unslothai/unsloth · error · Error

${parsed.error.message || "Stream error"}

Error message

${parsed.error.message || "Stream error"}

What it means

Thrown when a parsed SSE data payload contains an error object — the server pushed a mid-stream error event ({error: {message}}) instead of an OpenAI chat chunk. The thrown message is the server-provided error.message, falling back to 'Stream error'. This fires after the stream already started, so partial output may exist.

Source

Thrown at studio/frontend/src/features/chat/api/chat-api.ts:1378

        const dataLines = parseSseEvent(rawEvent);
        if (dataLines.length === 0) {
          separatorIndex = buffer.search(/\r?\n\r?\n/);
          continue;
        }

        const dataText = dataLines.join("\n");
        if (dataText === "[DONE]") {
          completed = true;
          sawTerminalSignal = true;
          throwIfReasoningOnlyLength();
          return;
        }

        const parsed = JSON.parse(dataText) as
          | OpenAIChatChunk
          | { type?: string; content?: string; error?: { message?: string } };
        if ("error" in parsed && parsed.error) {
          throw new Error(parsed.error.message || "Stream error");
        }
        // Tool status events are custom SSE payloads, not OpenAI chunks
        if ("type" in parsed && parsed.type === "tool_status") {
          yield {
            _toolStatus: parsed.content ?? "",
          } as unknown as OpenAIChatChunk;
          separatorIndex = buffer.search(/\r?\n\r?\n/);
          continue;
        }
        // Diffusion frame: a per-step canvas snapshot. Custom SSE payload (not an OpenAI chunk) with
        // no assistant text, surfaced as a transient marker for the in-place renderer, never the transcript.
        if ("type" in parsed && parsed.type === "diffusion_frame") {
          yield {
            _diffusionFrame: parsed,
          } as unknown as OpenAIChatChunk;
          separatorIndex = buffer.search(/\r?\n\r?\n/);
          continue;
        }

View on GitHub (pinned to 203007d190)

Solutions

1. Read the thrown message — it is the server's own description of the failure.
2. Retry once; transient upstream overload often clears.
3. If it is a moderation/content error, change the prompt rather than retrying.
4. Check backend logs for the upstream error that was forwarded.

function isStreamErrorEvent(parsed: unknown): parsed is { error: { message?: string } } {
  return typeof parsed === 'object' && parsed !== null && 'error' in parsed && !!(parsed as any).error;
}

try { for await (const c of stream) render(c); }
catch (e) {
  const msg = e instanceof Error ? e.message : '';
  if (/overloaded|rate/i.test(msg)) return retryOnce();
  showError(msg || 'Stream error');
}

When it happens

Trigger: The backend streams normally for a while, then emits a JSON event containing an error key — e.g. upstream provider failure, content filter, or model crash reported asynchronously over the same SSE channel.

Common situations: Upstream provider errors surfacing mid-stream (overloaded, context exceeded mid-conversation, moderation trigger); gateway converting a late upstream 5xx into an in-stream error event.

Related errors

• {path}: not valid JSON: {exc}
• ChatGPT returned a malformed stream.
• ChatGPT stream ended before completion.
• Stream response missing body
• Response interrupted: the connection dropped before the mode

AI-assisted analysis of unslothai/unsloth@203007d190 (2026-08-15). Data as JSON: /api/errors/2b8f552f9d2e9dcf. Report an issue: GitHub.