Chat Completions

The SDK provides powerful chat completion capabilities with support for streaming, expert routing, and conversation context.

Basic Usage

import { chat } from "@cpod/sdk";

const response = await chat.complete({
  message: "What is machine learning?",
});

console.log(response.content);

Streaming Responses

For real-time streaming output:

import { chat } from "@cpod/sdk";

for await (const chunk of chat.stream({ message: "Write a story" })) {
  process.stdout.write(chunk.content);
}

React Streaming Example

import { useState } from "react";
import { chat } from "@cpod/sdk";

function ChatComponent() {
  const [response, setResponse] = useState("");
  const [loading, setLoading] = useState(false);

  const sendMessage = async (message: string) => {
    setLoading(true);
    setResponse("");

    for await (const chunk of chat.stream({ message })) {
      setResponse((prev) => prev + chunk.content);
    }

    setLoading(false);
  };

  return (
    <div>
      <button onClick={() => sendMessage("Hello!")}>
        Send
      </button>
      <div>{response}</div>
    </div>
  );
}

Expert Routing

Route messages to specialized AI experts:

import { chat, listExperts } from "@cpod/sdk";

// List available experts
const experts = await listExperts();
console.log(experts.map(e => e.name));

// Route to a specific expert
const response = await chat.complete({
  message: "How should I price this product?",
  expertId: "sales-expert-id",
});

Conversation Context

Maintain conversation history:

import { chat } from "@cpod/sdk";

// Continue an existing conversation
const response = await chat.complete({
  message: "Tell me more about that",
  chatId: "existing-chat-id",
});

// Start a new chat in a project
const response = await chat.complete({
  message: "Analyze this document",
  projectId: "project-123",
});

Advanced Options

import { chat } from "@cpod/sdk";

const response = await chat.complete({
  message: "Write creative content",

  // Model parameters
  temperature: 0.8,      // Higher = more creative
  max_tokens: 2000,      // Maximum response length

  // Context
  expertId: "writer",    // Route to specific expert
  projectId: "proj-123", // Associate with project
  chatId: "chat-456",    // Continue conversation
});

Response Structure

interface ChatResponse {
  content: string;           // The AI response text
  chat_id: string;          // Conversation ID
  message_id: string;       // This message's ID
  expert_id?: string;       // Expert that responded
  tokens_used?: number;     // Token count
}