Managing Endpoints
Testing an Endpoint
Open an endpoint and go to the Test tab to send a live request and verify your configuration before running full test suites.
Rhesis reads your request body template and generates a fill-in row for each detected variable (input, files, image, etc.). Variables of type files or image show a file upload input instead of a text field.
Click Test to send the request. A collapsible request preview shows the fully-rendered HTTP request — method, URL, headers, and body — with auth values masked.
After a successful response, the JSON tree is clickable. Click any key to map it to an output variable (output, context, metadata, etc.). The mapping is saved to the endpoint’s response mapping automatically.
The response mapping you build here is the same one used during test runs. You can also edit it directly in the Overview tab.
Endpoint Detail Tabs
| Tab | Description |
|---|---|
| Overview | Name, URL, protocol, method, request body template, and response mapping |
| Headers | Authentication token and custom request headers |
| Test | Interactive test-and-map workbench |
| Test Runs | History of test runs executed against this endpoint |
Exploring an Endpoint Programmatically
External agents and API clients can launch an async Penelope exploration with POST /endpoints/{endpoint_id}/explore. Use this when you want endpoint capability discovery outside the web Architect chat — for example from an MCP client or automation script.
The route validates that the endpoint belongs to the current tenant, starts a Celery task, and returns a task_id. Poll GET /jobs/{task_id} until the job reaches SUCCESS; the result contains the exploration findings, conversation summary, strategy, duration, and endpoint ID.
Then poll the returned job:
Exploration request fields
Provide either strategy or a non-empty goal.
| Field | Required | Description |
|---|---|---|
strategy | No | One of domain_probing, capability_mapping, boundary_discovery, or comprehensive. When present, Penelope generates the goal and instructions for that strategy. |
goal | Conditional | What to learn about the endpoint. Required when strategy is omitted. |
instructions | No | Step-by-step probing guidance for Penelope. |
scenario | No | Persona or situational context to use while probing. |
restrictions | No | Constraints or forbidden behaviors to verify during exploration. |
previous_findings | No | Structured findings from a prior run, used to continue or deepen exploration. |
The same exploration strategies are available through the SDK ExploreEndpointTool; see SDK Agents.
Deleting Endpoints
Select one or more endpoints from the grid and click Delete.
Deleting an endpoint does not delete associated test configurations or historical test results. Your test data remains intact, but you cannot execute new tests against a deleted endpoint.
Using Endpoints in Tests
Executing Test Sets
Select an endpoint, configure the execution mode (parallel or sequential), and click Run Tests. Rhesis sends each test prompt through the endpoint and evaluates responses against your metrics.
Multiple Endpoints
Run identical tests against different endpoints to compare models, environments, or configurations. Each test run is independent, so differences in behavior and quality are easy to analyze side by side.
Disabling Tracing
Each endpoint has a Disable tracing toggle available in both the creation form and the detail page. When on, invocations skip trace and telemetry collection. Tracing is enabled by default.
Useful when the endpoint already emits its own telemetry or when tracing overhead is unwanted.
Environment Management
Organize endpoints by environment to match your deployment workflow.
- Development — Local or dev servers. Safe for rapid iteration and configuration changes.
- Staging — Pre-production systems for validation and integration testing before promotion.
- Production — Live production APIs. Use for regression testing and quality monitoring; modify with care.
Environment tags are shown on the Endpoints grid so you can quickly spot production-critical endpoints.