user-manual.md

User Manual

Welcome to LinX (灵枢)! This manual will guide you through using the platform to manage AI agents, submit tasks, and leverage the knowledge base.

Table of Contents

1. Getting Started
2. Dashboard Overview
3. Managing Agents
4. Task Management
5. Knowledge Base
6. Memory System
7. User Settings
8. Best Practices

Getting Started

First Login

1. Open your browser and navigate to the platform URL
2. Enter your username and password
3. Click "Sign In"

Default Admin Credentials (change immediately):

• Username: admin
• Password: admin123

Dashboard Overview

After logging in, you'll see the main dashboard with:

• Active Agents: Number of currently running agents
• Goals Completed: Total completed tasks
• Throughput: Tasks completed per hour
• Compute Load: System resource usage
• Task Distribution Chart: Visual breakdown of task types
• Recent Events: Timeline of recent activities

Managing Agents

Viewing Agents

Navigate to Workforce in the sidebar to see all your agents.

Each agent card shows:

• Agent name and type
• Current status (Working, Idle, Offline)
• Assigned skills
• Recent activity

Creating a New Agent

1. Click the "+ Add Agent" button
2. Choose a template or create custom:
   • Data Analyst: For data processing and analysis
   • Content Writer: For content generation
   • Code Assistant: For code-related tasks
   • Research Assistant: For research and information gathering
3. Configure agent settings:
   • Name: Give your agent a descriptive name
   • Description: Describe the agent's purpose
   • Skills: Select required skills
   • Resource Limits: Set CPU and memory limits
4. Click "Create Agent"

Agent Templates

Data Analyst:

• Skills: data_processing, sql_query, data_visualization
• Use for: Data analysis, report generation, database queries

Content Writer:

• Skills: text_generation, summarization, translation
• Use for: Article writing, content creation, document summarization

Code Assistant:

• Skills: code_generation, code_review, debugging
• Use for: Code development, bug fixing, code review

Research Assistant:

• Skills: web_search, document_analysis, information_extraction
• Use for: Research, information gathering, document analysis

Managing Agent Status

• Start Agent: Click the play button to activate an idle agent
• Pause Agent: Click the pause button to temporarily stop an agent
• Terminate Agent: Click the stop button to permanently stop an agent
• View Logs: Click "View Logs" to see agent activity

Agent Details

Click on an agent card to view detailed information:

• Overview: Agent configuration and status
• Activity Log: Recent actions and tasks
• Performance Metrics: CPU, memory, task completion rate
• Assigned Tasks: Current and completed tasks

Task Management

Submitting a Goal

1. Navigate to Tasks in the sidebar
2. Enter your goal in the text box
3. Click "Submit Goal"

Example Goals:

• "Analyze sales data from Q4 and create a summary report"
• "Write a blog post about AI trends in 2024"
• "Review the codebase and identify potential bugs"
• "Research competitors and create a comparison table"

Goal Clarification

If your goal is ambiguous, the system may ask clarification questions:

1. Review the questions
2. Provide answers
3. Click "Submit Answers"

Task Decomposition

The system automatically breaks down your goal into sub-tasks:

• View the task tree in the Task Timeline
• See dependencies between tasks
• Monitor progress for each sub-task

Task Flow Visualization

The Task Flow tab shows a visual graph of:

• Task nodes with status indicators
• Dependencies (arrows between tasks)
• Agent assignments
• Real-time progress updates

Status Indicators:

• 🔵 Blue: Pending
• 🟡 Yellow: In Progress
• 🟢 Green: Completed
• 🔴 Red: Failed

Monitoring Task Progress

• Timeline View: See tasks in chronological order
• Flow View: See task dependencies and relationships
• Details Panel: Click a task to see detailed information

Task Results

When a task completes:

1. Click on the completed task
2. View the result in the details panel
3. Download artifacts if available
4. Provide feedback (optional)

Knowledge Base

Uploading Documents

1. Navigate to Knowledge in the sidebar
2. Click "Upload Documents" or drag files to the upload zone
3. Supported formats:
   • Documents: PDF, DOCX, TXT, MD
   • Images: PNG, JPG (with OCR)
   • Audio: MP3, WAV (with transcription)
   • Video: MP4, AVI (audio extraction + transcription)
4. Wait for processing to complete

Document Processing

After upload, documents are:

1. Validated: File type and size checks
2. Processed: Text extraction, OCR, or transcription
3. Chunked: Split into manageable pieces
4. Indexed: Embedded and stored for search

Searching Documents

1. Enter search query in the search bar
2. Use filters:
   • Type: Filter by document type
   • Date: Filter by upload date
   • Tags: Filter by tags
3. View search results with relevance scores
4. Click a document to view details

