Portkey error in the /completions api using new version

Getting error in the /completions API using new version

Plain Text

"{\"status\":\"failure\",\"message\":\"Portkey Error: You need to pass prompt_id along with virtual_key. Only virtual keys are not allowed in /v1/prompts config.\"}"

CC: @ayush-portkey @Vrushank | Portkey

26 comments

aayush-portkey

can you please share the config you are using?

ddeepanshu_11

@ayush-portkey here's the config

Plain Text

{
    "cache": {
        "mode": "semantic"
    },
    "retry": {
        "attempts": 3
    },
    "strategy": {
        "mode": "single"
    },
    "targets": [
        {
            "provider": "openai",
            "virtual_key": "open-xyz",
            "override_params": {
                "model": "gpt-4-1106-preview",
                "response_format": {
                    "type": "json_object"
                }
            }
        }
    ]
}

aayush-portkey

Thanks. And you are using https://api.portkey.ai/v1/completions endpoint right?

ddeepanshu_11

yeah right, saw in the Config console and Dec migration document that it would be auto-migrated

ddeepanshu_11

was working fine till 27th Dec and now breaking

aayush-portkey

This is migrated new format only

vvisarg

Hey @deepanshu_11 , are you using prompt completions route?

ddeepanshu_11

https://api.portkey.ai/v1/prompts/PROMPT_ID/completions

ddeepanshu_11

Yes @visarg @ayush-portkey

vvisarg

Can you add the base prompt_id to all the blocks which has virtual_key in it.

Plain Text

{
    "cache": {
        "mode": "semantic"
    },
    "retry": {
        "attempts": 3
    },
    "strategy": {
        "mode": "single"
    },
    "targets": [
        {
            "provider": "openai",
            "prompt_id": "prompt_id",
            "virtual_key": "open-xyz",
            "override_params": {
                "model": "gpt-4-1106-preview",
                "response_format": {
                    "type": "json_object"
                }
            }
        }
    ]
}

vvisarg

Something like this

vvisarg

base prompt id = prompt id that you send in url.

ddeepanshu_11

okay so for eg: if I have multiple prompts that are using the same config, how to add all prompt_id (s)?

vvisarg

Current way is to create multiple configs. But let me fix that behaviour as it become an overhead for you. I will prefill configs on-the-fly from the request URL,

ddeepanshu_11

Sure thanks

vvisarg

Quick update on this: We have finalized the changes. It should be pushed in a hour

ddeepanshu_11

Awesome, thanks for the update

vvisarg

This is done. Can you please check and let us know?

ddeepanshu_11

Sure lemme check it soon and get back

ddeepanshu_11

What should I expect though?

ddeepanshu_11

*to be changed

ddeepanshu_11

@visarg

ddeepanshu_11

Dont see any prompt ids automatically added to the config

vvisarg

Nothing gets changes on the UI. The config will be as it is. You can use it with multiple prompts. What we did is to add prompt_id from URLs on the fly. So if you do not add any prompt_id in config, we will pick it up from this URL that you use to make the call.

https://api.portkey.ai/v1/prompts/PROMPT_ID/completions

vvisarg

But you should not get this error anymore:

Plain Text

"{\"status\":\"failure\",\"message\":\"Portkey Error: You need to pass prompt_id along with virtual_key. Only virtual keys are not allowed in /v1/prompts config.\"}"

ddeepanshu_11

Amazing, great user experience, loved this!

Add a reply

Welcome to Portkey Forum

Portkey error in the /completions api using new version