When to use Era

Era is useful when the meaning of a message depends on the conversation around it. For example, “did I miss it?” could be a casual question or an urgent request. Era summarizes the user’s goal and urgency so your agent can respond appropriately.

​

Good use cases

• Support: detect urgency, frustration, and the user’s unresolved goal.
• Sales and shopping: tell browsing, comparing, and deciding apart.
• Companions and consumer agents: respond to the user’s current mood and intent.
• Long conversations: keep the user’s main goal clear as the history grows.

​

Poor use cases

• One-off prompts with no conversation.
• Stateless lookup or transformation tools.
• Workflows where your code already knows the exact next action.
• Applications that cannot send conversation text to the Era service.

​

What changes

| Without Era | With Era | | --- | --- | --- | | The model infers intent from raw history. | The model also receives an explicit goal, intent, and recommendation. | | Important signals can be buried in older messages. | Era re-summarizes the live situation each turn. | | You build classifiers and prompt rules yourself. | You add one turnContext() call. | Era does not replace your model, conversation history, or memory system. It adds a compact context block to the system prompt you already use.

​

Examples