Streaming Guide

The ChatIQ API supports Server-Sent Events (SSE) for real-time streaming responses. This guide covers the exact format, parsing strategies, and common patterns.

SSE Format Specification

ChatIQ uses a standard SSE format with the following event types:

Content Chunks

Each chunk of the response is sent as:

data: {"choices":[{"delta":{"content":"chunk text here"}}]}

Format:

• Event type: data:
• Content: JSON object with choices array
• Each choice has a delta object with content string
• Multiple chunks are sent sequentially

Conversation ID

After all content chunks, the conversation ID is sent as a separate event:

data: {"conversationId":"uuid-here"}

Note: Field name is conversationId (camelCase), not conversation_id.

Multi-Message Responses

Some responses return multiple assistant messages at once (for example, booking flows). In that case the stream includes a responses array instead of choices deltas:

data: {"responses":["First message","Second message"]}

Clients should render each entry as its own assistant bubble.

Attachments & Suggestions

Image responses can include attachments and optional suggestions in the stream. These may arrive as their own event or alongside responses:

data: {"attachments":[{"type":"image","url":"https://.../delivery.jpg","assetId":"uuid"}]}

data: {"suggestions":[{"label":"See Pad Thai photo","payload":"__item__:docId:itemId"}]}

Client guidance:

• Attachments can arrive before or after text chunks.
• If attachments is present, render images even when response text is empty.

Completion Marker

The stream ends with:

data: [DONE]

Error Events

If an error occurs during streaming, it's sent as an SSE event before [DONE]:

data: {"error":{"code":"RATE_LIMIT","message":"Rate limit exceeded"}}

data: [DONE]

Complete Parsing Example

Here's a complete example that handles all event types correctly:

async function streamChatResponse(message, apiKey, conversationId = null) {
  const response = await fetch("https://chatiq.io/api/chat", {
    method: "POST",
    headers: {
      "Content-Type": "application/json",
      Authorization: `Bearer ${apiKey}`,
    },
    body: JSON.stringify({
      message,
      bot_slug: "your-bot-slug",
      stream: true,
      conversation_id: conversationId,
    }),
  });

  if (!response.ok) {
    throw new Error(`HTTP error: ${response.status}`);
  }

  const reader = response.body?.getReader();
  const decoder = new TextDecoder();

  if (!reader) {
    throw new Error("No response body");
  }

  let fullResponse = "";
  let receivedConversationId = null;
  let buffer = "";

  try {
    while (true) {
      const { done, value } = await reader.read();

      if (done) {
        // Process any remaining buffer
        if (buffer.trim()) {
          processBuffer(buffer.trim());
        }
        break;
      }

      buffer += decoder.decode(value, { stream: true });

      // Process all complete "data: {...}" entries
      while (buffer.includes("data: ")) {
        const dataIndex = buffer.indexOf("data: ");

        // Skip any content before "data: " (shouldn't happen in SSE)
        if (dataIndex > 0) {
          buffer = buffer.slice(dataIndex);
          continue;
        }

        // Find the end of this data entry
        const afterData = buffer.slice(6); // Skip "data: "
        const dataEnd = afterData.indexOf("data: ");
        const newlineIndex = afterData.indexOf("\n");

        // Use the earliest of: next "data: ", newline, or end of buffer
        let endIndex = dataEnd;
        if (
          newlineIndex !== -1 &&
          (endIndex === -1 || newlineIndex < endIndex)
        ) {
          endIndex = newlineIndex;
        }

        // If no end found, wait for more data
        if (endIndex === -1) {
          break;
        }

        // Extract the data content
        const dataContent = afterData.slice(0, endIndex).trim();

        // Remove processed part from buffer
        buffer = buffer.slice(6 + endIndex);

        // Handle [DONE] marker
        if (dataContent === "[DONE]") {
          return {
            response: fullResponse,
            conversationId: receivedConversationId,
          };
        }

        if (!dataContent) continue;

        try {
          const parsed = JSON.parse(dataContent);

          // Handle content chunks
          if (parsed.choices?.[0]?.delta?.content) {
            const chunk = parsed.choices[0].delta.content;
            fullResponse += chunk;
            // Emit chunk to UI or callback
            onChunk(chunk);
          }
          // Handle conversation ID
          else if (parsed.conversationId) {
            receivedConversationId = parsed.conversationId;
          }
          // Handle errors
          else if (parsed.error) {
            throw new Error(parsed.error.message || "Stream error");
          }
        } catch (parseError) {
          console.warn("Failed to parse SSE data:", dataContent, parseError);
        }
      }
    }
  } finally {
    reader.releaseLock();
  }

  return {
    response: fullResponse,
    conversationId: receivedConversationId,
  };
}

function onChunk(chunk) {
  // Update UI with new chunk
  console.log("Received chunk:", chunk);
}

