通过正反方对抗辩论 + AI 裁决的工作流，克服单一 LLM 的幻觉和盲目顺从问题。
--- ## ✨ Features - 🔄 **Multi-Persona Debate** — 正方支持、反方反驳、裁决方综合判断 - 🔀 **Parallel Execution** — 正反方并行生成，提升效率 - 🌊 **Streaming Output** — 实时流式输出，所见即所得 - 🎨 **Modern CLI** — 科技感 UI，丰富的颜色和视觉元素 - 🔌 **Multi-Provider** — 支持 DeepSeek、Gemini、DashScope (Qwen) - ⚙️ **Flexible Config** — 每个角色可独立配置不同的 Provider 和 Model - 🎯 **8 Model Combinations** — 交互模式提供 8 种预设模型组合，快速选择 - 📝 **Structured Input** — 交互模式支持问题+上下文文件的结构化输入 ## 🏗️ Architecture  ## 🚀 Quick Start ### Installation ```bash # Clone the repository git clone https://github.com/hrygo/dialecta.git cd dialecta # Build make build # Or install globally make install ``` ### Environment Setup ```bash # Set API keys (configure based on your providers) export DEEPSEEK_API_KEY="your-deepseek-api-key" export GEMINI_API_KEY="your-gemini-api-key" # or GOOGLE_API_KEY export DASHSCOPE_API_KEY="your-dashscope-api-key" ``` ### Usage ```bash # Analyze a file dialecta proposal.md # Read from stdin echo "我们应该启动 AI 创业项目" | dialecta - # Interactive mode (with structured input support) dialecta --interactive dialecta -i # Short form # Features: Model selection + Question + Optional context file # Quick access with Make commands make ui # Interactive mode with 8 model combinations make gemini # All Gemini (Pro, Con, Judge) make gemini-deepseek # Gemini Judge, DeepSeek debate make gemini-qwen # Gemini Judge, Qwen debate make deepseek-qwen # Gemini Judge, DeepSeek vs Qwen # Custom providers dialecta --pro-provider deepseek --con-provider dashscope --judge-provider gemini doc.md ``` ## ⚙️ Configuration ### Supported Providers | Provider | Environment Variable | Default Model | Description | | --------- | ----------------------------------- | ----------------------- | ----------------- | | DeepSeek | `DEEPSEEK_API_KEY` | `deepseek-chat` | DeepSeek Chat API | | Gemini | `GEMINI_API_KEY` / `GOOGLE_API_KEY` | `gemini-3-pro-preview` | Google Gemini API | | DashScope | `DASHSCOPE_API_KEY` | `qwen-plus` | Alibaba Qwen API | ### Default Role Configuration | Role | Provider | Model | Temperature | | ----------- | --------- | ----------------------- | ----------- | | Affirmative | DeepSeek | `deepseek-chat` | 0.8 | | Negative | DashScope | `qwen-plus` | 0.8 | | Adjudicator | Gemini | `gemini-3-pro-preview` | 0.1 | ### Interactive Mode Combinations When using `dialecta -i` or `make ui`, you can choose from 8 model combinations: | ID | Combination | Judge | Pro | Con | Description | |----|-------------|-------|-----|-----|-------------| | 1 | All Gemini | Gemini | Gemini | Gemini | Highest quality analysis | | 2 | Gemini Judge, DeepSeek Debate | Gemini | DeepSeek | DeepSeek | Gemini judgment + DeepSeek reasoning | | 3 | Gemini Judge, DeepSeek vs Qwen | Gemini | DeepSeek | Qwen | Mixed debate perspectives | | 4 | Gemini Judge, Qwen Debate | Gemini | Qwen | Qwen | Gemini judgment + Qwen debate | | 5 | All DeepSeek | DeepSeek | DeepSeek | DeepSeek | Unified DeepSeek experience | | 6 | DeepSeek Judge, DeepSeek vs Qwen | DeepSeek | DeepSeek | Qwen | DeepSeek judgment + mixed debate | | 7 | DeepSeek Judge, Qwen Debate | DeepSeek | Qwen | Qwen | DeepSeek judgment + Qwen debate | | 8 | All Qwen | Qwen | Qwen | Qwen | Unified Qwen experience | ### CLI Options ``` OPTIONS -pro-provider string Provider for affirmative (default "deepseek") -pro-model string Model for affirmative -con-provider string Provider for negative (default "dashscope") -con-model string Model for negative -judge-provider string Provider for adjudicator (default "gemini") -judge-model string Model for adjudicator -stream Enable streaming output (default true) -interactive Interactive input mode -i Interactive input mode (shorthand) ``` ## 📖 Examples ### Basic Usage ```bash # Analyze a business