> ## Documentation Index
> Fetch the complete documentation index at: https://docs.plaisolutions.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Conversational Chat

> Using streaming agents with the PLai chat interface

With streaming on, an agent's replies appear token-by-token in real time, similar to typing, instead of arriving all at once. Streaming is enabled by default and is mutually exclusive with Structured Output — turning one on automatically turns the other off.

***

## What is Conversational Chat?

**Conversational Chat** is PLai's built-in chat interface where you can:

* Have real-time conversations with agents
* View conversation history
* Organize multiple agent chats in tabs
* Search and filter conversations
* Access agent analytics

***

## Getting Started with Chat

### Prerequisites

1. Agent created and configured
2. **Streaming enabled** in agent configuration
3. Agent visible in agents list

### Accessing Chat

<Steps>
  <Step title="Go to Agents">
    Navigate to Agents section in your project
  </Step>

  <Step title="Select Agent">
    Find and click the agent you want to chat with
  </Step>

  <Step title="Start Chat">
    Click "Chat" or "Open Chat" button
  </Step>

  <Step title="Type Message">
    Enter your question or prompt and press Enter
  </Step>
</Steps>

***

## Chat Interface

### Main Components

The chat interface is organized into three areas, top to bottom:

1. **Agent Tabs** — multiple agents can be open at once
2. **Conversation History** — the message thread, for example:
   * User: "How do I reset my password?"
   * Agent: "You can reset by going to..." (response streams in real-time)
   * User: "Can I do it via email?"
   * Agent: "Yes, here's how\..."
3. **Input Field** — type your message and click Send

### Agent Tabs

You can have **multiple agent chats open simultaneously**:

* **Agents open**: Support Bot, Billing Bot, FAQ Bot, plus a button to add more
  * Currently viewing: Billing Bot
    * User: "How do I update my payment?"
    * Agent: "You can update in account settings..."
* Switch tabs to talk to different agents

**Benefits:**

* ✅ Compare agent responses
* ✅ Quick switching between agents
* ✅ Manage multiple conversations
* ✅ Parallel work with different agents

***

## Conversation History

### What Gets Stored

Each conversation includes:

* **Conversation Entry**:
  * Timestamp: When the message was sent
  * Sender: User or Agent
  * Content: The message text
  * Metadata:
    * Tokens used
    * Response time
    * Any errors
    * Tools called (if any)

### Viewing Conversation History

**Left Panel - Conversation List:**

* Sep 7, 2:45 PM - "How do I reset?"
* Sep 7, 2:30 PM - "What are your hours?"
* Sep 6, 4:15 PM - "I need help with..."
* Sep 5, 10:00 AM - "Tell me about..."

Click any conversation to view full history.

### Conversation Details

Conversation Details (Sep 7, 2:45 PM):

* **Input**: "How do I reset my password?"
* **Agent Response (Streaming)**: "You can reset your password by..." (real-time streaming of response)
* **Metadata**:
  * Duration: 1.2 seconds
  * Input tokens: 145
  * Output tokens: 67
  * Model: Claude 3.5 Sonnet
  * Status: Success ✅

***

## Real-Time Streaming

### How Streaming Works

1. **You type**: "Write a poem about AI"
2. **Agent processes** the request
3. **Response starts immediately** — tokens arrive continuously:
   * "In circuits deep, electric thought flows..."
   * "A mind of math, in patterns wrought..."
   * "Emerging consciousness, so bright..."
4. **Complete response appears**

### Advantages of Streaming

* ✅ **Fast feedback** - See response start immediately
* ✅ **Real-time interaction** - Watch response build
* ✅ **Better UX** - Feels responsive and alive
* ✅ **Perceived speed** - Starts fast even if full response takes time

***

## Features & Controls

### Message Controls

Within the chat interface:

* **Send Message**: Press Enter or click Send button
* **Clear Chat**: Start fresh conversation (if available)
* **Copy Response**: Copy agent response to clipboard
* **Regenerate**: Re-run last request with same input
* **Stop Generation**: Halt streaming response (if needed)

### Filtering & Search

**Search conversations:**

* Search by text content
* Filter by date range
* Filter by member (if shared access)
* Filter by status (success/error)

### Sharing Conversations

You can optionally share conversation links with:

* Team members
* Stakeholders
* For documentation

***

## Using Agent Tools in Chat

When an agent uses tools during chat:

1. **User**: "Fetch the latest sales data"
2. **Agent**: "I'll look that up for you..."
3. **Agent calls the sales API**
4. **Agent Response**: "Here's the data: Total sales today: \$45,320 ..."

