BerriAI/litellm · error · Exception

Only POST requests are supported

Error message

Only POST requests are supported

What it means

litellm's vendored DeepEval/Confident-AI Api client implements only POST: _http_request hard-fails on any other method string before any network I/O. It exists as a guard against porting call sites from the official deepeval SDK, where GET/PUT/DELETE are supported for some endpoints.

Source

Thrown at litellm/integrations/deepeval/api.py:68

    BASELINE_ATTACKS_ENDPOINT = "/generate-baseline-attacks"


class Api:
    def __init__(self, api_key: str, base_url=None):
        self.api_key = api_key
        self._headers = {
            "Content-Type": "application/json",
            # "User-Agent": "Python/Requests",
            "CONFIDENT_API_KEY": api_key,
        }
        # using the global non-eu variable for base url
        self.base_api_url = base_url or API_BASE_URL
        self.sync_http_handler = HTTPHandler()
        self.async_http_handler = get_async_httpx_client(llm_provider=httpxSpecialProvider.LoggingCallback)

    def _http_request(self, method: str, url: str, headers=None, json=None, params=None):
        if method != "POST":
            raise Exception("Only POST requests are supported")
        try:
            self.sync_http_handler.post(
                url=url,
                headers=headers,
                json=json,
                params=params,
            )
        except httpx.HTTPStatusError as e:
            raise Exception(f"DeepEval logging error: {e.response.text}")
        except Exception as e:
            raise e

    def send_request(self, method: HttpMethods, endpoint: Endpoints, body=None, params=None):
        url: Final = f"{self.base_api_url}{endpoint.value}"
        res: Final = self._http_request(
            method=method.value,
            url=url,
            headers=self._headers,

View on GitHub (pinned to 77b7c6c40c)

Solutions

1. Use HttpMethods.POST for every send_request call through this vendored client
2. If you need GET/PUT/DELETE against Confident AI, use the official deepeval SDK (pip install deepeval) or a plain httpx client instead

Example fix

# before
api.send_request(HttpMethods.GET, Endpoints.PROMPT_ENDPOINT, params={"id": p})  # Exception

# after (official SDK for reads)
from deepeval.confident_api import ...  # or plain httpx GET
# vendored client stays POST-only:
api.send_request(HttpMethods.POST, Endpoints.TRACING_ENDPOINT, body=trace)

from litellm.integrations.deepeval.api import HttpMethods

def assert_post(method: HttpMethods) -> None:
    if method is not HttpMethods.POST:
        raise ValueError(f"vendored DeepEval client is POST-only, got {method}")

from litellm.integrations.deepeval.api import HttpMethods
from typing import Any

def is_supported_method(method: Any) -> bool:
    """Narrow to the only verb the vendored client implements."""
    return method is HttpMethods.POST

When it happens

Trigger: Calling api.send_request(HttpMethods.GET, Endpoints.PROMPT_ENDPOINT, ...) or any non-POST verb; code migrated from the official confident-ai/deepeval client that fetches or deletes resources.

Common situations: Developers assuming the vendored client mirrors deepeval's full REST surface; automated call-site refactors switching the HttpMethods enum value.

Related errors

• DeepEval logging error: {e.response.text}
• res.json().get("error", res.text)
• Please set 'CONFIDENT_API_KEY=<>' in your environment variab
• Invalid environment: {environment}. Please use one of the fo
• BedrockFilesConfig does not support file retrieval

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/9c58c95286d5a55a. Report an issue: GitHub.