Closed
Description
What happened?
I've been using Perplexity API through LiteLLM proxy for a few months already. Specifically, I was using Sonar Pro model in the next way:
headers = {
"Authorization": f"Bearer {OPENAI_API_KEY}",
"Content-Type": "application/json"
}
payload = {
"model": model,
"messages": messages,
"temperature": temperature,
"web_search_options": {"search_context_size": search_context_size},
**kwargs
}
async with httpx.AsyncClient() as client:
response = await client.post(
OPENAI_BASE_URL + "chat/completions",
headers=headers,
json=payload,
timeout=request_timeout # seconds
)
if response.status_code != 200:
raise Exception(f"Perplexity API request failed with status {response.status_code}: {response.text}")
res = response.json()Recently, I've updated LiteLLM to the latest stable release version, and started getting this error (added full log in the corresponding section):
perplexity does not support parameters: ['web_search_options'], for model=sonar-proI've been using web_search_options for at least last 2 months with older LiteLLM version, and all was working perfectly fine. Now, after update, it started failing, forcing me to drop this param, after which API seems work again.
According to PPLX API docs, this parameter is totally valid, and I've not seen any updates about it deprecation, so I tend to believe the issue is on LiteLLM side.
Would appreciate any help here!
Relevant log output
LiteLLM Router: INFO: router. py:1131 - Litellm. acompletion(model=perplexity/sonar-pro) Exception litellm. UnsupportedParamsError: perplexity does not support parameters: ['web_search_options'], for model=sonar-pro. To drop these, set 'litellm.drop_params=True' or for proxy:
'litellm_settings:
drop_params: true'
If you want to use these params dynamically send allowed_openai_params=['web_search_options'] in your request.Are you a ML Ops Team?
No
What LiteLLM version are you on ?
v1.72.6.post1
Twitter / LinkedIn details
No response