API Connection Issues

API connection issues prevent reviews and chat from functioning because these features require communication with AI provider APIs (OpenAI, Anthropic, Google). Most connection issues fall into a few categories with straightforward solutions.

If you are using BYOK (Bring Your Own Key), the most common issue is an invalid, expired, or misconfigured API key. Verify your key by testing it directly with the provider's API or dashboard. Common problems include copying the key with trailing whitespace, using a key that has been rotated or revoked, or using a key that does not have the required permissions for the models you are trying to access. Re-enter your key in TeamBench settings if you suspect a copy error.

Rate limiting is another common BYOK issue. Each API provider imposes limits on how many requests you can make per minute or per day. If your team is submitting many reviews in a short period, you may hit these limits. Check your provider dashboard for current rate limit settings and usage. Increasing your rate limit or spreading reviews across time can resolve this.

AI provider outages affect all TeamBench users, not just BYOK users. OpenAI, Anthropic, and Google occasionally experience service disruptions. If reviews are failing across all models, check the status pages for each provider. TeamBench cannot process reviews when the underlying AI provider is experiencing downtime. These outages are typically resolved within minutes to hours by the provider.

Network issues on your side can prevent the TeamBench application from communicating with the backend. Ensure your browser can make HTTPS requests, your network does not block connections to TeamBench's API endpoints, and no VPN or proxy is interfering with the connection. If you are on a corporate network, check with your IT team whether external API connections are permitted.

If you receive specific error messages, note them carefully -- they often indicate the exact cause. "Invalid API key" means the key needs to be checked or replaced. "Rate limit exceeded" means you need to reduce request frequency or increase limits. "Model not found" means you are trying to use a model that is not available with your current key or plan. "Timeout" suggests a provider-side latency issue that often resolves on its own.

For intermittent issues that come and go, the cause is often transient provider-side latency. AI providers occasionally experience brief slowdowns during peak usage. If a review fails, waiting a minute and retrying often succeeds. Persistent failures across multiple retries and multiple models suggest a deeper issue worth investigating.

Clear your browser cache and cookies if you are experiencing persistent issues that other team members are not. Occasionally, cached authentication tokens or stale session data can cause connection problems. A fresh login often resolves these cases.

If none of the above resolves the issue, contact TeamBench support with the specific error message, the model you were trying to use, whether you are using BYOK or TeamBench credits, and the approximate time the issue started. This information helps the support team diagnose the problem quickly.