Starting a Conversation

Overview

This guide walks you through creating a new conversation in the Chat App and sending your first message to an AI model.

Time Required: ~2 minutes Skill Level: Beginner

Prerequisites

Before starting:

• You have access to the Chat App within FlowState
• At least one LLM provider is configured (Anthropic, OpenAI, or LM Studio)
• Your API keys are set up in FlowState configuration

Steps

Step 1: Open the Chat App

Navigate to the Chat App from the FlowState app launcher. The app opens with the sidebar visible on the left and the main chat area on the right.

Step 2: Create a New Conversation

Click the New Chat button at the top of the sidebar. This action:

1. Creates a new conversation record in the database
2. Automatically selects the new conversation
3. Navigates you to the chat view

The conversation will have a default title like "New Conversation" which you can edit later.

Step 3: Select Your AI Model

Before sending your first message, choose your preferred AI provider and model:

1. Locate the model dropdown in the chat header
2. Click to expand the dropdown
3. Select a provider:
   • Anthropic: Claude Sonnet 4.5, Claude Opus 4.1, Claude Haiku 4.5
   • OpenAI: GPT-4o, GPT-4 Turbo, GPT-3.5 Turbo
   • LM Studio: Your locally loaded model
4. The selection is saved to the conversation settings

Step 4: Compose Your Message

Type your message in the input field at the bottom of the screen:

• The input field supports multiple lines
• Press Shift+Enter to add a new line
• The field automatically resizes as you type

Example prompts to try:

Explain the concept of closures in JavaScript with examples.

Write a Python function that validates email addresses using regex.

Help me design a database schema for a task management application.

Step 5: Send Your Message

Send your message using one of these methods:

• Press Enter on your keyboard
• Click the Send button

After sending:

1. Your message appears in the chat area (right-aligned)
2. The input field clears and becomes disabled during streaming
3. The AI's response begins streaming token by token

Step 6: Watch the Streaming Response

The AI response appears in real-time:

• Tokens appear one by one as they're generated
• A streaming indicator shows the response is in progress
• The chat automatically scrolls to show new content
• Once complete, the full message is saved to the database

Step 7: Continue the Conversation

After the response completes:

1. The input field becomes active again
2. Type your follow-up message
3. The conversation history provides context for the AI's next response

The AI maintains context from all previous messages in the conversation.

Expected Results

After completing this workflow, you should have:

• A new conversation visible in the sidebar
• Your message and the AI's response in the chat area
• The conversation saved and ready for future messages
• The ability to continue the conversation with follow-ups

Optional: Customize Conversation Settings

Edit the Title

1. Click on the conversation title in the header
2. Type a descriptive name
3. Press Enter to save or Escape to cancel

Set a System Prompt

System prompts guide the AI's behavior. Configure via conversation settings to:

• Set a specific tone or personality
• Provide context about your project
• Define output format preferences

Troubleshooting

Issue: No response after sending message

Solution:

• Check your API key configuration
• Verify you've selected a valid model
• Ensure network connectivity
• For LM Studio, confirm the local server is running

Issue: Model dropdown is empty

Solution:

• Verify API keys are configured for at least one provider
• For LM Studio, ensure the application is running with a loaded model

Issue: Response stops mid-stream

Solution:

• Wait a moment - the stream may resume
• If not, refresh the page - partial messages are saved
• Try a shorter prompt or different model

Related Workflows

• Managing Conversations - Edit, delete, and organize conversations
• Using MCP Tools - Enable AI to execute tools
• Switching Models - Change providers mid-conversation