Skip to content

JiuwenSwarm is an intelligent AI Agent built on openJiuwen. It extends the powerful capabilities of large language models directly to your fingertips through various communication apps you use daily.

Python
6.9 KB
Raw Blame

JiuwenSwarm page overview (web)

Goal: Help you understand the web UI structure, where features live, and how to interact with them. This is a foundation for deeper use.

Simplified Chinese: 页面概览


Overall layout

The JiuwenSwarm web app uses a classic three-column layout—common in professional tools and dev platforms—so information stays organized and actions are easy to reach.

Layout at a glance

Area Position Main purpose
Left navigation Left edge Feature menu, version info
Main workspace Center Chat, task execution, controls
Right info panel Right edge Status, task list, system info

JiuwenSwarm web UI: three-column layout

Layout notes

  1. Responsive design: Regions resize with screen width.
  2. Collapsible sidebars: You can fold the left and right areas to give the center more room.
  3. Layered information: Core interaction is in the center; supporting details on the right.

Tip: A wide display (e.g. 1920×1080 or higher) works best.


Left navigation

The left bar is the main entry to features.

Menu items

Item What it is When to use it
Chat AI conversation: text in, multi-turn context Q&A, tasks, code help, day-to-day use
Agents Agent setup: switch personas, create custom agents, tune parameters When you need a different style or domain expert
Sessions Session list: history, restore context, new session When continuing or finding past chats
Team Team workspace management: create teams, view team directories, manage team sessions, share team resources (skills, artifacts, workspaces, etc.) Multi-user collaboration, resource sharing, team project management
Heartbeat Heartbeat / health: runtime checks, scheduled job signals To confirm the system and cron-style work are healthy
Scheduled tasks Cron-style jobs: create, edit, delete Recurring work (e.g. daily reports, reminders)
Skills Skills library: browse, install, configure extensions Extra capabilities (e.g. deep search, PPT)
Channels Outbound channels: Feishu, WeChat, Telegram, etc. Push AI messages to other apps
Harness Harness Package management: select Agent runtime mode, import/export extension packages, manage different version configurations Switching native/extended mode, customizing Agent capabilities
Configuration System and model settings Change behavior or switch models
Browser Browser automation When the agent should drive a real browser
Logs Application logs and traces Debug issues and audit what happened
Update App updates: check for new versions, download, install and restart When upgrading to the latest version

Version info

  • Where: bottom-left of the nav

JiuwenSwarm version in the web UI

  • What: current JiuwenSwarm version
  • Why: identify builds when reporting issues or checking compatibility

Note: Some advanced features need the right permissions or config.


Main workspace (center)

This is where most interaction happens.

1. Chat view

The primary surface for talking to JiuwenSwarm.

Input

  • Type in the text box and send (e.g. Enter, depending on settings).

What you see

  • Conversation history: full thread of messages
  • AI reply: final answer and intermediate steps when shown
  • Tool calls: tools the agent used and their results

JiuwenSwarm chat: messages, reply, and tools

2. Execution modes

JiuwenSwarm offers three execution modes. Pick the one that fits the task. Names in the app follow the UI (English strings are typically Planning mode, Performance mode, and Cluster mode in en.json).

Scope of this page: only switching modes in the app. It does not cover changing config files or environment variables; for that, see Configuration.

Mode (UI) How it works When to use it
Planning mode Breaks the request into steps and runs them in order Complex work, step-by-step review, when you need to confirm each step
Performance mode More flexible, can run parallel work Simpler tasks, faster responses, when parallel helps
Cluster mode Multi-agent: a leader coordinates specialists; subtasks in parallel; leader merges the output Large jobs (e.g. PPT, deep research) that need many roles

Switching modes

  • Choose the mode in the input area of the main chat.
  • Modes change how the agent plans and runs; in Cluster mode you can usually see how work is split and parallel work (as the UI shows).

Execution mode selector

3. Task control bar

While a task is running, you can often manually control the current run in the same area:

State Meaning What you can do (typical)
Running The model is replying to your input Wait; you can click Stop to terminate the current execution
Stopped Stopped; the same user instruction will not continue You can send a new input

💡 Tip: In Cluster mode, clicking Stop pauses the current execution and you can resume it.

Running (example)
Task running

Stopped (example)
Task stopped


Right info area

The right side shows system status and supporting context.

1. Top status row

Item What it shows
Connection Backend link state (connected / disconnected / reconnecting)
Language UI language (e.g. Chinese / English)
Theme Light / dark / system

Status and auxiliary area

2. Info panel

Block What it is Why it helps
Task list Current and queued work See progress and queue
Context compression Context compression state and compression ratio Understand long threads
Memory usage Current system memory consumption Monitor resource usage
Heartbeat System heartbeat status Monitor system health

Tip: The info panel content updates in real time; you usually do not need to refresh manually.


FAQ

Q: The page feels slow?
A: Check your network, clear cache for the site, or try refreshing the page.

Q: Connection shows “disconnected”?
A: Check that the backend is running and your network is OK.


Next steps: Quick start and Skills for more detail (Chinese: 快速开始, 技能).