SubtitleEdit/subtitleedit · error · Exception

{error}

Error message

{error}

What it means

Thrown inside RunLlmOcr as a fail-fast guard: on the FIRST group only, if the OCR call returned empty text AND the engine's getError() is non-empty, the scan aborts immediately instead of grinding through every frame and reporting 'no subtitles found'. This catches a globally-broken engine early — most commonly a wrong API key or URL for the GLM/Ollama/LlamaCpp backends.

Source

Thrown at src/ui/Features/Video/VideoOcr/VideoOcrViewModel.cs:1053

        Func<VideoOcrFrameGroup, Task<string>> ocr,
        Func<string> getError,
        Action reportProgress,
        Action<VideoOcrFrameGroup> addPreviewLine,
        CancellationToken cancellationToken)
    {
        var isFirst = true;
        foreach (var group in ocrGroups)
        {
            cancellationToken.ThrowIfCancellationRequested();

            group.Text = VideoOcrLineBuilder.CleanOcrResult(await ocr(group));

            // Fail fast on a broken engine (wrong API key/URL) instead of grinding
            // through the whole video and reporting "no subtitles found".
            var error = getError();
            if (isFirst && string.IsNullOrEmpty(group.Text) && !string.IsNullOrEmpty(error))
            {
                throw new Exception(error);
            }

            isFirst = false;
            reportProgress();
            addPreviewLine(group);
        }
    }

    private async Task<bool> EnsureEngineIsAvailable()
    {
        var engineType = SelectedEngine.EngineType;

        if (engineType == OcrEngineType.Glm && string.IsNullOrWhiteSpace(GlmApiKey))
        {
            await MessageBox.Show(
                Window!,
                Se.Language.General.Error,
                "An API key is required for the GLM API engine.",

View on GitHub (pinned to 17a9f07487)

Solutions

1. Read the engine.Error text in the exception — it is the backend's own error (auth/model/HTTP).
2. For GLM: verify the API key is set and valid in settings.
3. For Ollama/LlamaCpp: confirm the URL is reachable and the named model is loaded (curl the /api/tags or /v1/models endpoint).
4. Correct the URL/model/key and re-run; only after the first frame succeeds will the rest proceed.

Example fix

// before
var error = getError();
if (isFirst && string.IsNullOrEmpty(group.Text) && !string.IsNullOrEmpty(error))
    throw new Exception(error);

// after - name the engine and frame so the user knows where it died
var error = getError();
if (isFirst && string.IsNullOrEmpty(group.Text) && !string.IsNullOrEmpty(error))
    throw new Exception($"OCR engine error on first frame ({group.RepresentativeFileName}): " + error);

// Validate credentials/endpoint before the scan for each LLM backend.
if (engineType == OcrEngineType.Glm && string.IsNullOrWhiteSpace(GlmApiKey)) throw new Exception("GLM API key is required.");
if (engineType == OcrEngineType.Ollama && !await IsReachable(OllamaUrl)) throw new Exception("Ollama server is not reachable.");

try { await RunLlmOcr(...); }
catch (Exception ex) // fail-fast guard re-throws the engine error
{
    SeLogger.Error(ex, "LLM OCR failed on first frame - check API key/URL/model");
    await MessageBox.Show(Window!, Se.Language.General.Error, ex.Message, MessageBoxButtons.OK, MessageBoxIcon.Error);
}

When it happens

Trigger: First frame yields no text and the engine reported an error: invalid/missing API key (GLM), wrong base URL, model name not found on the server, network/auth failure, or the inference server returning an error status.

Common situations: GLM API key empty or revoked; Ollama/LlamaCpp URL wrong or server not running; model id typo; rate-limit/auth error from the provider; server started but model failed to load.

Related errors

• API key invalid (or perhaps billing/API is not enabled)?
• llama.cpp returned {(int)resp.StatusCode}: {json}
• Ollama returned {(int)resp.StatusCode}: {json}
• API key invalid (or perhaps billing/API is not enabled)?
• Error calling Cloud Vision API: {httpException.Message}

AI-assisted analysis of SubtitleEdit/subtitleedit@17a9f07487 (2026-08-13). Data as JSON: /api/errors/0ba435757f399ce7. Report an issue: GitHub.