Helicone
One row per request: the model, the upstream provider, the token counts, the latency, the cost Helicone computed and the properties you set. The query asks for inputs to be left out, so no prompt and no response is returned.
Access
- Kind
- A read-only key you paste
- What to give
- Read-only API key
- Why that one
- A Helicone API key. Valistry calls only the request query endpoint, a read, and asks it not to return prompts or responses.
What it attributes on its own
- From the source
- request · model · upstream provider · custom properties · end user (hashed) · latency · what Helicone computed the request cost
- Needs a rule or a trace
- the provider's invoice — Helicone computes a cost, it does not bill you for inference · outcome unless a custom property carries it
The addresses it calls
- https://api.helicone.ai/v1/request/query
Only these. A call anywhere else is refused before it leaves, and every call is logged.
Connecting it
- Open your Helicone API keys. Sign in as an owner. EU accounts use eu.helicone.ai. Open ↗
- Generate a key for Valistry. Name it so you recognise it in your own audit later.
- Choose your region and click Test and connect. Valistry reads the last seven days, reports how many requests it found, then fills thirty.
Helicone computes a cost; the provider bills it. Connecting both Helicone and a provider behind it counts the same request twice — connect one. Helicone is the better of the two when you need per-property or per-user attribution; the provider is better when you need the invoice.