────────────────────────────────────────────────────────┐
│                    Autodesk Fusion360                         │
│  ┌────────────────────────────────────────────────────────┐  │
│  │  fusion360_cam_ai.py (内置Python脚本)                    │  │
│  │  ┌─────────────┐  ┌──────────┐  ┌────────────────────┐ │  │
│  │  │ 下拉框选择   │  │ 查询按钮  │  │ 结果弹窗展示       │ │  │
│  │  │ 特征·材料·机床│  │ →调用API  │  │ 刀具|S|F|ap       │ │  │
│  │  └─────────────┘  └────┬─────┘  └────────────────────┘ │  │
│  └────────────────────────┼────────────────────────────────┘  │
└───────────────────────────┼───────────────────────────────────┘
                            │ HTTP POST /get_craft
                            ▼
┌──────────────────────────────────────────────────────────────┐
│              Windows 11 本机 (D:\CAM_CLOUD_API)               │
│  ┌────────────────────────────────────────────────────────┐  │
│  │  FastAPI 中转服务 (cam_cloud_api.py :8000)              │  │
│  │  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐ │  │
│  │  │ 1.参数校验    │→│ 2.拼装Prompt  │→│ 3.调用本地AI  │ │  │
│  │  │ 材料/特征枚举 │  │ 知识库前缀   │  │ Ollama API   │ │  │
│  │  └──────────────┘  └──────────────┘  └──────┬───────┘ │  │
│  └──────────────────────────────────────────────┼─────────┘  │
└─────────────────────────────────────────────────┼─────────────┘
                                                  │ HTTP (本地)
                                                  ▼
┌──────────────────────────────────────────────────────────────┐
│              Ollama 本地大模型服务 (端口11434)                 │
│  ┌────────────────────────────────────────────────────────┐  │
│  │  qwen2.5:7b-instruct-q4_K_M (temperature=0.1)            │  │
│  │  → 生成标准化切削参数 (OpenAI兼容API)                    │  │
│  └────────────────────────────────────────────────────────┘  │
└──────────────────────────────────────────────────────────────┘

✨ Features

🆕 v1.8 — OpenRouter/Fusion 3D 模型在线生成

🎨 OpenRouter Fusion 3D: 文本→3D + 图片→3D, 基于 OpenRouter API 调用 fusion 模型, AI 描述驱动几何生成
🔗 多后端支持: Hunyuan3D (腾讯, 免费20次/天) | Meshy (付费, 4K纹理) | OpenRouter/Fusion (通用, 按量计费)
🖼️ 图片→3D: 上传图片自动分析形状/尺寸/材质并生成对应3D模型
⚡ 运行时热切换: /admin/api/ai_3d/{backend}/key 运行时配置 API Key, 无需重启
📦 自动导入 Fusion360: 生成的模型自动导入当前设计文档

🆕 v1.3 — AI-Powered Auto Process Planning

🔍 Auto Model Feature Detection: Scans Fusion360 BRep geometry, detects planes/holes/pockets/bosses/chamfers automatically (no manual selection needed)
🤖 AI-Generated Full Process Plan: Multi-step process plan with toolpath strategies, cutting parameters for every detected feature
🗺️ Toolpath Strategy Knowledge Base: 8 strategy categories (face milling, pocket, contour, drill cycles, surface finish, adaptive, threading, chamfer) referencing open-source CAM projects (FreeCAD Path, OpenCAMLib, FabexCNC, LinuxCNC)
🔧 Tool Material Knowledge Base: 12 tool material types with hot hardness, coatings, and application guidance
📤 Personal Craft Library: Upload/manage your own shop-validated process parameters, persisted to personal_craft_library.json

Core

🔗 End-to-end pipeline: Fusion360 script → local FastAPI → cloud AI → formatted parameters
🎯 Fixed output format: 工序号. 工序名称 | 刀路策略 | 特征 | 刀具型号 | S转速 | F进给 | ap切深 | 备注
📚 Built-in knowledge base: 8 machining features × 11 materials with validated baseline data
🌡️ Low-temperature inference: temperature=0.1 ensures deterministic, stable outputs
🔒 Offline fallback: Knowledge base lookup works without internet / API calls

