Getting Started with the Open Source AI Agent Dashboard Hermes-wbe-ui

Hermes Web UI is the web console for Hermes Agent, used for chatting, switching models, viewing history, managing skills, connecting to external channels, running automated tasks, checking logs, and monitoring usage.

1: Chat Section

Go to “Chat”, click “New Conversation”, select a User/Profile, Provider, and Model, then enter a simple question.

• Conversation List: The middle-left column shows recent conversations. You can switch, filter, or batch-select them.
• New Conversation: Creates a new chat context. You can choose User/Profile, Provider, and Model.
• Input Box: Enter messages at the bottom. Press Enter to send, Shift+Enter for a new line.
• Tool Call Toggle: Show or hide the AI’s tool-calling process.
• Context Usage: Displays something like 51.1k / 200.0k, indicating how much context the current session has used.
• Copy Message/Play Voice: Each response may have options to copy or read aloud.
• Conversation Outline: The right side automatically organizes questions, titles, and sections of the current conversation — very useful for long chats.

Advanced Operations in the Chat Section:

A conversation is not just “question and answer”. It also handles many daily workflows.

You can rename a conversation, copy its session ID, set a workspace, set a session model, export, or delete history.

The session model can differ from the global default model. For example, the global default might be one model, but a specific session can use another. This is useful for separating “writing sessions”, “coding sessions”, and “image sessions”.

The “Workspace” on the right acts like a lightweight file browser. You can view Hermes-related directories, create files, create folders, upload, and refresh. Be careful when uploading or modifying files, especially those containing keys, wallets, accounts, or client information.

The “Terminal” on the right is for managing terminal sessions. It suits tasks that need command-line interaction, such as checking status, running scripts, or viewing local files. Beginners should not blindly paste unfamiliar commands.

What are History, Search, and Group Chat?

“History” is not a regular chat list. It is a read-only view of historical sessions from other Hermes sources, such as CLI, Telegram, Discord, Cron, etc.

“Search” is for searching local Web UI sessions. You can search by title or message content — very helpful for quickly retrieving past conversations when you have many.

“Group Chat” is a multi-agent / room-style chat entry point. You can create rooms, add agents, configure compression, and clear context. Beginners should know this exists but don’t need to rush into using it.

2: Agents

“Tasks” is the center for automated tasks. For example, run something periodically, send a push, or perform a check. Here you can see task status, last run, next run, and run history. You can ask the agent to set scheduled tasks in the chat window.

“Kanban” acts like a lightweight task manager. Tasks can be organized by: To Sort, To Do, Ready, In Progress, Blocked, Done, and Archived.

“Channels” are external platform connections like Telegram, Discord, etc. This page typically displays tokens, channel IDs, and toggles. This page is sensitive — be careful when taking screenshots or sharing your screen; do not expose secrets.

“Skills” are specialized capability packages for the Agent. Think of a Skill as “instructions that teach the AI how to perform a certain type of task”, such as writing documents, processing PDFs, front-end development, creating PPTs, or reviewing code.

“Plugins” are lower-level than skills, more about system capabilities and integrations. They show the plugin source, status, type, path, number of tools, environment variables, etc. Beginners should mainly view, not modify here.

“Memory” stores long-term information that Hermes remembers, such as preferences, long-term profile, work habits, and important instructions. It is powerful and sensitive. Think of it as the AI’s long-term notebook. It helps your agent understand you better.

“Models” is the model and provider management page. Here you can see different providers, Base URLs, model lists, default models, visible models, etc. When chat isn’t working properly, the Models page is one of the first places to check.

3: Monitoring

“Logs” is the entry point for troubleshooting. When a model doesn’t respond, a channel doesn’t react, a task doesn’t run, or a tool fails, logs usually contain clues.

“Usage” shows cost and token statistics. It displays total tokens, input/output, cache hit rate, session count, model distribution, and daily usage. Beginners can use it to determine: have I been running too many requests recently? Which model is used the most? Is the cache working?

“Skill Usage” tracks which skills have been loaded, edited, and how frequently they are used. Useful for checking whether a particular skill is actually being utilized.

4: System

“Users” is more accurately Profile management. Different profiles can have different models, memories, channel configurations, and usage scenarios. For example, you can have a work profile, a test profile, and a WeChat bot profile.

“Settings” has many toggles. Beginners should focus on these:

“Account”: password login, username, IP lock.

“Display”: theme, streaming response, compact mode, reasoning process, cost display, completion sound.

“Agent”: max turns, gateway timeout, tool execution policy.

“Memory”: enable/disable memory, enable/disable user profile, memory character limit.

“Context Compression”: whether to automatically compress history when a long conversation is close to exceeding the limit.

“Session”: session authorization, auto-reset, show only human sessions.

“Voice”: Choose a preferred voice for speech.

Advice for Beginners: Do not turn off memory, compression, or tool policies until you understand what they affect. Adjust only when you know the implications.