Web fetching and search: headless page capture (web.fetch) plus multi-engine free search (web.search). 网页抓取与搜索:无头抓取网页内容(web.fetch)+ 多引擎免费搜索(web.search)。
5,032 Steam results · Checked
Agent-facing todo tool; drives the visible task row. 任务清单:展示 agent 的计划与执行进度,让每一步目标和状态清晰可见。
Agents: pick a channel, pair it, and get a local agent named after it. Multi-account = multi-bot (WeChat, Telegram, QQ, Feishu, Lark, Signal): each agent owns at most one bot. Message it there and it asks a few things, one at a time. Each agent keeps its o
Bridge Doubao chat via web session; exposes a compatible API on loopback after login. 通过网页会话桥接豆包对话,登录后在本地回环提供兼容接口。
Orchestrates parallel subagents to split tasks and aggregate results. 并行子代理编排:拆分任务派发给多个小助理并行执行,再汇总处理结果。
Search inside the Fatiao data directory by filename or content (file.search by name, file.grep by content, backed by the native fff index). 在Fatiao数据目录内按文件名/内容快速搜索(file.search 按名、file.grep 查内容,fff 原生索引)
Turn data into a real table: comparisons, inventories, budgets, trackers, exports. The deliverable is one .csv (RFC 4180, literal values, no formulas) you edit as a grid in the app and export to Excel or PDF. It also reads an existing .xlsx / .xls / .csv —
Write any document as Markdown: reports, memos, letters, proposals, specs, READMEs, summaries, articles. The deliverable is one .md file you edit in the app and export to Word or PDF. It also reads an existing .docx (text, tables, styles) and edits one in
Agent-facing ask-user HITL tool, discovered via its tools.json description (Describer). 向用户提问确认:通过选择题、自由输入或确认弹窗收集你的决定。
Extract text from PDFs/scans (pymupdf, marker-pdf). 从 PDF 与扫描件中提取文本(pymupdf、marker-pdf 等),补充 read_file 仅读文本层的盲区。
Typed judgment: state + questions in, choice / noul / score answers with probabilities out — no generated text. Engines: local model (default; no key, no network) or cloud Jev (your key). Use when candidates are already listed and only the judgment is miss