Dashboard

The dashboard is a React web application that runs at http://localhost:3000 alongside the bot. It provides real-time monitoring, group management, a memory editor, a knowledge base, and task scheduling — all without touching the command line.

Prerequisite

The dashboard assets must be built before the backend can serve them. If you see a blank page, run npm run build:dashboard first. See Building & Running for details.

First login

Navigate to http://localhost:3000. You will see a login screen. Enter the value you set for DASHBOARD_ACCESS_CODE in your .env file. This sets a session cookie that persists across page reloads.

If DASHBOARD_ACCESS_CODE is not set, the dashboard is accessible without a password. This is fine on localhost but must not be used in a publicly accessible deployment.

Adding a group

Groups must be registered before the bot will respond to them. There are two ways to register a group.

Via the dashboard

1. Go to the Overview page.
2. Click Add Group.
3. Paste the Telegram group chat ID (a negative number like -1001234567890).

How to find your group chat ID

Add @userinfobot or @getmyid_bot to your group and send any message. The bot will reply with the chat ID.

4. Give the group a display name.
5. Click Register.

Via Telegram

If the bot is already in the group, you can auto-register it by sending the registration command from inside the group. Check the bot's help message for the exact command syntax.

TIP

The Telegram method is faster when the bot is already present in a group. The dashboard method is useful when you want to pre-register a group before adding the bot.

Configuring a group

Click on any registered group on the Overview page to open the Group Detail view. Here you can configure:

Persona

Select from built-in personas or enter a custom system prompt:

Persona	Description
Professional	Formal, precise responses suited for work environments
Creative	Expressive and imaginative, ideal for brainstorming
Concise	Short answers only — perfect for busy groups
Custom	Enter any system prompt you like

Custom personas are written to the group's GEMINI.md file (which you can also edit directly in Memory Studio).

Model

Override the default Gemini model for this specific group. Useful when you want one group to use a faster model and another to use a more capable one.

Trigger name

Override the @Name trigger for this group. Defaults to the global ASSISTANT_NAME set in .env. This lets different groups address the bot by different names.

Web search

Toggle whether the web search tool is available in this group. When enabled, the bot can search the web to answer questions about current events.

Fast path

Toggle direct API mode per group. When enabled, simple text queries bypass the container and respond in milliseconds instead of seconds.

WARNING

Disabling the fast path for a group means all queries go through the container, which takes 5–15 seconds per response. Only do this for groups that specifically need container-based code execution.

Memory Studio

The Memory Studio page provides a Monaco editor (the same editor used in VS Code) for directly editing the AI context files for each group:

• System prompt (GEMINI.md) — The instructions and persona that shape every response. Edit this to give the bot specialized knowledge or a custom personality.
• Conversation summary — A compressed summary of past conversations fed as context to reduce token usage.

To edit:

1. Click the folder icon to select a group.
2. Choose which file to edit (system prompt or summary).
3. Make your changes in the editor.
4. Press Save.

Changes take effect on the next message the bot receives in that group.

TIP

You can write rich Markdown in GEMINI.md. Include headings, bullet points, and code blocks — Gemini reads them as structured instructions.

Knowledge Base

The Knowledge page lets you upload documents to a per-group full-text search index (powered by SQLite FTS5).

To add a document:

1. Select a group from the dropdown.
2. Click Upload Document and select a .txt, .md, or .pdf file.
3. The document is chunked and indexed in the database automatically.

Once indexed, users in that group can query the knowledge base:

@Andy search the knowledge base for refund policy

The bot retrieves the most relevant chunks and uses them to answer the question.

Supported file types

Format	Notes
.txt	Plain text, indexed as-is
.md	Markdown, indexed as plain text
.pdf	Text is extracted before indexing

TIP

For best results, use focused documents rather than large generic ones. A 10-page product manual will produce more accurate results than a 500-page company handbook.

Scheduled Tasks

The Tasks page lets you create tasks that run a Gemini prompt on a schedule and send the response as a message to the assigned group.

Creating a task

1. Click New Task.
2. Select the target group.
3. Write the prompt (e.g., Summarize the top tech news today and share it in a friendly tone.).
4. Set the schedule using natural language or a cron expression.
5. Click Save.

Schedule formats

Natural language

every day at 8am
every Monday at 9:00
every weekday at 7:30am
every hour

Cron expression

0 8 * * *        # 8:00 AM daily
0 9 * * 1        # 9:00 AM every Monday
*/30 * * * *     # every 30 minutes

One-time

in 30 minutes
tomorrow at noon
in 2 hours

Timezone matters

Scheduled tasks run in the timezone set by TZ in your .env file. Make sure it matches your local time. See Configuration for details.

Monitoring tasks

The Tasks page shows:

• Next run time for each active task
• Last run time and status (success or error)
• Run log — click any task to see its execution history and the bot's output

Tasks can be paused, resumed, or deleted at any time without restarting the bot.