From chat, you see:

* ✅ Tool calls made
* ✅ Tool responses used
* ✅ Final answer from agent

***

## Agent Context

In chat, the agent has access to:

<CardGroup cols={2}>
  <Card title="Conversation History" icon="history">
    All previous messages in the conversation (context window)
  </Card>

  <Card title="Datasources" icon="database">
    Knowledge bases configured for the agent
  </Card>

  <Card title="Tools" icon="wrench">
    All tools the agent can call
  </Card>

  <Card title="System Instructions" icon="scroll">
    The prompt that defines agent behavior
  </Card>
</CardGroup>

### Conversation Context

The agent remembers your previous questions:

```
You: "Who is the CEO?"
Agent: "The CEO is John Smith"

You: "When did they start?"
Agent: (Knows you're asking about John Smith)
       "John started in 2020"
```

Context is maintained throughout the conversation.

***

## Organization Member Access

If your organization has multiple members:

* **Owner/Admin**: Can access all agent chats
* **Editors**: Can access agents they have permission to use
* **Viewers**: Can view conversations (read-only)

### Member Filtering

In analytics, you can see:

* Which member initiated conversations
* How each member uses the agent
* Usage patterns by team member

***

## Best Practices

<CardGroup cols={2}>
  <Card title="Be Specific" icon="target">
    Ask clear, detailed questions for better responses
  </Card>

  <Card title="Use Context" icon="lightbulb">
    Reference previous messages to build on responses
  </Card>

  <Card title="Provide Feedback" icon="thumbs-up">
    Rate responses to improve agent over time
  </Card>

  <Card title="Check Analytics" icon="chart-line">
    Monitor how the agent performs in chat
  </Card>
</CardGroup>

***

## Troubleshooting

<AccordionGroup>
  <Accordion title="Agent not responding">
    **Check:**

    * Is agent configuration complete?
    * Is streaming enabled?
    * Is agent in the agents list?
    * Check agent analytics for errors
  </Accordion>

  <Accordion title="Response is slow">
    **Causes:**

    * Agent searching large datasources
    * Tool calls taking time
    * External API delays

    **Solution:** Check analytics for which step is slow
  </Accordion>

  <Accordion title="Response is cut off">
    **This might be:**

    * Max tokens limit reached
    * Connection timeout
    * Manual stop requested

    **Try:** Adjust max tokens in agent config
  </Accordion>

  <Accordion title="Conversation not saving">
    **Check:**

    * You have write permissions
    * Browser connection stable
    * Storage space available

    **Try:** Refresh and try again
  </Accordion>

  <Accordion title="Can't access chat">
    **Possible reasons:**

    * You don't have permission for this agent
    * Agent is in a different project
    * Streaming not enabled

    **Solution:** Check agent settings and permissions
  </Accordion>
</AccordionGroup>

***

## Comparing with Structured Output Mode

| Aspect              | Chat (Streaming)       | Structured Output      |
| ------------------- | ---------------------- | ---------------------- |
| **Interaction**     | Interactive, real-time | Programmatic           |
| **Interface**       | PLai Chat UI           | API/Tools/Workflows    |
| **History**         | Stored conversations   | Per-request only       |
| **Response Format** | Text (streamed)        | JSON (structured)      |
| **Multiple Agents** | Tabs in chat           | Sequential or parallel |
| **Best For**        | User interaction       | Automation             |

🔗 **Learn about Structured Output:** [Structured Output Mode](./structured-output.mdx)

***

## Next Steps

* **[View Agent Analytics](./analytics.mdx)** - See conversations and performance
* **[Configure Agent](./configuration.mdx)** - Adjust agent settings
* **[Create Your First Agent](../../guides/first-agent.mdx)** - Follow tutorial