Docs / AI Features
AI Features PRO
Custom AI Providers
Configure any provider, set up fallback chains, and diagnose connection errors when a key stops working.

Why “bring your own key”

DaazSync PRO never proxies your AI calls through daazlabs.com — you connect directly to the provider using your own API key, and pay that provider’s own rates. No middleman markup, and your article content never passes through our servers.

Configuring a provider

  1. Go to DaazSync → Settings → AI and open the section for your chosen provider (OpenAI, Claude, Gemini or Groq).
  2. Paste your API key into that provider’s own key field.
  3. Click the single Test Connection button at the bottom of the AI tab — it automatically tests whichever provider is currently active, so there’s no need to hunt for a provider-specific test button.
  4. Save your settings.

Using a fully custom endpoint

Not using one of the built-in providers? DaazSync PRO also accepts any custom API provider that follows an OpenAI-compatible chat completion format — point it at your own endpoint and key.

Fallback chains

You can configure a fallback order per provider, so if one model or key fails mid-run, the next one in the chain picks up the remaining articles instead of failing the whole import. This is especially useful when mixing a paid provider with a free-tier one that has stricter rate limits.

Tip: keep a free-tier provider (like Groq) as a fallback behind your primary paid provider — if you hit a temporary quota limit, imports keep completing instead of stalling.

Current model catalogue

ProviderNotes
OpenAIGPT-4o and compatible chat models.
Anthropic ClaudeAny Claude model your API key has access to.
Google GeminiCurrent catalogue: gemini-2.5-flash, gemini-2.5-flash-lite, gemini-3.5-flash-lite. Older models (gemini-2.0-flash, gemini-1.5-flash, gemini-1.5-flash-8b) were retired by Google and will 404.
GroqFree tier has strict per-minute token limits (e.g. 8,000 TPM on some models) — good as a fallback, or for smaller batches.

Troubleshooting connection errors

  • “Unexpected response (HTTP 404)” on Gemini: almost always means the configured model was retired — switch to one of the current models listed above.
  • Test button reports the wrong provider’s key is missing: make sure you’re pasting the key into that specific provider’s own field on the AI tab, not a shared field.
  • Import succeeds but no summaries appear: check Max AI calls per import run under the AI tab — if it’s set low, only the first N articles in that run receive a summary.