Developer Experience

📝 Fully commented code — every function, class, and complex block
🧹 Clean variable naming — craft_params, VALID_MATERIALS, CRAFT_KNOWLEDGE_BASE
🛡️ Input validation — Pydantic models with enum constraints
📊 Structured logging — timestamped, level-tagged log output
🔍 Health check endpoint — API connectivity verification

Operations

🚀 One-click Windows startup — start_service.bat with auto-dependency check
🤖 Auto-start support — VBScript for silent boot-time launch
📖 Auto-generated API docs — Swagger UI at http://127.0.0.1:8000/docs

🚀 Quick Start

Prerequisites

Windows 11 (primary target; works on Win10+)
Python 3.10+ (download)
Ollama 本地大模型服务 (下载安装)
Ollama 模型 (已拉取, 如 ollama pull qwen2.5:7b-instruct-q4_K_M)
Autodesk Fusion360 with CAM workspace

1. Install Ollama & Pull Model

# 下载安装 Ollama: https://ollama.com/download
# 安装完成后, 拉取所需模型:
ollama pull qwen2.5:7b-instruct-q4_K_M
# 或其他 qwen 版本:
# ollama pull qwen2.5:14b
# ollama pull qwen2.5:32b

# 启动 Ollama 服务 (通常开机自启, 也可手动启动):
ollama serve

2. Install Python Dependencies

mkdir D:\CAM_CLOUD_API
cd D:\CAM_CLOUD_API

# Clone or copy all project files into this directory, then:
pip install -r requirements.txt

3. Configure Ollama

Edit start_service.bat lines 18-19 (可选, 默认即可):

set OLLAMA_BASE_URL=http://127.0.0.1:11434/v1
set OLLAMA_MODEL=qwen2.5:7b-instruct-q4_K_M

4. Start the Service

# Double-click:
start_service.bat

# Or command line:
python cam_cloud_api.py

You should see:

模型: qwen2.5:7b-instruct-q4_K_M | 温度: 0.1 | 端口: 8000
Ollama地址: http://127.0.0.1:11434/v1
Uvicorn running on http://0.0.0.0:8000

5. Verify

# Health check
Invoke-RestMethod http://127.0.0.1:8000/health

# Test AI query
$body = '{"feature":"平面铣削","material":"6061铝","machine":"三轴立式加工中心"}'
Invoke-RestMethod -Uri http://127.0.0.1:8000/get_craft -Method Post -Body $body -ContentType "application/json"

6. Run in Fusion360

Tools → Scripts & Add-Ins → Create → Python
Name it CAM_AI_Craft
Paste fusion360_cam_ai.py contents
Save → Run

📁 Project Structure

D:\CAM_CLOUD_API\
├── cam_cloud_api.py              # FastAPI relay service (core backend) v1.8
├── fusion360_cam_ai.py           # Fusion360 Python script (client) v1.7
├── static/                       # 管理后台 Web UI (index.html)
│   └── index.html
├── personal_craft_library.json   # 个人工艺库 (持久化存储, 自动创建)
├── machine_registry.json         # 机床注册表 (持久化, 运行时可增删)
├── generated_models/             # AI 3D 生成模型输出目录 (v1.5)
├── requirements.txt              # Python dependency manifest
├── start_service.bat             # Windows one-click launcher
├── start_service.ps1             # PowerShell launcher
├── auto_start.vbs                # Silent auto-start script for boot
├── test_api.py                   # Quick API diagnostic script
├── DESIGN.md                     # 系统设计文档 (架构/数据流/接口/已知问题)
├── README.md                     # This file
├── CHANGELOG.md                  # Version history
├── LICENSE                       # MIT License
└── .gitignore                    # Git ignore rules

📡 API Reference

Base URL

http://127.0.0.1:8000

`POST /get_craft`

Generate AI-recommended cutting parameters.

