跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
开发与测试

agentcode

7 ücretsiz modelin en iyi yeteneklerini tek çatı altında toplayan akıllı coding agent'ı

mrcbrbn5361 · 已有 AI 审核记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add mrcbrbn5361/agentcode

技能说明


name: agentcode description: "Smart routing across 7 verified free AI models with system detection and multi-language support for OpenCode." license: MIT version: "0.0.2" author: "AgentCode Contributors" category: "coding-agent"

AgentCode - Smart Multi-Model Coding Agent

What It Does

AgentCode routes coding tasks to the best free AI model automatically, with system detection and multi-language support.

Features

v0.0.2 New Features

  • System Model Detection: Detect AI models available on your system
  • Interactive Setup Wizard: Choose models based on detected capabilities
  • Session Management: Persistent sessions with context preservation
  • Multi-Language Support: 14+ programming languages with templates
  • OpenClaw-Inspired Features: Dynamic model switching and progress tracking

Core Features

  • Smart task routing to optimal AI models
  • 7 verified free models with official sources
  • MIT licensed, production-ready

Models

ModelProviderBest ForContext
MiMo-V2.5XiaomiMultimodal1M
DeepSeek V4 FlashDeepSeekSpeed (126 tok/s)1M
Laguna S 2.1NVIDIATerminal1M
Ling-3.0-flashAlibabaEfficiency256K
North Mini CodeNVIDIALocal256K
Nemotron 3 UltraNVIDIAEnterprise1M

Install

mkdir -p ~/.config/opencode/skills/agentcode
curl -fsSL https://raw.githubusercontent.com/mrcbrbn5361/agentcode/main/SKILL.md -o ~/.config/opencode/skills/agentcode/SKILL.md

Quick Start

from agentcode import route_task, ModelType

# Basic routing
model = route_task("Create FastAPI endpoint")
print(model)  # ModelType.DEEPSEEK

# System detection
from detector import detect_system_models
models = detect_system_models()

# Interactive setup
from wizard import run_setup_wizard
preferences = run_setup_wizard()

API

route_task(task_description, has_image=False, context_size=0, is_local_only=False) -> ModelType

Routes a task to the optimal model.

get_model_info(model: ModelType) -> Dict

Gets model information (name, provider, strength).

detect_system_models() -> List[DetectedModel]

Detects AI models available on the user's system.

run_setup_wizard() -> UserPreferences

Runs interactive setup wizard for model selection.

get_session_manager() -> SessionManager

Gets session manager for persistent coding sessions.

detect_project_language(project_path) -> ProgrammingLanguage

Detects primary language of a project.

System Detection

AgentCode automatically detects:

  • CLI Tools: OpenAI, Anthropic, Ollama, llama.cpp, vLLM, LM Studio
  • IDE Extensions: VS Code Copilot, Cursor, Windsurf, JetBrains AI
  • Cloud APIs: OpenAI, Anthropic, Google, DeepSeek, Mistral, Groq
  • Local Models: Ollama models, LM Studio models

Multi-Language Support

LanguageExpert ModelLintersFormatters
PythonMiMo-V2.5flake8, pylint, mypyblack, autopep8
JavaScriptMiMo-V2.5eslint, jshintprettier
TypeScriptMiMo-V2.5tsc, eslintprettier
GoLaguna S 2.1golangci-lintgofmt
RustLaguna S 2.1clippyrustfmt
JavaNemotron 3 Ultracheckstyle, spotbugsgoogle-java-format
C#Nemotron 3 Ultradotnet formatdotnet format
C++Laguna S 2.1cppcheck, clang-tidyclang-format
PHPLing-3.0-flashphpcs, phpstanphp-cs-fixer
RubyLing-3.0-flashrubocoprubocop
SwiftMiMo-V2.5swiftlintswiftformat
KotlinMiMo-V2.5ktlint, detektktlint
ScalaNemotron 3 Ultrascalastylescalafmt
ShellLaguna S 2.1shellcheckshfmt

Routing Rules

  1. Image/Audio → MiMo-V2.5
  2. Docker/Bash → Laguna S 2.1
  3. Speed keywords → DeepSeek V4 Flash
  4. Context >256K → Nemotron 3 Ultra
  5. Local only → North Mini Code
  6. Default → Ling-3.0-flash

Session Management

from sessions import get_session_manager, get_model_switcher

# Create session
manager = get_session_manager()
session = manager.create_session("mimo", "Create API", "python")

# Switch models
switcher = get_model_switcher()
switcher.switch_model("deepseek")

# Get statistics
stats = manager.get_session_stats()

Configuration

Configuration saved to ~/.agentcode/config.json:

{
  "version": "0.0.2",
  "selected_models": {
    "primary": "mimo",
    "fallback": ["deepseek", "laguna"],
    "terminal": "laguna",
    "multimodal": "mimo",
    "local": "north"
  },
  "user_preferences": {
    "priority": "speed",
    "privacy_mode": false,
    "auto_fallback": true
  }
}

License

MIT