TL;DR: Symptom: server logs show the tool call and the tool result, but the model produces no follow-up response. The chat just stalls. Cause: the route passed raw UIMessages to streamText without converting them.

## Fix
1. Symptom: server logs show the tool call and the tool result, but the model produces no follow-up response.
   Expected: this specific failure stops.
2. If you see messages passed straight from req.json() into streamText with no convertToModelMessages call, that is the bug.
   Expected: this specific failure stops.
3. Check the network stream contains tool-result parts followed by text parts.
   Expected: the value or setting is exactly right, nothing ambiguous.

## Details
The chat just stalls. Cause: the route passed raw UIMessages to streamText without converting them. Tool calls and results in UIMessage format are not valid model input, so the follow-up step has nothing to work with. Confirmation: 1. Look at the route handler. 2. Since v6 convertToModelMessages is async. If you see messages: convertToModelMessages(messages) without await, the model receives a Promise object. Same silence. Fix: import { convertToModelMessages, createUIMessageStreamResponse, streamText, toUIMessageStream } from 'ai'; export async function POST(req: Request) { const { messages } = await req.json(); const result = streamText({ model: 'your-model-id', messages: await convertToModelMessages(messages), }); return createUIMessageStreamResponse({ stream: toUIMessageStream({ stream: result.stream }) }); } Verification: send a message that triggers a tool, confirm the tool executes and the model writes a final answer referencing the result. A single-step limit can also cut the conversation off right after the tool result.

## When to use
You hit exactly this: Diagnose useChat going silent when tools are involved in Diagnose.

## When not to use
A different error, or the same symptom in a different tool. This page only covers the failure above.

## Compatibility
Diagnose (versions mentioned here: v6).

## Edge cases
- Also confirm stopWhen allows multiple steps (e.g. stopWhen: isStepCount(5)).