Chat and streaming

A chat completion

const reply = await client.chat.completions.create({
  model: "fta/zai/glm-5.3",
  messages: [{ role: "user", content: "Name three uses for a paperclip." }],
  temperature: 0.7,
  max_tokens: 400,
});
console.log(reply.choices[0]?.message.content);

Any extra field is sent as it is, so every Chat Completions parameter works: temperature, top_p, max_tokens, stop, seed, tools, tool_choice, response_format, reasoning_effort, and web_search_options.

Streaming text

for await (const text of client.chat.completions.streamText({
  model: "fta/kimi/k3",
  messages: [{ role: "user", content: "Tell me a short story." }],
})) {
  process.stdout.write(text);
}

Streaming chunks

With stream: true, create resolves to an async iterable of chat.completion.chunk objects. The last chunk carries usage.

const stream = await client.chat.completions.create({
  model: "fta/kimi/k3",
  messages,
  stream: true,
});

for await (const chunk of stream) {
  for (const choice of chunk.choices) {
    process.stdout.write(choice.delta.content ?? "");
  }
  if (chunk.usage) console.log("\n", chunk.usage);
}

If the stream stops before the model finishes, the loop throws StreamInterruptedError with the text received so far in .partial.

Reasoning

Thinking models return their reasoning in message.reasoning_content (and delta.reasoning_content when streaming). Turn thinking off with reasoning_effort: "none":

const reply = await client.chat.completions.create({
  model: "fta/zai/glm-5.3",
  messages,
  reasoning_effort: "none",
});

Some models write their thinking at the start of the reply inside <think>...</think>; the client moves it to reasoning_content so content holds only the answer.

Images in messages

Models that read images take OpenAI-style content parts:

const reply = await client.chat.completions.create({
  model: "fta/zai/glm-5.3-flash",
  messages: [
    {
      role: "user",
      content: [
        { type: "text", text: "What is in this picture?" },
        { type: "image_url", image_url: { url: "https://example.com/cat.jpg" } },
      ],
    },
  ],
});

A text-only model throws ImagesNotSupportedError; with fallback models set, the next model is tried.

Which model answered

With fallback models, reply.model names the model that answered.