Mintplex-Labs/anything-llm · error · Error

e.message

Error message

e.message

What it means

GeminiLLM.getChatCompletion wraps the OpenAI-compatible chat completion call in a .catch that logs the original error to console and rethrows new Error(e.message). This is a passthrough of the underlying SDK error: typical causes are 400 INVALID_ARGUMENT (unsupported parameter or unsupported model for the OpenAI-compat layer), 429 quota exhaustion, 401/403 key problems, or 404 for a deprecated model id. The console.error means the full structured error (status, code) is visible in server logs even though only the message propagates.

Solutions

  1. Check the server console — the original error with HTTP status is logged just above; fix per status (429=quota/billing, 401=key, 404=model gone)
  2. Update the model to a current id (default here is gemini-2.0-flash-lite) via workspace pref or GEMINI_LLM_MODEL_PREF
  3. If 429, enable billing or wait for the quota window; check Google AI Studio quotas
  4. If the inner message mentions system role / system_instruction, switch away from the gemma no-system-prompt models or adjust the prompt construction
Defensive patterns

Strategy: try-catch

Try / catch

try {
  return await llm.getChatCompletion(messages);
} catch (err) {
  // message is Gemini's raw error; the full object was console.error'd server-side
  if (/429|RESOURCE_EXHAUSTED|quota/i.test(err.message)) return respond("Gemini quota hit — retry later.");
  if (/404|not found/i.test(err.message)) return respond("Model retired — pick a current Gemini model.");
  if (/401|403|API key/i.test(err.message)) return respond("Invalid GEMINI_API_KEY.");
  throw err;
}

Prevention

When it happens

Trigger: chat.completions.create against generativelanguage.googleapis.com/v1beta/openai/ failing: e.g. temperature outside allowed range for a thinking model, model id like a retired gemini-1.x, free-tier rate limit (429 RESOURCE_EXHAUSTED), or invalid API key. Note the class routes experimental models to a v1beta endpoint and some models (gemma list) can't take a system prompt.

Common situations: Google deprecated/renamed the pinned model (e.g. old gemini-pro ids); free tier quota hit at peak; key restricted by API key settings (referrer/IP restrictions); using a gemini model variant that rejects the system-role message the prompt builder always prepends.

Related errors

AI-assisted analysis of Mintplex-Labs/anything-llm@3aec848f28 (2026-08-18). Data as JSON: /api/errors/14ac8ef9517ca3db. Report an issue: GitHub.

Appendix: source

Thrown at server/utils/AiProviders/gemini/index.js:390

      ...formatChatHistory(chatHistory, this.#generateContent),
      {
        role: "user",
        content: this.#generateContent({ userPrompt, attachments }),
      },
    ];
  }

  async getChatCompletion(messages = null, { temperature = 0.7 }) {
    const result = await LLMPerformanceMonitor.measureAsyncFunction(
      this.openai.chat.completions
        .create({
          model: this.model,
          messages,
          temperature: temperature,
        })
        .catch((e) => {
          console.error(e);
          throw new Error(e.message);
        })
    );

    if (
      !result.output.hasOwnProperty("choices") ||
      result.output.choices.length === 0
    )
      return null;

    return {
      textResponse: result.output.choices[0].message.content,
      metrics: {
        prompt_tokens: result.output.usage.prompt_tokens || 0,
        completion_tokens: result.output.usage.completion_tokens || 0,
        total_tokens: result.output.usage.total_tokens || 0,
        outputTps: result.output.usage.completion_tokens / result.duration,
        duration: result.duration,
        model: this.model,

View on GitHub (pinned to 3aec848f28)