Adovo

Applied AI

LangGraph React submit errors need a tested UI path

Check the hook-level handler when per-submit error callbacks do not fire.

November 17, 2026

A user sends a message to an agent, the request fails, and the retry control never appears. A LangGraph JS pull request describes a specific cause: in the @langchain/langgraph-sdk/react useStream path, submit(input, { onError }) was typed but the per-submit callback was not called, while the hook-level onError did fire. The PR says the gap affected server and custom transport paths and also existed in an LTS branch. Check the installed package rather than assuming the fix is already in the deployed release.

Trace the error through the UI

Instrument a deliberate failing submit in a test environment. Confirm whether the network request failed, the hook-level error handler ran, the per-submit handler ran, and the UI moved out of its loading state. Use a visible recovery action that preserves the user's input. A silent loading spinner makes a recoverable failure look like an unresponsive agent.

Keep one recovery owner

Until the relevant SDK fix is in the installed version, put essential recovery behavior in the handler that actually fires, or wrap submit in a tested application-level error boundary. Avoid showing two retry prompts if both handlers fire after an upgrade. Test both server and custom transport paths if the product supports both. The LangGraph design guidance distinguishes transient errors from user-fixable failures; the UI should make that distinction visible.

Put failure in the user’s language

Show whether the message was sent, whether the run started and whether retry is safe. Keep the original input available. A generic “something went wrong” notice gives the user too little information to decide whether to resubmit.

Adovo’s view

An error handler is part of the product promise. The agent's reasoning can be sound while the interface leaves a person stranded. Recovery belongs in the path the user actually sees.

Continue exploring

Applied AI

LangGraphRun control

A busy LangGraph agent may queue work you meant to reject

Verify multitaskStrategy reject under a busy LangGraph JS stream so duplicate requests are not queued.

Read more

Applied AI

LangGraphAgent Reliability

A cancelled LangGraph dev run may still finish its work

Test late cancellation in LangGraph's in-memory dev runtime and bound side effects inside long-running nodes.

Read more

Applied AI

LangGraphStreaming

A hidden Location header can make an agent stream look finished

Test LangGraph JS run-stream reconnection through proxies and CORS before trusting a clean close.

Read more
Explore applied AI services →