Chat and Streaming - Tensormesh User Documentation

Generate A Single Response

tensormesh(modelId), tensormesh.chatModel(modelId), and tensormesh.languageModel(modelId) create AI SDK chat models backed by /v1/chat/completions.

import { generateText } from "ai";
import { tensormesh } from "@tensormesh/ai-sdk-provider";

const result = await generateText({
  model: tensormesh("Qwen/Qwen3-Coder-30B-A3B-Instruct"),
  prompt: "Write a concise implementation plan for adding streaming chat.",
});

console.log(result.text);

Stream A Response

import { streamText } from "ai";
import { tensormesh } from "@tensormesh/ai-sdk-provider";

const result = streamText({
  model: tensormesh("Qwen/Qwen3-Coder-30B-A3B-Instruct"),
  prompt: "Explain how serverless inference helps developer tooling.",
});

for await (const text of result.textStream) {
  process.stdout.write(text);
}

Text Completions

Use completionModel when you specifically need the /v1/completions route.

import { generateText } from "ai";
import { tensormesh } from "@tensormesh/ai-sdk-provider";

const result = await generateText({
  model: tensormesh.completionModel("openai/gpt-oss-20b"),
  prompt: "Complete this sentence: Serverless inference is useful because",
});

console.log(result.text);

For most applications, prefer tensormesh(modelId) because chat models support richer AI SDK features such as tool calling and structured output.

⌘I