Request Body:

{
  "feature": "平面铣削",
  "material": "6061铝",
  "machine": "三轴立式加工中心"
}

Field	Type	Required	Options
`feature`	string	✅ Yes	`平面铣削`, `型腔加工`, `键槽加工`, `钻孔`, `攻丝`, `曲面精加工`
`material`	string	✅ Yes	`6061铝`, `45#钢`, `304不锈钢`, `H62黄铜`
`machine`	string	No (default: `三轴立式加工中心`)	`三轴立式加工中心`, `数控铣床`, `钻攻中心`, `五轴加工中心`, `龙门铣床`

Response 200 OK:

{
  "craft_params": "Φ63端铣刀(5刃) | S6000 | F1200 | ap1.0",
  "feature": "平面铣削",
  "material": "6061铝",
  "machine": "三轴立式加工中心",
  "status": "ok"
}

Error Responses:

Code	Detail	Cause
`400`	不支持的材料: 'xxx'	Material not in knowledge base
`400`	不支持的特征: 'xxx'	Feature not in knowledge base
`404`	模型未找到	Ollama模型未拉取, 运行 `ollama pull`
`503`	无法连接到Ollama	Ollama服务未启动
`500`	AI服务调用失败	SDK / 推理异常

`GET /knowledge_base`

List full knowledge base (no API call).

`GET /knowledge_base/lookup?feature=X&material=Y`

Look up baseline parameters offline.

`POST /auto_craft` 🆕 v1.3

AI-powered full process planning from detected model features.

Request Body:

{
  "features": [
    {"feature_type": "平面", "name": "顶面", "dimensions": "约100×80mm", "count": 1, "area_mm2": 8000},
    {"feature_type": "通孔", "name": "Φ10通孔", "dimensions": "Φ10×深约25mm", "count": 4, "diameter": 10, "depth": 25}
  ],
  "material": "6061铝",
  "machine": "三轴立式加工中心",
  "part_name": "底板",
  "overall_dimensions": "100×80×30mm"
}

Response 200 OK:

{
  "process_plan_text": "工艺总览...\n---\n1. 平面铣削 | 面铣之字形 | 顶面 | ...",
  "steps": [
    {"step": 1, "operation": "平面铣削", "toolpath_strategy": "面铣之字形", "feature_ref": "顶面", "tool": "Φ63端铣刀(5刃)", "spindle_speed": "6000", "feed_rate": "1200", "depth_of_cut": "1.0", "note": "粗铣顶面"}
  ],
  "features_detected": 2,
  "material": "6061铝",
  "machine": "三轴立式加工中心",
  "status": "ok"
}

`POST /craft_library/upload` 🆕 v1.3

Upload custom process parameters to personal library (persisted as JSON).

`GET /craft_library/query?material=6061铝&feature=钻孔` 🆕 v1.3

Search personal craft library with optional filters.

`GET /health`

Service health + API connectivity check.

Interactive Docs

Visit http://127.0.0.1:8000/docs for the auto-generated Swagger UI.

