Chat

Overview

The chat feature provides real-time AI conversations using Server-Sent Events (SSE) for streaming responses.

Endpoints

How It Works

1. User sends a message from the chat UI
2. Server validates session and checks credit balance
3. Credits are deducted (10 credits per message)
4. Request is sent to Volcano Engine (Doubao model) — demo route only
5. Response streams back via SSE to the client
6. Message is saved to chatMessage table

Data Model

Chat Sessions (chatSession)

Each conversation is a session:

• id — Session identifier
• userId — Owner
• model — AI model used
• totalCreditsUsed — Running total of credits spent

Chat Messages (chatMessage)

Individual messages within a session:

• sessionId — Parent session
• role — user or assistant
• content — Message text
• creditsUsed — Credits consumed for this message

Configuration

Model parameters can be adjusted in the API route:

const response = await chat({
  model: 'doubao-1-5-thinking-pro-250415',
  messages: [...],
  temperature: 0.7,
  top_p: 0.9,
  max_tokens: 2048,
});