search_domain_filter doesn't seem to work

[jcmag]

Hi,

I'm trying to use the 'search_domain_filter' setting, this way:

curl --request POST
--url https://api.perplexity.ai/chat/completions
--header 'Authorization: Bearer .......'
--header 'Content-Type: application/json'
--data '{
"messages": [
{
"content": "How much does the Feed Reader bot cost?",
"role": "user"
}
],
"model": "sonar",
"search_domain_filter": [
"tfrbot.com"
]
}'

but the answer gives citations for other domains than tfrbot.com:

{
"id": "10b54496-242f-4680-8b97-3862f522d7a5",
"model": "sonar",
"created": 1737536367,
"usage": {
"prompt_tokens": 9,
"completion_tokens": 192,
"total_tokens": 201
},
"citations": [
"https://thefeedreaderbot.com/premium.html",
"https://rss.app/en/pricing",
"https://thefeedreaderbot.com",
"https://zapier.com/blog/best-rss-feed-reader-apps/",
"https://monitorss.xyz"
],
"object": "chat.completion",
"choices": [
{
"index": 0,
"finish_reason": "stop",
"message": {
"role": "assistant",
"content": "The Feed Reader Bot offers different subscription plans with varying costs:\n\n- Free Account: This is the basic plan with no cost, but it has limitations such as a refresh rate of 4 hours and support for up to 10 feeds and 5 channels/groups[1].\n- Premium Account: Costs \u20ac5 per month or a lifetime subscription of \u20ac150. It offers a faster refresh rate of 30 minutes, supports up to 50 feeds, and includes additional features like custom buttons and translation[1].\n- Elite Account: Costs \u20ac10 per month or a lifetime subscription of \u20ac300. It provides the fastest refresh rate of 10 minutes, supports up to 100 feeds, and includes advanced features like AI-powered summaries and custom bot usage[1].\n\nAdditionally, for more frequent refresh rates (e.g., 2 minutes, 1 minute), the cost increases significantly, ranging from \u20ac10 to \u20ac120 per month per feed[1]."
},
"delta": {
"role": "assistant",
"content": ""
}
}
]
}

Am I doing something wrong?

[AidanShipperley]

I just looked at their usage tier documentation, it looks like they implemented a tiered system with the release of "Sonar" models, similar to OpenAI's tiered API.

To use search domain filters, you need to be in tier 3, which means you have to have to have paid at least $500 to their API (all time) to get access.

I'm glad the feature is released, because it's been in a closed beta for over a year, but they should really put a note for limited parameters in their API reference page, which doesn't indicate any restriction to the parameter.

More so, they should really trigger a failure in the API response for parameters that you don't have access to, because I just assumed this parameter was broken until I stumbled into their usage page. I spent a couple hours struggling with structured outputs as well not realizing I didn't even have access.

[jcmag]

good catch! I better understand now

[ak222002]

So what does this all time mean ?
so what i can understand is by this term all time is the credit we have spend for all previous fro day we go the perplexity api ?
For example : i started using api a year back does that mean the total money in this year ?