AI Skill Hub 强烈推荐：AI代码侦探 是一款优质的Agent工作流。AI 综合评分 8.0 分，在同类工具中表现稳健。如果你正在寻找可靠的Agent工作流解决方案，这是一个值得深入了解的选择。
AI代码侦探 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
AI代码侦探 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 方式一：pip 安装（推荐）
pip install ai-code-sherlock
# 方式二：虚拟环境安装（推荐生产环境）
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install ai-code-sherlock
# 方式三：从源码安装（获取最新功能）
git clone https://github.com/signupss/ai-code-sherlock
cd ai-code-sherlock
pip install -e .
# 验证安装
python -c "import ai_code_sherlock; print('安装成功')"
# 命令行使用
ai-code-sherlock --help
# 基本用法
ai-code-sherlock input_file -o output_file
# Python 代码中调用
import ai_code_sherlock
# 示例
result = ai_code_sherlock.process("input")
print(result)
# ai-code-sherlock 配置文件示例（config.yml） app: name: "ai-code-sherlock" debug: false log_level: "INFO" # 运行时指定配置文件 ai-code-sherlock --config config.yml # 或通过环境变量配置 export AI_CODE_SHERLOCK_API_KEY="your-key" export AI_CODE_SHERLOCK_OUTPUT_DIR="./output"
| Feature | Description |
|---|---|
| ⚙️ **Auto-Improve Pipeline** | Set a goal, run a script, let AI iterate autonomously up to N times |
| 🕵️ **Sherlock Mode** | Automated root-cause analysis with confidence scoring |
| ⚡ **Surgical Patching** | Exact SEARCH/REPLACE — never rewrites whole files |
| 🧠 **Multi-Model Engine** | Ollama (offline), any OpenAI-compatible API, ZennoPoster File Signal |
| 🗜️ **Context Compression** | 120k tokens → 4k without losing signal using AST skeleton extraction |
| 🗂️ **Error Map** | Persistent database of errors + confirmed solutions |
| 📉 **Log Compressor** | Smart log compression preserving 100% of errors and tracebacks |
| ⏱️ **Version Control** | Every patch backed up automatically, one-click restore |
| 🔄 **Universal File Reader** | Reads xlsx, parquet, numpy, pickle, HTML, JSON for AI context |
| 🌍 **i18n** | Full English / Russian UI, switch live in Settings |
| 🎨 **Themes** | 4 UI themes (Dark, Light, Monokai, Dracula) + custom accent color, font size |
| 🏗️ **Workflow Constructor** | Visual drag-and-drop AI agent pipeline builder — 50+ node types, unlimited complexity |
| 🤖 **AI Agent Nodes** | Code Writer, Reviewer, Planner, Tester, Orchestrator, Patcher, Image Gen/Analyst, and more |
| 🔀 **Automation Snippets** | If/Else, Loop, Switch, Variable Set, HTTP Request, Delay, Log, Notifications, JS Snippets |
| 🌐 **Browser Automation** | Launch, click, screenshot, profile management — all as visual nodes |
| 🖥️ **Desktop Automation** | Open programs, click by image, screenshot, run actions — all visual nodes |
| 📋 **Lists & Tables** | Project-scoped lists and tables with file-backed persistence and runtime access |
| 🔗 **Conditional Routing** | LLM router, conditional branches, fallback agents, retry with backoff |
| 🧩 **Skill Registry** | Built-in and custom skills injected into agent system prompts automatically |
| 🔄 **Infinite Loop Guard** | Auto-detects repeated errors in pipeline logs, kills process, rolls back, re-patches |
| 📖 **Companion Scripts** | Add read-only context scripts to pipeline — AI sees them but never runs or patches them |
| 🤝 **Consensus Engine** | Query multiple models simultaneously, pick the best answer |
| 📡 **Signal Monitor** | Real-time ZennoPoster folder watcher |
| 🧩 **New Project Wizard** | Create a project with correct language, files auto-generated by AI |
| ▶️ **Live Script Execution** | Run any script with real-time output streaming + interactive stdin |
---
pip install -r requirements.txt
```bash
```bash
Windows: C:\Users\<user>\.ai_code_sherlock\settings.json
Linux: ~/.ai_code_sherlock/settings.json
macOS: ~/.ai_code_sherlock/settings.json---
ollama serve
ollama pull deepseek-coder-v2 # best for code
ollama pull llama3.2
ollama pull codestral
ollama pull qwen2.5-coder:7b| Provider | Base URL | Model example |
|---|---|---|
| OpenAI | https://api.openai.com | gpt-4o |
| Anthropic (via proxy) | https://api.openai.com | claude-3-5-sonnet |
| Groq | https://api.groq.com/openai | llama-3.3-70b-versatile |
| Together AI | https://api.together.xyz | mixtral-8x7b |
| Mistral | https://api.mistral.ai | codestral-latest |
| Gemini (via proxy) | https://generativelanguage.googleapis.com/v1beta/openai | gemini-2.0-flash |
| LM Studio | http://localhost:1234 | any local model |
| Kobold.cpp | http://localhost:5001 | any local model |
The most powerful feature — runs your script, reads output, generates a patch, validates syntax, applies it, and repeats autonomously.
Goal: "achieve f1 > 0.85 on validation set"
Script: train_model.py
Strategy: SAFE_RATCHET
Max iter: 20高质量的开源AI工作流构建工具，具有较强的自动化和代码分析能力
AI Skill Hub 为第三方内容聚合平台，本页面信息基于公开数据整理，不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后，再部署至生产环境，并做好必要的安全评估。
✅ MIT 协议 — 最宽松的开源协议之一，可自由商用、修改、分发，仅需保留版权声明。
总体来看，AI代码侦探 是一款质量优秀的Agent工作流，在同类工具中具备一定竞争力。AI Skill Hub 将持续追踪其更新动态，建议收藏备用，结合自身场景选择合适时机引入使用。
| 原始名称 | ai-code-sherlock |
| Topics | aiautomationautonomouscode-analysiscode-assistant |
| GitHub | https://github.com/signupss/ai-code-sherlock |
| License | MIT |
| 语言 | Python |
收录时间：2026-06-29 · 更新时间：2026-06-29 · License：MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型，复制安装指令后粘贴到对应客户端