Code Semantic Service
Uses a local Ollama instance as the inference backend for code refactoring and semantic extraction. The server calls Ollama with a configurable host, model (default deepseek-coder), and timeout, and gracefully degrades to a rule-based engine when Ollama is unavailable.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
Code Semantic Service · 代码重构与语义提取
一个可自托管、也可直接调用的代码智能服务:代码重构 + 代码语义提取,同时提供 REST API、MCP Server、x402 付费接口三种接入方式。
核心逻辑不依赖任何前端框架,后端只有 FastAPI,推理走本地 Ollama,没有大模型时自动降级为规则引擎,保证服务永不 500。
在线试用
入口 | 地址 | 说明 |
交互页面 | 粘贴代码直接看重构结果 | |
MCP Server | Agent 可直接连接 | |
OpenAPI | https://ai-interview-coach-80785.app.workbuddy.host/openapi.public.json | 平台导入插件用 |
能力清单 | https://ai-interview-coach-80785.app.workbuddy.host/llms.txt | 给 Agent 读的纯文本说明 |
Related MCP server: FrameworkIA MCP Server
它解决什么问题
重构代码这件事,人做起来慢,用大模型直接做又不稳定:模型可能改掉业务逻辑、可能只给一段解释而不给完整代码、可能把缩进搞乱。
这个服务做了三件事来保证结果可用:
只输出代码,不夹带解释文字。提示词和输出清洗双管齐下,
refactor_code()返回的一定是可直接替换的代码。强制校验可运行性。Python 代码会过一遍
ast.parse,语法不合法就退回原代码并说明原因,绝不返回一段跑不起来的"重构结果"。没有模型也能干活。
fallback_refactor.py是纯规则引擎:== None→is None、清理未使用 import、统一缩进、补类型注解占位。机械改造语义等价,不会引入新 bug。
快速开始
方式一:MCP(推荐给 Agent)
任何支持 MCP 的客户端填这个地址即可,无需鉴权:
{
"mcpServers": {
"code-semantic": {
"type": "http",
"url": "https://ai-interview-coach-80785.app.workbuddy.host/mcp"
}
}
}提供的工具:
工具 | 作用 |
| 按任务重构代码,返回完整代码 |
| 提取函数/类/导入/调用的结构化摘要 |
| 按岗位 JD 生成面试题 |
| 5G NWDAF 网络分析查询 |
方式二:REST API
# 代码重构
curl -X POST https://ai-interview-coach-80785.app.workbuddy.host/api/code/refactor \
-H "Content-Type: application/json" \
-d '{"code":"def f(a):\n if a == None:\n return 0\n return a\n","language":"python"}'{
"result": "def f(a):\n if a is None:\n return 0\n return a\n",
"model": "deepseek-coder",
"language": "python",
"billable_units": 1,
"degraded": false,
"elapsed_ms": 842
}degraded=true 表示当前环境没有大模型、结果由规则引擎产出——服务不会假装模型可用。
方式三:x402 付费调用(Agent 原生付费)
不需要注册、不需要 API Key、不需要预充值。请求先拿到报价,再用钱包签名支付:
# 1) 首次请求会收到 402 + PAYMENT-REQUIRED 报价头
curl -i -X POST https://ai-interview-coach-80785.app.workbuddy.host/paid/refactor \
-H "Content-Type: application/json" -d '{"code":"x=1"}'
# 2) 带上 X-PAYMENT 头重试即可拿到结果
curl -X POST https://ai-interview-coach-80785.app.workbuddy.host/paid/refactor \
-H "Content-Type: application/json" \
-H "X-PAYMENT: <base64 支付载荷>" \
-d '{"code":"x=1"}'发现文件:/.well-known/x402
本地运行
python -m venv .venv && .venv/Scripts/activate # Windows
pip install -r requirements.txt
# 可选:接本地大模型(不接则走规则引擎)
ollama pull deepseek-coder
export OLLAMA_MODEL=deepseek-coder
python run.py # 默认 8000 端口,读 PORT 环境变量打开 http://127.0.0.1:8000/docs 看接口文档,或 http://127.0.0.1:8000/ 用交互页面。
环境变量
变量 | 默认 | 说明 |
| 8000 | 监听端口 |
| Ollama 地址 | |
| deepseek-coder | 模型名 |
| 120 | 推理超时(秒) |
| 1 | 模型不可用时是否降级到规则引擎 |
| 空 | 收款地址,留空则付费接口不接单 |
项目结构
core/
main.py FastAPI 入口(/refactor /semantic /health)
refactor.py 重构核心:提示词、Ollama 调用、输出清洗、AST 校验
code_semantic.py 语义提取:AST 优先,正则降级(支持多语言)
fallback_refactor.py 规则引擎:无模型时的等价改造
models.py Pydantic 请求/响应模型
adapters/ 各平台上架适配层(fal.ai / x402 / ClawMart / 扣子 / TapLaunch)
tests/ 接口级测试,19 项已知取巧与限制
不用 tree-sitter。非 Python 语言走正则降级,精度不如 AST,但零编译依赖,部署省心。要更高精度可自行接 tree-sitter。
降级模式不做语义理解。规则引擎只做机械等价的改造,重构质量远低于模型模式,所以响应里会明确标
degraded=true。x402 端点不做本地验签。签名验证交给链上/ facilitator,因为本地验签无法判断 nonce 是否已用,会漏掉重放攻击。
License
MIT
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
MCP-Native LLM Orchestration Agent
Stateless agentic tools over MCP: concept extraction, long-context, knowledge graph, planning.
Governed app access for AI agents: 1,000+ apps & 12,000+ tools via Code Mode MCP.
OCR, transcription, file extraction, and image generation for AI agents via MCP.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to search code by meaning, explore codebase structure, store and query knowledge with temporal facts, and read source code through a set of MCP tools.306 npm7MIT
- FlicenseNot gradedqualityDmaintenanceEnables AI coding assistants to manage persistent memory, search project context, generate tests, review architecture, and orchestrate workflows through 15 MCP tools, enhancing context-aware code generation.-
- FlicenseNot gradedqualityBmaintenanceEnables AI agents to access unified development tools including code generation, documentation synchronization, test case rendering, and architecture graph queries through a single MCP server.-
- FlicenseNot gradedqualityAmaintenanceEnables AI assistants to save, retrieve, and semantically search code snippets, generate language-specific style guides, and produce comprehensive documentation through discoverable MCP tools and multi-agent workflows.-