Default Insurance Chatbot (Rosalind)
Rosalind is a pre-configured insurance chatbot powered by Google’s Gemini, available to all new users as a demo endpoint during onboarding. It lets you generate tests, run evaluations, and explore results before connecting your own application.
Chat Endpoint
POST /chat
Response:
Parameters:
message(string, required): Your question or message to Rosalindsession_id(string, optional): Identifier for conversation continuity. If omitted, a new session is createduse_case(string, optional): The domain context. Defaults to “insurance”
Requests may include a bearer token in the Authorization: Bearer <your-api-key> header; authenticated requests get a higher rate limit. Requests without a token fall back to the public tier.
Echo use case
Setting "use_case": "echo" returns the input message verbatim, skips LLM invocation, and is exempt from rate limits. Useful for connector smoke tests, pipeline checks, and request-mapping validation.
Sessions
Reuse the session_id from a response in follow-up requests to keep conversation context across turns.
- GET
/sessions/{session_id}— retrieve the conversation history for a session - DELETE
/sessions/{session_id}— delete a session and its history
Rate Limits
Without authentication, the service allows 100 requests per day per IP address; with authentication, 1000 per day per user (configurable via CHATBOT_RATE_LIMIT). Limits reset on a fixed daily window; exceeding them returns 429 Too Many Requests.
Example
Parameter Awareness
Rosalind is a parameter-aware endpoint: its system prompt, use case, model, temperature, and output mode are managed through Parameters & Experiments. Create a new experiment in the project and promote it to the default environment to see the chatbot pick up the new behavior after the next cache refresh. For how experiment values reach the request mapping, see Using Experiment Parameters.
Limitations
- Domain-Specific: Rosalind answers insurance questions only and may decline off-topic queries
- Rate Limited: Usage is restricted to prevent abuse and ensure availability
- Session Expiry: Sessions expire after 24 hours of inactivity (default)
- No SLA: This is a demo service without uptime guarantees
Next Steps - Create custom endpoints for your own AI services - Explore test generation to validate your LLM applications - Set up metrics to measure response quality