Row-Bot
An AI agent execution framework that provides large language models with a browser sandbox environment, visual feedback, and other toolchains. It supports models in verifying execution results after generating code, following a plan–act–observe loop architecture.
Core Facts
Timeline (last 90 days)
The test ran the model through the developer's self-built agent framework, Row-Bot.
Row-Bot v4.6.0 introduced resumable document ingestion capability, enabling processing of large-scale documents to continue from the breakpoint after an interruption
Row-Bot is an open-source project that is still actively soliciting community feedback
Row-Bot v4.7.0 reduces prompt overhead through on-demand loading of tools and skills, saving token costs in multi-tool and multi-skill scenarios
Row-Bot v4.8.0 added dynamic OpenCode transport discovery and stricter custom endpoint validation
Row-Bot v4.7.1 is a reliability patch that fixes issues with agent restart recovery, process detachment, workspace locking, Telegram startup, Docker inspection, and Ollama capability detection, and adds optional offline SenseVoice speech-to-text support
Row-Bot v4.6.0 added authenticated remote access and headless server mode, and strengthened Docker deployment
Row-Bot v4.8.0 added per-model inference control, stricter context validation for custom endpoints, and 64K "Auto" context support for Ollama
Row-Bot v4.7.0 introduced full context metering and rolling compression functionality
Row-Bot v4.5.0 added foundational constraint mechanisms including agent execution budgets, concurrency limits, and loop protection
13 more timeline events
All Facts (20)
Skills采用渐进式披露设计，平时只向模型提供技能的名称和描述，具体执行规则在需要时才加载
65%UnverifiedRow-Bot is able to resume from a saved state after restarting mid-task, rather than starting over from the beginning.
60%UnverifiedThe test ran the model through the developer's self-built agent framework, Row-Bot.
50%UnverifiedRow-Bot v4.6.0 introduced resumable document ingestion capability, enabling processing of large-scale documents to continue from the breakpoint after an interruption
50%UnverifiedRow-Bot v4.7.0 introduced full context metering and rolling compression functionality
50%UnverifiedRow-Bot v4.8.0 added per-model inference control, stricter context validation for custom endpoints, and 64K "Auto" context support for Ollama
50%UnverifiedRow-Bot v4.6.0 added authenticated remote access and headless server mode, and strengthened Docker deployment
50%UnverifiedRow-Bot v4.7.1 is a reliability patch that fixes issues with agent restart recovery, process detachment, workspace locking, Telegram startup, Docker inspection, and Ollama capability detection, and adds optional offline SenseVoice speech-to-text support
50%UnverifiedRow-Bot v4.8.0 added dynamic OpenCode transport discovery and stricter custom endpoint validation
50%UnverifiedRow-Bot v4.7.0 reduces prompt overhead through on-demand loading of tools and skills, saving token costs in multi-tool and multi-skill scenarios
50%UnverifiedRow-Bot is an open-source project that is still actively soliciting community feedback
50%UnverifiedRow-Bot v4.5.0 added foundational constraint mechanisms including agent execution budgets, concurrency limits, and loop protection
50%UnverifiedRow-Bot v4.8.0取消了自定义端点自动继承假设上下文窗口大小的默认行为，并将模型探测严格限定在被测试的模型范围内
50%UnverifiedRow-Bot v4.8.0为滚动压缩（rolling compaction）机制增加了Preflight（预检）、Recovery（失败恢复）、Validation（结果验证）和Persistence safeguards（持久化保护）四重保障
50%UnverifiedRow-Bot v4.8.0让每一个会话都能为当前使用的具体模型保留一个有效的推理选择
50%UnverifiedRow-Bot 是一款开源 AI 聊天客户端
50%UnverifiedRow-Bot v4.8.0 引入了 provider-aware reasoning controls（服务商感知的推理控制机制），让每个会话都能为当前使用的具体模型保留有效的推理选择
50%Unverifiedv4.8.0 取消了自定义端点自动继承假设上下文窗口大小的做法，避免内容被意外截断或请求失败
50%Unverifiedv4.8.0 中模型探测（model probes）被严格限定在被测试的模型范围内，不会跨模型污染配置信息
50%Unverifiedv4.8.0 为滚动压缩（rolling compaction）机制增加了 Preflight、Recovery、Validation 和 Persistence safeguards 四重保护措施，防止历史内容压缩造成数据丢失
50%