Skip to main content

Creating Conversations

This guide explains how to create conversation-type tasks in the Wegent frontend to interact with AI agents.

πŸ“‹ Table of Contents​

πŸ’¬ What is a Conversation​

In Wegent, a conversation is the primary way users interact with AI agents. Each conversation creates a Task that records the complete conversation history.

Core Concept:

Conversation = User Message + Agent + Context + Conversation History

Conversation Components​

ComponentDescriptionExample
MessageUser's question or instruction"Help me analyze this report"
AgentAI team executing the taskData Analysis Agent
ContextAttached knowledge bases, files, etc.Project docs, data files
StatusConversation execution statusPENDING β†’ RUNNING β†’ COMPLETED

πŸš€ Creating a Conversation​

Step 1: Navigate to Chat Page​

  1. Click "Chat" in the left navigation bar
  2. The system displays the conversation list and input area

Step 2: Select an Agent​

Above the input area, click the agent selector:

  1. Click the agent dropdown - Shows available agents
  2. Select an appropriate agent

Step 3: Configure Conversation Options (Optional)​

Configure the following options as needed:

Model Selection​

Click the model selector to override the agent's default model:

  • Select model: Choose from the dropdown list
  • Force override: When enabled, uses your selected model even if the agent has a configured model

Code Repository (Code Type)​

If you selected a Code type agent:

  1. Select repository: Click the repository selector, choose the target repository
  2. Select branch: Choose the branch to work on

Knowledge Base Context​

Click the context button to add knowledge bases:

  1. Click the "#" button - Opens the context selector
  2. Select knowledge bases - Check the ones to add
  3. Confirm selection - Knowledge bases appear as tags

File Attachments​

Click the attachment button to upload files:

  1. Click the attachment icon - Opens file selector
  2. Select files - Supports images, documents, code files
  3. Wait for upload - Files show preview after upload

Skill Selection​

If the agent supports skills:

  1. Click the skill button - Opens skill selector
  2. Select skills - Check the needed skills
  3. Or use "/" command - Type / in the input box to trigger skill selection

Step 4: Enter Message and Send​

  1. Type your message in the input box - Describe your needs
  2. Press Enter to send - Or click the send button
  3. Wait for response - Agent starts processing and streams results

πŸ–₯️ Conversation Interface​

Input Area​

The input area contains the following controls:

ControlFunctionLocation
Agent SelectorSelect the agent for the taskAbove input box
Model SelectorOverride default modelControl bar
Context ButtonAdd knowledge basesControl bar
Attachment ButtonUpload filesControl bar
Skill ButtonSelect skillsControl bar
Clarification ButtonEnable clarification modeControl bar
Correction ButtonEnable correction modeControl bar
Send ButtonSend messageRight side of input box

Message Area​

The message area displays conversation history:

  • User Messages: Your sent messages, displayed on the right
  • AI Responses: Agent's replies, displayed on the left
    • Text content
    • Code blocks (with syntax highlighting)
    • Thinking process (if enabled)
    • Tool call indicators

The left sidebar displays:

  • Conversation List: All historical conversations
  • Search Box: Search conversation content
  • New Chat Button: Start a new conversation

⚑ Advanced Features​

Per-Conversation Model Selection​

Select a different model for a single conversation without modifying agent configuration:

  1. Click the model selector - In the input control bar
  2. Select a model - From the available model list
  3. Enable force override - Ensures the selected model is used

Use Cases:

  • Use a more powerful model for complex tasks
  • Use a faster/cheaper model for simple queries
  • Test different models' effectiveness

Clarification Mode​

Enable clarification mode for the agent to confirm requirements before execution:

  1. Click the clarification icon (πŸ’¬) - In the input control bar
  2. Agent asks questions first - Confirms your requirement details
  3. After answering questions - Agent starts executing the task

πŸ“– For detailed information, see Clarification Mode Guide

Correction Mode​

Enable correction mode to have another AI model verify and improve responses:

  1. Click the correction icon (βœ“) - In the input control bar
  2. Select correction model - Choose from the popup dialog
  3. View evaluation results - Scores and improvement suggestions appear after agent responds
  4. Apply improvements (optional) - Click "Apply" button to adopt the improved version

πŸ“– For detailed information, see Correction Mode Guide

File Attachments​

Upload files to provide context:

  1. Click the attachment button (πŸ“Ž)
  2. Select files - Supports multiple formats
  3. Supported formats: Images, PDF, Word, code files
  4. Paste upload: Directly paste images from clipboard

Knowledge Base Context​

Add knowledge bases to enhance agent capabilities:

  1. Click the context selector
  2. Select knowledge bases - Multiple selection supported
  3. Agent searches knowledge bases - Provides more accurate answers

Skill Selection​

Add additional capabilities to the agent:

  1. Click the skill button - Opens skill selector
  2. Select skills - Check the needed skills
  3. Or type "/" command - Quick skill selection
  4. Skills load on-demand - Dynamically loaded during execution

✨ Best Practices​

1. Writing Effective Prompts​

βœ… Be Specific and Clear​

  • Clearly describe your needs
  • Provide necessary background information
  • Specify expected output format
  • Include acceptance criteria

❌ Avoid Vague Requests​

  • Avoid overly brief descriptions
  • Avoid requests lacking context
  • Avoid asking too many things at once

2. Choosing the Right Agent​

3. Task Granularity​

Recommended granularity:

  • Small task: Single clear objective
  • Medium task: Contains a few related steps
  • Large task: Split into multiple smaller tasks

4. Providing Sufficient Context​

  • Upload relevant files
  • Add knowledge bases
  • Reference previous conversation content
  • Specify technical constraints or preferences

5. Leveraging Conversation History​

  • Continue conversations for iterative refinement
  • Reference previous messages for context

⚠️ Common Issues​

Q1: Conversation stuck in PENDING status?​

Possible reasons:

  1. Agent is unavailable
  2. System resources are limited
  3. Repository configuration error

Solutions:

  • Check agent status in Settings β†’ Agents
  • Verify repository access permissions
  • Try selecting a different agent

Q2: Agent response is incomplete?​

Solutions:

  • Click "Continue" to resume generation
  • Select a model with larger context window
  • Split complex tasks into smaller parts

Q3: How to stop a running conversation?​

Method:

  1. Click the "Stop" button in the input area
  2. Agent stops processing
  3. You can continue the conversation or start a new task

Q4: How to retry a failed conversation?​

Method:

  1. Click the "Retry" button on the failed message
  2. Optionally modify the message before retrying
  3. Agent will attempt the task again

Q5: How to share a conversation?​

Method:

  1. Click the share button in the message area
  2. Copy the generated link
  3. Share with team members (requires access permissions)

Q6: How to export conversation history?​

Method:

  1. Click the export button in the task menu
  2. Choose export format (Markdown, JSON)
  3. Download the conversation history

πŸ”— Related Resources​

Prerequisites​

Reference Documentation​

Detailed Feature Documentation​

For more details on advanced features, see:

πŸ’¬ Get Help​

Need assistance?

  • πŸ“– Check FAQ
  • πŸ› Submit GitHub Issue
  • πŸ’¬ Join community discussions

Start your first conversation and let AI agents work for you! πŸš€