FluxiIA/smart_tools

Smart Tools Dataset是一个葡萄牙语合成数据集，用于训练语言模型在使用工具（函数调用）时的智能和上下文操作，包括从简单的工具调用到复杂的交互式工作流。主要特点包括：100%葡萄牙语，自然真实的对话，多种工具编排模式，包括与人类的交互，多态和上下文相关的工具。数据集包含多个子集，每个子集针对不同的训练目标，如顺序执行、并行执行、工具链、上下文累积、交互式工作流和完整集合。数据集适用于LLM的微调、推理能力的提升、多轮交互的改进，以及用于研究、比较工具调用策略和开发智能助手等应用。

Smart Tools Dataset is a synthetic Portuguese dataset for training language models in intelligent and contextual use of tools (function calling), ranging from simple tool invocations to complex interactive workflows. Key features include: 100% in Portuguese, natural and realistic conversations, multiple orchestration modalities, including human interaction, polymorphic and context-aware tools. The dataset contains several subsets, each tailored for different training objectives, such as sequential execution, parallel execution, tool chaining, context accumulation, interactive workflows, and a complete set. The dataset is suitable for LLM fine-tuning, improving reasoning capabilities, enhancing multi-turn interactions, and for research applications, comparing tool calling strategies, and developing smart assistants.