BerriAI/litellm · error · OCIError
param `{key}` is not supported on OCI
Error message
param `{key}` is not supported on OCI What it means
Error "param `{key}` is not supported on OCI" thrown in BerriAI/litellm.
Source
Thrown at litellm/llms/oci/chat/transformation.py:357
for key, value in {**non_default_params, **optional_params}.items():
alias = param_map.get(key)
if alias is False:
# max_retries is a litellm-level control param (litellm applies
# retries itself); it is never a generation param OCI accepts, so
# drop it silently. The litellm proxy injects it on every request,
# which otherwise 500s OCI calls unless drop_params is set.
if key == "max_retries":
continue
# n=1 (or None) is the OpenAI default: a single generation, which
# every OCI model produces anyway. Drop it silently so standard
# clients that always send n=1 (e.g. the MLflow gateway) are not
# rejected; only n>1 is genuinely unsupported on Cohere, which
# has no numGenerations field.
if key == "n" and (value is None or value == 1):
continue
if drop_params or litellm.drop_params:
continue
raise OCIError(
status_code=400,
message=f"param `{key}` is not supported on OCI",
)
if alias is None:
adapted_params[key] = value
continue
adapted_params[alias] = value
# Preserve the original OpenAI ``response_format`` key alongside the
# OCI-mapped ``responseFormat`` so downstream litellm framework code
# (e.g. ``json_mode`` detection, logging) that inspects
# ``optional_params["response_format"]`` continues to work.
if alias == "responseFormat":
adapted_params["response_format"] = value
return adapted_params
def sign_request(
self,View on GitHub (pinned to 6c2dcb801b)
Solutions
- Remove the unsupported parameter for OCI or set drop_params=True.
When it happens
Trigger: Thrown at litellm/llms/oci/chat/transformation.py:357 when the library encounters an invalid state.
Common situations: See trigger scenarios.
AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15).
Data as JSON: /api/errors/dd4a69240ecde26c.
Report an issue: GitHub.