Helicone Community Page

Hey all, new to Helicone coming from arize, I run a lot of models with Ollama and I was wondering what the recommended setup is because I followed to docs to use the SDK and it seems like a lot of things aren't properly wired up there for Ollama? Is it recommended to use the OpenAI API interface and pretend that Ollama isn't Ollama, or what is the intended setup for Ollama that will properly capture statistics?
TypeError: Cannot read properties of undefined (reading 'map')
on the Experiments page.
Hi all, I am attempting to test out the prompt tracking in Helicone and am unsure if I have come across an issue. I am using OpenAI' asssitants and have utilized hpf tags to handle prompts. Only thing is when I go to the prompts page, I don't see the text of my prompt, only a preview of a prompt variable. You can see in the attached image, what my view looks like. Any thoughts or pointers on this? Thanks! Here is the code snippet where I am sending through my prompt.
is there any way to store user feedback or create an eval on a request that is neither binary nor a score, but rather qualitative i.e. the "expected" model output?
7 comments
J
d
heya, currently trying to see if helicone would work for us. want to know if aws bedrock + google generative ai are supported via the vercel ai sdk
2 comments
o
Cant log in via github

I reported this issue half a year ago and was dismissed that it is a github issue, but github is fine.

Forwarded

I receive: {"code":400,"error_code":"validation_failed","msg":"Unsupported provider: provider is not enabled"}

"Cant log in with github" • 28/06/2024
i want to use geminin models through helicone llama-index what should be the API base and target URL setup
hi everyone
can we use vercel ai with async logging helicone feature? ?
2 comments
J
Getting random "KV PUT failed" failures, causing my app to fail at random times. And it's NOT during heavy loads from my app - you can check our logs, we're pretty boring quite users haha. Here's the message (our app does auto-retry 5x times):

Plain Text
(PID 2922) 2025-01-13T14:02:15.885Z  [ERROR] callChatModel.js:2868: Requesting https://oai.hconeai.com/v1/chat/completions hit HTTP error '500' (Request failed with status code 500) {"helicone-message":"Helicone ran into an error servicing your request: Error: KV PUT failed: 429 Too Many Requests","support":"Please reach out on our discord or email us at [email protected], we'd love to help!","helicone-error":"{}"}


I'm going to have to turn off routing request thru helicone because we cannot afford to have our users experience this "random crash" - but any way around this?
1 comment
J
I need help, I can't set an admin now because I accidentally set myself as member.
2 comments
m
J
Hi, I'm trying to set up helicone to work with litellm. Have the following code:

Plain Text
resp = completion(
        model="azure/gpt-4o-mini",
        messages=[
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": user_prompt}
        ],
        response_format=PoliticalBiasScore,
        fallbacks=["gpt-4o-mini"],
        user="varsity_topstories_scorearticlepoliticalbias",
        metadata={
            "user": "varsity_topstories_scorearticlepoliticalbias",
            "Helicone-Property-User": "varsity_topstories_scorearticlepoliticalbias"
        }
    )


which seems to be hit or miss on if it tags this to a user at all. What is the best way to do this - thank you!
4 comments
J
d
Hi, I am trying out the Pro mode and added the prompt add on. But I don't see a prompts ui. Is this expected?
2 comments
C
Very long delay between messages sent and messages appearing in requests.
2 comments
C
Happy New Year, Everyone
Can we use our AWS Bedrock integration or any OSS models with the playground, or only the closed models shown in the Demo?
1 comment
J
Hey
I can do it for vercel ai sdk + groq with the following code:
Plain Text
createGroq({
    apiKey,
    baseURL: "https://groq.helicone.ai/openai/v1",
    headers: {
        "Helicone-Auth": `Bearer ${process.env.PLASMO_PUBLIC_HELICONE_API_KEY}`,
    },
});

so, my questions are:
  1. is it possible to do the similar thing for vercel ai sdk + gemini.
  2. if yes, then what's the baseURL for gemini?
thanks
6 comments
H
J
Editing prompts doesn't work, you can't add post types, also at creation time you can't add.
1 comment
J
If you delete a prompt and create a new one with the same name, the prompt doesn't work "Failed to get compiled prompt versions"
1 comment
J
Helicone has been great to help us get end of year stats for our usage so thanks, really useful tool! Checking out your own Open Stats and out of curiosity what happened with OpenAI on 5th Dec this year?
Error 400 when calling Get User Data in https://docs.helicone.ai/rest/user/post-v1userquery

I tried both in JS and Postman.
3 comments
L
Hey all! Happy users of Helicone here 🙂

We recently launched Martin AI on Product Hunt and we're super close to getting #1 Product of the Day (within 10 upvotes)! Would really appreciate if y'all have some time to spare to share our product some love and help us get number 1!

https://www.producthunt.com/posts/martin-3

Thanks!
Ethan
1 comment
J
Info:- SAP Stocks , a platform for traders to get current market sentiment wrt desired tickers alongside a visualised trend of latest closing prices and a chatbot to query the news

Hello everyone,this is a small demo of my current project

https://x.com/aspalve2004/status/1859443976400482744


Linkedin post with demo video

https://www.linkedin.com/posts/aaryan-palve-6124372b7_aiinfinance-sentimentanalysis-stockmarketinsights-activity-7264979310308937728-sb80?utm_source=share&utm_medium=member_desktop

Looking forward to hearing some feedback and additional features
Hey @Cole, just signed up to try the pro tier -- I'm looking to create prompts from the web dashboard, but this window appears when I try to do so as if my account wasn't pro. Any idea what's going on here? Thanks for your help
4 comments
C
g
And right now I'm in free trial.