📚 Built-in Knowledge Base

Material	Face Milling	Pocket	Keyway	Drilling	Tapping	Surface Finish
6061 Al	Φ63端铣刀 S6000 F1200 ap1.5	Φ12立铣刀 S8000 F1500 ap1.0	Φ8键槽铣刀 S5000 F800 ap0.5	Φ6麻花钻 S4000 F300	M6丝锥 S800 F800	R5球刀 S10000 F2000 ap0.2
7075 Al	Φ63端铣刀 S5500 F1100 ap1.5	Φ12立铣刀 S7500 F1400 ap1.0	Φ8键槽铣刀 S4800 F750 ap0.5	Φ6麻花钻 S3800 F280	M6丝锥 S750 F750	R5球刀 S9500 F1900 ap0.2
45# Steel	Φ63端铣刀 S2500 F500 ap1.0	Φ12立铣刀 S3500 F600 ap0.5	Φ8键槽铣刀 S2500 F400 ap0.3	Φ6麻花钻 S1800 F150	M6丝锥 S300 F300	R5球刀 S5000 F800 ap0.15
40Cr Alloy	Φ63端铣刀 S1800 F400 ap0.8	Φ12立铣刀 S2800 F500 ap0.4	Φ8键槽铣刀 S2200 F350 ap0.25	Φ6含钴钻 S1500 F120	M6丝锥 S250 F250	R5球刀 S4200 F650 ap0.12
Cr12MoV Die	Φ63端铣刀 S1000 F250 ap0.5	Φ12立铣刀 S2000 F350 ap0.3	Φ8键槽铣刀 S1500 F220 ap0.2	Φ6含钴钻 S1000 F80	M6丝锥 S180 F180	R5球刀 S3000 F500 ap0.1
304 SS	Φ63端铣刀(AlTiN) S1200 F250 ap0.5	Φ12立铣刀(AlTiN) S2000 F300 ap0.3	Φ8键槽铣刀(AlTiN) S1500 F200 ap0.2	Φ6含钴钻 S800 F80	M6含钴丝锥 S150 F150	R5球刀(AlTiN) S3500 F500 ap0.1
316 SS	Φ63端铣刀(AlTiN) S1000 F200 ap0.4	Φ12立铣刀(AlTiN) S1700 F250 ap0.25	Φ8键槽铣刀(AlTiN) S1300 F170 ap0.15	Φ6含钴钻 S650 F65	M6含钴丝锥 S120 F120	R5球刀(AlTiN) S3000 F430 ap0.1
HT250 Cast Iron	Φ63端铣刀 S1500 F400 ap1.0	Φ12立铣刀 S2500 F500 ap0.5	Φ8键槽铣刀 S1800 F300 ap0.3	Φ6硬质合金钻 S2000 F200	M6丝锥 S350 F350	R5球刀 S4000 F700 ap0.15
TC4 Titanium	Φ63端铣刀(AlTiN) S350 F100 ap0.3	Φ12立铣刀(AlTiN) S800 F150 ap0.2	Φ8立铣刀(AlTiN) S600 F100 ap0.15	Φ6硬质合金钻(内冷) S400 F40	M6丝锥 S80 F80	R5球刀(AlTiN) S1800 F350 ap0.08
Inconel 718	Φ63端铣刀(AlTiN) S200 F60 ap0.2	Φ12立铣刀(AlTiN) S500 F100 ap0.15	Φ8立铣刀(AlTiN) S400 F70 ap0.1	Φ6硬质合金钻(内冷) S250 F25	M6丝锥 S50 F50	R5球刀(AlTiN) S1200 F250 ap0.06
H62 Brass	Φ63端铣刀 S5000 F1000 ap1.5	Φ12立铣刀 S7000 F1200 ap1.0	Φ8键槽铣刀 S4000 F600 ap0.5	Φ6麻花钻 S3500 F250	M6丝锥 S600 F600	R5球刀 S8000 F1500 ap0.2

💡 These are shop-floor-validated starting parameters. Always adjust based on your specific setup, tool holder, coolant, and rigidity conditions. 💡 v1.3 adds tool material knowledge (HSS→PCD, 12 types), toolpath strategy knowledge (8 categories from open-source CAM projects), and CAM Assist-style UI.

⚙️ Configuration

Parameter	Location	Default	Description
`OLLAMA_BASE_URL`	`cam_cloud_api.py:48` or env var	`http://127.0.0.1:11434/v1`	Ollama OpenAI兼容API地址
`OLLAMA_MODEL`	`cam_cloud_api.py:49` or env var	`qwen2.5:7b-instruct-q4_K_M`	Ollama模型名称
`TEMPERATURE`	`cam_cloud_api.py:53`	`0.1`	LLM sampling temperature
`MAX_TOKENS`	`cam_cloud_api.py:54`	`200` (600 for auto_craft)	Max output tokens
`TOP_P`	`cam_cloud_api.py:55`	`0.1`	Nucleus sampling parameter
`HUNYUAN3D_API_KEY`	env var	(空)	腾讯 Hunyuan3D API Key — 文本/图片→3D, 免费20次/天
`MESHY_API_KEY`	env var	(空)	Meshy API Key — 付费3D生成, 4K纹理
`OPENROUTER_API_KEY`	env var	(空)	OpenRouter API Key — 通用AI网关, 调用 `fusion` 3D模型
`OPENROUTER_3D_MODEL`	env var	`fusion`	OpenRouter 3D 模型名称
Host/Port	`cam_cloud_api.py` main / env `HOST`/`PORT`	`127.0.0.1:8000`	FastAPI server binding (设 `HOST=0.0.0.0` 开放局域网)
`API_BASE_URL`	`fusion360_cam_ai.py:39`	`http://127.0.0.1:8000`	Fusion360 → API endpoint
`PERSONAL_LIBRARY_FILE`	auto: `personal_craft_library.json`	same dir as API	Personal craft library persistence

