different-ai/openwork · error · AgentContextDiagnosticsTransportError

agent_context_diagnostics_response_too_large

agent_context_diagnostics_response_too_large

Error message

Agent context diagnostics response exceeded the 1 MiB limit.

What it means

readBoundedResponseText first inspects the declared Content-Length header. If it is a valid integer exceeding AGENT_CONTEXT_DIAGNOSTICS_RESPONSE_MAX_BYTES (1 MiB), it cancels the body and throws responseTooLargeError() without reading anything. This pre-read guard avoids downloading a body known in advance to be too big.

Source

Thrown at apps/app/src/app/lib/agent-context-diagnostics-transport.ts:60

function cancelUnlockedBody(response: Response): void {
  if (!response.body || response.body.locked) return;
  void response.body.cancel().catch(() => undefined);
}

async function readBoundedResponseText(
  response: Response,
  signal: AbortSignal,
): Promise<string> {
  const declaredLength = response.headers.get("content-length")?.trim() ?? "";
  if (/^\d+$/.test(declaredLength)) {
    const declaredBytes = Number(declaredLength);
    if (
      !Number.isSafeInteger(declaredBytes)
      || declaredBytes > AGENT_CONTEXT_DIAGNOSTICS_RESPONSE_MAX_BYTES
    ) {
      cancelUnlockedBody(response);
      throw responseTooLargeError();
    }
  }

  if (!response.body) return "";

  const reader = response.body.getReader();
  const chunks: Uint8Array[] = [];
  let bytesRead = 0;
  const cancelOnAbort = () => {
    void reader.cancel().catch(() => undefined);
  };
  signal.addEventListener("abort", cancelOnAbort, { once: true });

  try {
    while (true) {
      if (signal.aborted) throw timeoutError();
      const chunk = await reader.read();
      if (chunk.done) break;

View on GitHub (pinned to 2b7df46e8a)

Solutions

Fix the diagnostics endpoint to cap/truncate its JSON payload to under 1 MiB
Verify the client URL points at the real agent-context diagnostics endpoint, not another route
Check for proxies/middleware that inflate the response (error pages, injected content)
If the payload is legitimately larger, the endpoint must paginate or summarize rather than grow the response

Example fix

// before: endpoint serializes full context
return Response.json(fullContext);
// after: cap the payload server-side
return Response.json(truncateContext(fullContext, 1_000_000));

Defensive patterns

Strategy: try-catch

Validate before calling

const contentLength = Number(response.headers.get("content-length") ?? "");
if (Number.isSafeInteger(contentLength) && contentLength > 1024 * 1024) {
  throw new Error("Diagnostics response would exceed the 1 MiB limit");
}

Type guard

null

Try / catch

try {
  const { payload } = await requestAgentContextDiagnosticsPayload({ fetchImpl, url, init });
} catch (e) {
  if (e instanceof AgentContextDiagnosticsTransportError && e.code === "agent_context_diagnostics_response_too_large") {
    // endpoint returned >1MiB; request a truncated/paginated payload
  } else throw e;
}

Prevention

Cap the diagnostics payload server-side below 1 MiB
Confirm you are calling the real diagnostics endpoint, not a debug route
Check status/content-type so error pages aren't mistaken for diagnostics JSON
Paginate or summarize large context data instead of one big response

When it happens

Trigger: The diagnostics endpoint responds with a Content-Length header whose numeric value is greater than 1048576 bytes — e.g. a server that isn't the bounded diagnostics endpoint or that failed to truncate its payload.

Common situations: Pointing the diagnostics client at a wrong/debug endpoint that returns huge payloads; a server bug that serializes the full agent context unbounded; a proxy injecting large HTML error pages with explicit Content-Length.

Related errors

AI-assisted analysis of different-ai/openwork@2b7df46e8a (2026-09-01). Data as JSON: /api/errors/8b91521bca7173b5. Report an issue: GitHub.