React Hook Example

Here's a reusable React hook for streaming:

import { useState, useCallback } from "react";

interface UseStreamingChatOptions {
  apiKey: string;
  botSlug: string;
  onChunk?: (chunk: string) => void;
}

export function useStreamingChat({
  apiKey,
  botSlug,
  onChunk,
}: UseStreamingChatOptions) {
  const [response, setResponse] = useState("");
  const [conversationId, setConversationId] = useState<string | null>(null);
  const [isLoading, setIsLoading] = useState(false);
  const [error, setError] = useState<Error | null>(null);

  const sendMessage = useCallback(
    async (message: string) => {
      setIsLoading(true);
      setError(null);
      setResponse("");

      try {
        const result = await streamChatResponse(
          message,
          apiKey,
          conversationId
        );
        setResponse(result.response);
        if (result.conversationId) {
          setConversationId(result.conversationId);
        }
      } catch (err) {
        setError(err instanceof Error ? err : new Error("Unknown error"));
      } finally {
        setIsLoading(false);
      }
    },
    [apiKey, botSlug, conversationId, onChunk]
  );

  return {
    sendMessage,
    response,
    conversationId,
    isLoading,
    error,
  };
}

Error Handling

HTTP Errors

Check response status before processing stream:

if (!response.ok) {
  const errorData = await response.json().catch(() => ({}));
  throw new Error(errorData.error?.message || `HTTP ${response.status}`);
}

Stream Errors

Errors can be sent within the SSE stream:

if (parsed.error) {
  throw new Error(parsed.error.message || "Stream error");
}

Network Errors

Handle connection issues:

try {
  // ... streaming code
} catch (error) {
  if (error instanceof TypeError && error.message.includes("fetch")) {
    // Network error
    throw new Error("Network error: Unable to connect to ChatIQ API");
  }
  throw error;
}

Common Patterns

Pattern 1: Server-Side Proxy (Recommended)

Use a Next.js API route to proxy requests and hide your API key:

// app/api/chatbot/route.ts
export async function POST(request: Request) {
  const { message, conversation_id } = await request.json();

  const chatIQResponse = await fetch("https://chatiq.io/api/chat", {
    method: "POST",
    headers: {
      "Content-Type": "application/json",
      Authorization: `Bearer ${process.env.CHATIQ_API_KEY}`,
    },
    body: JSON.stringify({
      message,
      bot_slug: process.env.CHATIQ_BOT_SLUG,
      stream: true,
      conversation_id,
    }),
  });

  // Stream the response back
  return new Response(chatIQResponse.body, {
    headers: {
      "Content-Type": "text/event-stream",
      "Cache-Control": "no-cache",
      Connection: "keep-alive",
    },
  });
}

Benefits:

• API key stays on server (never exposed to client)
• No CORS issues
• Can add custom error handling/logging

Pattern 2: Direct Client-Side (Requires CORS)

If ChatIQ CORS is configured for your domain:

// Direct fetch from browser
const response = await fetch("https://chatiq.io/api/chat", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    Authorization: `Bearer ${apiKey}`, // ⚠️ Exposes API key to client
  },
  body: JSON.stringify({
    message: "Hello",
    bot_slug: "my-bot",
    stream: true,
  }),
});

Note: This exposes your API key in client-side code. Use with caution.

Pattern 3: Server-Side Only (No Streaming)

For server-side applications, use JSON mode:

const response = await fetch("https://chatiq.io/api/chat", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    Authorization: `Bearer ${apiKey}`,
  },
  body: JSON.stringify({
    message: "Hello",
    bot_slug: "my-bot",
    stream: false, // JSON response
  }),
});

const data = await response.json();
console.log(data.response); // Full response text
console.log(data.conversationId); // Conversation ID

Troubleshooting

Issue: Buffer Not Processing Correctly

Symptom: Chunks are missing or incomplete

Solution: Ensure you're handling incomplete buffers correctly:

// Wait for complete entries
if (endIndex === -1) {
  break; // Wait for more data
}

Issue: Conversation ID Not Received

Symptom: conversationId is always null

Solution: Check that you're parsing the separate conversationId event:

if (parsed.conversationId) {
  receivedConversationId = parsed.conversationId;
}

Issue: Errors Not Caught

Symptom: Stream fails silently

Solution: Always check for error events:

if (parsed.error) {
  throw new Error(parsed.error.message);
}

Best Practices

1. Always handle errors - Check for error events in the stream
2. Use server-side proxy - Keep API keys secure
3. Handle incomplete buffers - Don't parse incomplete JSON
4. Release readers - Always call reader.releaseLock() in finally block
5. Track conversation ID - Maintain context across messages
6. Implement retry logic - Handle network failures gracefully

Next Steps

• See API Reference for all endpoints
• See Getting Started for onboarding