Creating Endpoints

Fill in the Overview

Enter the endpoint Name, URL, HTTP method, protocol (REST or WebSocket), project, and environment.

Environment tags the endpoint as Development, Staging, or Production — useful for keeping configurations separate and quickly identifying production-critical endpoints on the grid.

Disable tracing — when on, invocations skip trace and telemetry collection. Useful when the endpoint already emits its own telemetry. Tracing is enabled by default.

Set up authentication

In the Headers tab, set your API token and any custom request headers.

The token is stored securely and available as {{ auth_token }} in your request template. If you set a token without an explicit Authorization header, Rhesis adds Authorization: Bearer <token> automatically.

Two authentication types are supported:

• Bearer token — paste your API key or token directly
• Client credentials (OAuth) — provide client_id, client_secret, token_url, and scopes; Rhesis fetches and refreshes the token automatically

Write the request and response mapping

In the Mapping tab, define how Rhesis formats requests to your API and extracts values from the response.

Request body — write a Jinja2 template that maps Rhesis variables to the fields your API expects. At minimum, place {{ input }} wherever your API expects the user’s message:

{
  "model": "gpt-4",
  "messages": [
    {
      "role": "user",
      "content": "{{ input }}"
    }
  ]
}

The editor provides clickable variable chips organized by category — click any chip to insert it at the cursor position.

For the full list of available variables and filters, see Request Mapping.

Response — run a test request, then click any key in the JSON response tree to map it to a Rhesis output variable (output, context, metadata, etc.). The mapping is saved automatically.

For response variable details and extraction syntax, see Response Mapping.

Verify with a connection test

In the Connection Test tab, fire a live request to confirm everything works end-to-end before saving.

Fill in the input variable rows on the left, then click Run test. You’ll see the rendered request preview alongside the raw API response and the extracted mapped output.

Save

Click Create endpoint to save.

Fill in the Overview and Headers tabs

Provide the endpoint Name, URL, and API Token before running Auto-Configure.

Click Auto-Configure

The Auto-configure button (magic wand icon) becomes active once the URL is filled in.

Paste reference material

Paste any of the following — the AI accepts multiple formats:

curl command (most reliable):

curl -X POST https://api.example.com/v1/chat/completions \
-H "Authorization: Bearer sk-..." \
-H "Content-Type: application/json" \
-d '{
    "model": "gpt-4",
    "messages": [{"role": "user", "content": "Hello"}]
}'

Python code (requests call or Flask/FastAPI handler):

response = requests.post(
    "https://api.example.com/chat",
    json={"query": "Hello", "model": "gpt-4"},
    headers={"Authorization": "Bearer sk-..."}
)
print(response.json()["response"]["text"])

Sample request/response JSON:

// Request
{"messages": [{"role": "user", "content": "What is AI?"}], "model": "gpt-4"}

// Response
{"choices": [{"message": {"content": "AI is..."}}]}

Plain text API docs — any description of the endpoint’s expected fields and response structure also works.

Run and review

Click Auto-configure. Behind the scenes Rhesis runs a multi-step pipeline:

1. Parse — AI analyzes your input to identify the URL, method, request fields, and response structure
2. Probe (optional) — Rhesis sends a live request to capture the real response format, retrying up to 3 times if needed
3. Self-correct — if the probe fails, AI analyzes the error and adjusts the request before retrying
4. Generate — using the confirmed schema, AI produces Jinja2 request templates and JSONPath response mappings

Review the confidence level and any warnings before applying:

• High (70%+) — verified via live probe
• Medium (40–70%) — generated, may need minor adjustments
• Low (below 40%) — best guess, review manually

Click Show probe response to see the raw JSON your endpoint returned — useful for verifying the response mapping extracts the right fields.

Apply and verify

Click Apply to Endpoint to populate the Mapping tab, then switch to Connection Test to verify end-to-end.

If only some fields were mapped, apply the partial result as a starting point and fill in the missing fields manually. Even partial results save significant time compared to configuring from scratch.