peteromallet/my-dataclaw-data

这是一个包含7亿token的AI代理会话日志数据集，记录了peteromallet使用不同AI模型（如GPT-5系列、Claude系列、GLM-5.1等）进行代码生成和对话的交互过程。数据集包含3125个会话，来自3个不同来源，共161个分片。每个会话记录包含会话ID、项目信息、使用的模型、git分支、时间戳、消息记录（包括用户请求和AI助手的响应、思考过程和工具使用情况）以及统计信息（如消息数量、token数量等）。数据集采用JSONL格式存储，路径和用户名已匿名化处理。

This is a dataset containing 0.7 billion tokens of AI agent conversation logs, recording interactions between peteromallet and various AI models (such as GPT-5 series, Claude series, GLM-5.1, etc.) for code generation and conversation tasks. The dataset includes 3,125 sessions from 3 different sources, divided into 161 shards. Each session record contains session ID, project information, model used, git branch, timestamps, message logs (including user requests, AI assistant responses, thinking processes, and tool usage), and statistics (such as message counts, token counts, etc.). The data is stored in JSONL format, with paths and usernames anonymized.