proposal dialecta business-plan.md # Quick test via stdin echo "公司应该全面采用远程办公模式" | dialecta - ``` ### Interactive Mode ```bash # Start interactive mode dialecta -i # Or use Make make ui # You'll see a menu to choose from 8 model combinations: # 🌟 Gemini Judge: # [1] All Gemini # [2] Gemini Judge, DeepSeek Debate # [3] Gemini Judge, DeepSeek vs Qwen # [4] Gemini Judge, Qwen Debate # ⚡ DeepSeek Judge: # [5] All DeepSeek # [6] DeepSeek Judge, DeepSeek vs Qwen # [7] DeepSeek Judge, Qwen Debate # 🔷 Qwen Judge: # [8] All Qwen ``` **Structured Input with Context File**: Interactive mode now supports two-step input for complex analysis: ``` # Step 1: Select model combination ▸ Enter your choice (1-8): 1 # Step 2: Enter your question ① Enter your question or instruction: (Press ENTER twice to finish) ▸ 请分析这份代码的优缺点 ▸ 重点关注性能和可维护性 ▸ # Step 3: Optionally provide a context file ② Enter context file path (optional, press ENTER to skip): ▸ ./internal/cli/input.go ✓ Question + Context file loaded • Question: 45 characters • Context file: ./internal/cli/input.go (5234 characters) ``` The program will automatically structure your input: ``` # 用户问题 请分析这份代码的优缺点 重点关注性能和可维护性 --- # 上下文文件：./internal/cli/input.go [File content...] ``` ### Multi-Provider Setup ```bash # Use DeepSeek for reasoning, Gemini for judgment dialecta --judge-provider gemini --judge-model gemini-3-pro-preview proposal.md # All Qwen models dialecta --pro-provider dashscope --con-provider dashscope --judge-provider dashscope doc.md # Mixed providers for diversity dialecta --pro-provider deepseek --con-provider dashscope --judge-provider gemini idea.txt ``` ### Production Workflow ```bash # High-quality analysis with specific models dialecta \ --pro-provider deepseek --pro-model deepseek-chat \ --con-provider dashscope --con-model qwen-max \ --judge-provider gemini --judge-model gemini-3-pro-preview \ important-decision.md ``` ## 🛠️ Development ### Prerequisites - Go 1.21+ - Make ### Build & Test ```bash # Format, lint, test, and build make all # Run tests make test # Run tests with coverage make cover # Build for current platform make build # Cross-platform build make build-all # Run linter make lint # Interactive mode with model selection make ui # Quick model combinations make gemini # All Gemini make gemini-deepseek # Gemini Judge + DeepSeek debate make gemini-qwen # Gemini Judge + Qwen debate make deepseek-qwen # Gemini Judge + DeepSeek vs Qwen # Show all available commands make help ``` ### Project Structure ``` dialecta/ ├── cmd/dialecta/ # CLI entry point ├── internal/ │ ├── cli/ # CLI components (UI, input, flags) │ ├── config/ # Configuration management │ ├── debate/ # Debate orchestration │ ├── llm/ # LLM client implementations │ └── prompt/ # Prompt templates ├── Makefile ├── go.mod └── README.md ``` ## 📊 Test Coverage | Package | Coverage | | ----------------- | -------- | | `internal/config` | 100% | | `internal/prompt` | 100% | | `internal/debate` | 70.8% | | `internal/cli` | 60.9% | | `internal/llm` | 40%* | \* LLM package requires integration tests with real API calls ## 🤝 Contributing Contributions are welcome! Please feel free to submit a Pull Request. 1. Fork the repository 2. Create your feature branch (`git checkout -b feature/amazing-feature`) 3. Commit your changes (`git commit -m 'feat: add amazing feature'`) 4. Push to the branch (`git push origin feature/amazing-feature`) 5. Open a Pull Request ## 📄 License This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details. ## 🙏 Acknowledgments - [DeepSeek](https://deepseek.com/) for their powerful AI models - [Google Gemini](https://ai.google.dev/) for their multimodal capabilities - [Alibaba DashScope](https://dashscope.aliyun.com/) for Qwen models ---Made with ❤️ by hrygo