Message-to-LLM Converter

toAiMessages() converts FluxyChat messages into the message format expected by AI SDKs (OpenAI, Anthropic, Vercel AI SDK, etc.).

When to Use

• Feeding chat history to an AI agent
• Converting a conversation thread into context for generateText or streamText
• Building AI-powered chat features on top of FluxyChat

Usage

import { toAiMessages } from "@fluxy-chat/sdk";

const messages = await client.getMessages("room-1", { limit: 50 });
const history = await toAiMessages(messages, {
  botUserId: "agent-bot-123",
});

// Use with any AI SDK
const response = await openai.chat.completions.create({
  model: "gpt-4",
  messages: history,
});

Options

botUserId

Identifies which user is the bot/agent. Messages from this user are mapped to role: "assistant", all others to role: "user".

toAiMessages(messages, { botUserId: "my-bot-id" });

includeNames

When true, prefixes user messages with [userId]: for multi-user context:

toAiMessages(messages, { includeNames: true });
// User message → "[alice]: hello there"

transformMessage

Hook to modify or skip individual messages:

toAiMessages(messages, {
  transformMessage: (aiMessage, sourceMsg) => {
    // Add prefix to all messages
    return { ...aiMessage, content: `[${sourceMsg.roomId}] ${aiMessage.content}` };
    // Or return null to skip
  },
});

onUnsupportedAttachment

Callback for attachment types that can't be converted (audio, video):

toAiMessages(messages, {
  onUnsupportedAttachment: (att, msg) => {
    console.warn(`Can't include ${att.kind} from message ${msg.id}`);
  },
});

How It Works

1. Messages are sorted chronologically by createdAt
2. Empty/whitespace-only content is filtered out
3. Bot messages become assistant, user messages become user
4. Link previews are appended as text
5. Image and text files are converted to FilePart using their URL
6. Unsupported attachments trigger the callback

Output Format

type AiMessage =
  | { role: "user"; content: string | AiMessagePart[] }
  | { role: "assistant"; content: string };

This format is compatible with the Vercel AI SDK's ModelMessage type and most LLM provider APIs.

See Also

• Provider Registry Guide — Managing AI providers
• LLM Middleware Guide — Transforming AI SDK calls