Reliability

Most of what keeps an assistant dependable happens behind the scenes. You never see it directly, but it answers the question every buyer asks: can I trust this not to break?

This section explains what ChatHQ does so that an assistant that works today still works after the next update.

What ChatHQ does

• Assistant versions — Every assistant keeps numbered versions. You edit a draft, publish when ready, and can restore an earlier version if something gets worse.
• Release testing — A fixed set of real questions is run against the assistant on every release, so a change that breaks an answer is caught before it reaches you.
• Failure visibility — When the assistant cannot answer, ChatHQ records why. A knowledge gap and a provider outage are never confused, and visitors are never told "I don't know" when the real cause is a temporary failure.
• Choice of AI model — ChatHQ does not depend on a single AI provider. You choose the model behind each assistant and can change it without rebuilding anything.

Related

• Trust & accuracy — What the assistant does to make sure each individual answer can be trusted.
• Troubleshooting widget — What to check when the widget does not load or returns an error.