Harden cron article generation with Gemini model fallback chain

[Copilot]

The cron GitHub Actions job was failing because /api/cron/generate returned 500 {"error":"Failed to generate content"} when content generation failed on a single Gemini model. This change makes generation resilient to model-level failures instead of failing immediately.

• Root-cause alignment

  • The cron workflow itself was functioning; failure occurred in lib/ai.ts where generation used one hardcoded model and threw on first provider/model error.

• Generation fallback strategy

  • Added ordered model candidates in generateArticle:
    • process.env.GEMINI_MODEL (if set)
    • gemini-2.5-flash-lite
    • gemini-1.5-flash
  • Deduplicates candidates and attempts each model sequentially.

• Failure handling behavior

  • Per-model errors are logged and retried with the next candidate.
  • Existing final behavior is preserved: if all candidates fail, the function throws Failed to generate content.

const modelCandidates = [
  process.env.GEMINI_MODEL,
  "gemini-2.5-flash-lite",
  "gemini-1.5-flash",
].filter((model, index, self): model is string => !!model && self.indexOf(model) === index);

for (const modelName of modelCandidates) {
  try {
    const model = genAI.getGenerativeModel({ model: modelName });
    const result = await model.generateContent(prompt);
    // parse/return response...
  } catch (error) {
    // log and continue to next model
  }
}
throw new Error("Failed to generate content");

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
genpost	Ready	Preview, Comment	Jul 16, 2026 12:30pm