中文 · English
The details of Yan live here, one topic per page. In-app usage notes are under Settings → Docs (Chinese only for now).
| Page | What it covers |
|---|---|
| Getting started | Quick start, use inside VS Code, opening index.html directly, environment variables, the local bridge |
| Storage and backup | What the storage folder holds, sharing one copy across browsers, moving it, backup export and import |
| Chat | Sessions and input (groups, context, quotes, side threads), generation and versions (follow-ups, continuing, branches), rendering (in-page visuals, themes) |
| Attachments and Files | Attachment parsing and limits, Files, output cards and preview, the audio player |
| Models and tools | Connecting Anthropic and ChatGPT subscription, usage and tokens left, thinking levels, tool calls and sources, ask_user |
| Using SiliconFlow in Yan | Pick SiliconFlow as the API and paste a key: three steps. The full illustrated guide is in SiliconFlow's official docs |
| Presets | Bundle a prompt, tools, MCP, a model and permissions into one set |
| Chat and Agent | The folder decides the mode, file tools, background commands that wake the model, the ledger and the Intent (constraints and the conclusion graph of a long job), steps, subagents, stability limits |
| Permissions and sandbox | The three levels Ask first / Auto-review / Full auto, and the strict and loose sandbox |
| MCP | Connecting external MCP servers, how they're handed to the model, secrets |
| Skills | SKILL.md in the Claude Code format: where it goes, how to write one, how the model reads it |
| Browser panel | The model's browser, docked in a panel on the right: setup, usage, how it works |
| Sandbox environment | A bundled Python and toolchains for the model, download mirrors |
| Memory | How memories are written, read and consolidated |
| Code structure and development | How the code is split, concatenation, tests, formatting and type checks |