Document Viewer

Click on a document to:

• View content
• See metadata (upload date, size, type)
• Download original file
• Edit access permissions
• Add tags
• Delete document

Access Control

Set document permissions:

• Public: All users can access
• Internal: Only authenticated users
• Confidential: Specific users/roles only
• Restricted: Admin only

Memory System

The platform uses a reset-era memory pipeline:

User Memory

• Long-term facts about the user
• Stores preferences, relationships, background, skills, goals, and important events
• Used for personalization and better future assistance
• Private to the owning user by default

Skill Proposals

• Agent-owned learned successful paths
• Stores reusable execution methods that worked in practice
• Requires review before becoming a published skill
• Scoped to the owning agent account

Knowledge Base

• Shared documents and reference knowledge
• Use the Knowledge Base for company or project documents
• Not mixed into long-term user memory

Browsing Memory

1. Navigate to Memory in the sidebar
2. Select a product surface:
   • User Memory
   • Skill Proposals
3. Use search to find specific memories
4. Filter by:
   • Product type
   • Date range
   • Tags
5. View memory details and relevance scores

Skill Proposal Review

Review learned successful paths:

1. Select a skill proposal
2. Click "审核" or the review action in the detail panel
3. Choose whether to publish, reject, or request revision
4. Add an optional review note
5. Confirm the review action

User Settings

Profile Settings

1. Click your avatar in the top right
2. Select "Settings"
3. Update:
   • Display name
   • Email
   • Password
   • Language preference
   • Theme (Light/Dark/System)

Notifications

Configure notification preferences:

• Task completion alerts
• Agent status changes
• System updates
• Email notifications

API Keys

Generate API keys for programmatic access:

1. Go to Settings → API Keys
2. Click "Generate New Key"
3. Copy and save the key securely
4. Set expiration date (optional)

Best Practices

Writing Effective Goals

Good Goals:

• ✅ "Analyze Q4 sales data and identify top 3 products"
• ✅ "Write a 500-word blog post about renewable energy"
• ✅ "Review pull request #123 and suggest improvements"

Poor Goals:

• ❌ "Do something with data"
• ❌ "Write stuff"
• ❌ "Fix things"

Tips:

• Be specific and clear
• Include context and constraints
• Specify desired output format
• Mention any deadlines

Organizing Knowledge

• Use descriptive filenames
• Add relevant tags
• Set appropriate access levels
• Keep documents up to date
• Remove outdated content

Agent Management

• Start with templates for common use cases
• Monitor resource usage to optimize performance
• Review agent logs regularly
• Terminate unused agents to free resources
• Update agent skills as needs change

Security

• Change default passwords immediately
• Use strong passwords (12+ characters)
• Enable two-factor authentication (if available)
• Review access permissions regularly
• Don't share credentials
• Log out when finished

Performance

• Limit concurrent tasks to avoid overload
• Use appropriate agent types for tasks
• Monitor system resources
• Archive old data periodically
• Optimize large documents before upload

Keyboard Shortcuts

• Ctrl/Cmd + K: Quick search
• Ctrl/Cmd + N: New agent
• Ctrl/Cmd + T: New task
• Ctrl/Cmd + U: Upload document
• Ctrl/Cmd + ,: Settings
• Esc: Close modal

Troubleshooting

Task Not Starting

• Check if agents are available
• Verify agent has required skills
• Check system resource limits
• Review task requirements

Document Upload Failing

• Check file size (regular files max 200MB, ZIP archives max 3GB)
• Verify file format is supported
• Check storage quota
• Try uploading smaller files

Agent Not Responding

• Check agent status
• Review agent logs
• Restart agent if needed
• Contact administrator

Search Not Working

• Check search query syntax
• Verify documents are indexed
• Try different keywords
• Check access permissions

Getting Help

• Documentation: Browse docs at /docs
• FAQ: Common questions at /faq
• Support: Email support@example.com
• Community: Join our Discord/Slack
• GitHub: Report issues on GitHub

Glossary

• Agent: AI worker that performs tasks
• Goal: High-level objective submitted by user
• Task: Sub-component of a goal
• Skill: Capability that an agent can perform
• Template: Pre-configured agent type
• Memory: Stored knowledge and experiences
• Knowledge Base: Document repository
• Embedding: Vector representation of text

Updates and Changelog

Check the Changelog for latest updates and new features.

Feedback

We value your feedback! Please:

• Report bugs on GitHub
• Suggest features via email
• Rate your experience
• Share success stories

Thank you for using LinX (灵枢)!