🖥 Fusion360 Integration

Script Location

Fusion360 scripts are stored at:

%APPDATA%\Autodesk\Autodesk Fusion 360\API\Scripts\CAM_AI_Craft\

Installation Steps

In Fusion360: Tools → Scripts & Add-Ins
Click Create → Python
Name: CAM_AI_Craft
Replace default content with fusion360_cam_ai.py
Save & Run

UI Overview

┌──────────────────────────────────────────┐
│  ⭐ 云端AI工艺参数推荐系统 ⭐ v1.3       │
├──────────────────────────────────────────┤
│  加工特征: [▼ 平面铣削              ]    │
│  工件材料: [▼ 6061铝                ]    │
│  机床类型: [▼ 三轴立式加工中心      ]    │
├──────────────────────────────────────────┤
│  🆕 v1.3 自动工艺规划                    │
│  [🔍 自动识别模型特征]  ← 扫描3D几何体   │
│  [🤖 AI生成完整工艺流程] ← 多步工序+策略 │
├──────────────────────────────────────────┤
│  📊 检测结果: 平面×1 / 通孔×4 / 倒角×2  │
│  📋 AI工艺:                              │
│  1. 平面铣削 | 面铣之字形 | S6000 F1200  │
│  2. 钻孔 | G83啄钻 | S3500 F250 ...      │
├──────────────────────────────────────────┤
│  [🔍 查询工艺参数]     ← 单步手动 (收费) │
│  [📖 查看知识库基准]    ← 离线 (免费)    │
└──────────────────────────────────────────┘

🔧 Troubleshooting

Symptom	Probable Cause	Fix
`ModuleNotFoundError: openai`	Missing dependency	`pip install -r requirements.txt`
Ollama connection refused	Ollama not running	Run `ollama serve` first
Model not found error	Model not pulled	`ollama pull qwen2.5:7b-instruct-q4_K_M`
Connection refused in Fusion360	Service not running	Double-click `start_service.bat` first
Port 8000 occupied	Another process	`netstat -ano \| findstr :8000` → `taskkill /PID N /F`
Port 11434 occupied	Ollama already running	Normal — Ollama runs on this port
Script error in Fusion360	Paste truncated	Re-copy full `fusion360_cam_ai.py`
AI output malformed	Rare model variance	Use offline KB button; or re-query
Firewall blocks Python	Windows Defender	Allow `python.exe` in firewall settings

Full diagnostic check:

# 1. Python version
python --version

# 2. Ollama status
ollama list

# 3. Dependencies
python -c "import fastapi, uvicorn, openai, pydantic; print('OK')"

# 4. Service health
Invoke-RestMethod http://127.0.0.1:8000/health

# 5. AI connectivity
Invoke-RestMethod http://127.0.0.1:8000/get_craft -Method Post `
  -Body '{"feature":"平面铣削","material":"6061铝","machine":"三轴立式加工中心"}' `
  -ContentType "application/json"

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

🙏 Acknowledgments

Ollama — 本地大模型运行平台
Qwen (通义千问) — 阿里云开源大模型系列
Autodesk Fusion360 — Python API for CAM automation
FastAPI — High-performance Python web framework
Uvicorn — Lightning-fast ASGI server

Built with ❤️ for CNC machinists and CAM programmers

FilesExpand file tree

README.md

Latest commit

History