Messaging

Build conversations around a defined character, the user's recent messages and the context your product is allowed to retain.

Conversation flow

1. Authenticate the user in your product.
2. Load the permitted character configuration and relevant conversation context.
3. Submit the message from your backend.
4. Stream or return the response to the user.
5. Store the resulting conversation according to your project's memory and retention settings.

Illustrative request

The following shows an OpenAI-compatible generation pattern for a proposed integration. Your assigned production endpoint and model identifier are confirmed during onboarding.

curl "$AMORA_API_BASE_URL/v3/chat/completions" \
  -H "Authorization: Bearer $AMORA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_ASSIGNED_MODEL",
    "messages": [
      {"role": "system", "content": "Use the approved character configuration."},
      {"role": "user", "content": "How has your day been?"}
    ],
    "stream": true
  }'

This example is illustrative and is not a currently executable Amora endpoint guarantee.

Character and knowledge context

Keep the character's identity and behavior in configuration owned by your project. Add selected knowledge and memory context on the server. Maintain clear boundaries between different users, characters and customer workspaces.

Streaming responses

Plan for connection interruption, partial responses and user cancellation. Preserve the response accumulated so far and distinguish incomplete output from a fully completed message. Do not automatically repeat a paid generation when its outcome is unknown.

Usage

Token accounting, model availability and any fallback behavior are part of your enabled service configuration. Agree those limits before publishing an allowance to your customers.